Introduction Ray Jackendoff , Paul Bloom ,

and Karen Wynn

It is with a deep sense of gratitude and humility that we offer this collec -
tion of essaysin memory of John Macnamara by his students, colleagues,
and friends . The essays speak eloquently to the intellectual influence he
had on all of us, but it is seemly that we begin with John himself .
John grew up in County Limerick , Ireland , and spent the early part of
his adult life as a member of the Vincenti an community . He taught at
Castleknock and at St. Patrick 's College , Drumcondra . In 1963 he gained
a doctorate from the University of Edinburgh . His controversial research
findings cast doubt on the value of some of the educational priorities of
postindependenceIreland; they were published in a book entitled Bilin-
gualism and Primary Education . In 1966 John moved to Canada , where he
became Professor of Psychology at McGill University , a post he held till
the end of his life in 1996 .
John ' s intellectual impact in the period after he came to McGill was as
a scholar in the study of the mind and how it develops. Before reviewing
some of the contributions he made in different areas of psychology , it is
important to understand his broader intellectual commitments . These
pervade his work and are themselves interesting and important .
There is a tendency for psychologists to conclude that the human mind
is really quite simple . Psychologists influenced by the ideas of Watson and
Skinner have long argued that thinking and learning can be explained
through a few basic laws. Some reductionists believe that psychological
phenomena can be entirely accounted for in terms of principles of biology

We are grateful to John 's wife Joyce for her assistance in writing this introduction .
Some paragraphs are adapted from a previously published essay by Paul Bloom .
We also wish to thank Albert Bregman for his assistance and encouragement in
the initial stages of developing this volume .
 x
Jackendoff, Bloom, and Wynn

and chemistry . Many contemporary scholars are entranced with the idea
that the mind works like a computer (according to some, a rather un-
complicated computer ) . And surely , it is often said, we only need a simple
theory to account for the inner workings of a child - who is, after all , a
quite simple creature !
In John 's writings , one finds the sometimes unfashionable conviction
that the mind of the child is extraordinarily rich and complex . He begins
his 1982 book Names for Things by noting that psychologists typically
ignore the complexity of language learning - at the cost of not being able
to explain it . His favorite example , as will be seen throughout this vol -
ume, was that something as apparently simple as learning the name of a
dog requires considerable mental resources. For instance , children must
be able to appreciate the intentions of others . The 2-year -old must realize
that when an adult uses the name " Freddie " in certain contexts , he or she
intends it to refer to the dog and nothing else. Children must also possess
certain logical resources. The child has to understand how this new word
" Freddie " relates to other parts of speech, such as the common noun
" dog " and the adjective " brown ," and how it contributes to the meanings
of the sentencesthat it appearsin . Perhapsmost importantly , children can
use words to refer : to a real animal for a proper name like " Freddie , " to
an abstract kind for a common noun like " dog ," and to different kinds of
entities altogether when it comes to words like " two " and " Santa Claus ."
John argued that explaining the child 's learning and understanding of
language requires a psychology that includes notions such as inten -
tionality , reference, and truth . This means that psychology and philoso -
phy are more related than many would have thought . It also means that a
complete theory of the mind cannot be found in the fields of biology and
computer science , as such sciences cannot capture these essential proper -
ties of our mental life .

The implications of this perspective can be pursued through three re-

lated aspects of John 's research: his studies of language learning , his con -
tributions to the debate over the nature of child development , and his
work on a psychology of ideals- only a minority of his writings . In all the
best ways , John had the style of a nineteenth -century intellectual . He
wrote about an extraordinary variety of topics , from the nature of free
will to the demise of Freudian psychoanalysis to what formal logic has to
say' about the Holy Trinity . He also did not shy away from disagreement ,
and he was involved in productive and civil debate with scholars such as
the philosopher Mario Bunge and the linguist N oam Chomsky .
Introduction Xl.

Finally , he was one of the very few psychologistswho isn't a chore to

read . He wrote with style and wit , and most of his work was directed
to the educated public ; it was not crafted for a small group of fellow
scholars .

In 1972 John published a classic article with the title " The Cognitive
Basis of Language Learning in Infants ." This article presented a proposal
that has grown to be highly influential . John suggested that although
there are important regards in which language learning is special, distinct
from other sorts of learning that children succeed at, it also relies crucially
on the child 's general understanding of the situations in which sentences
are used. He proposed that the child 's learning of basic grammar (what is
a noun and what is a verb , how words are ordered within a sentence , and

so on) requires knowing what words and sentences mean , and how they
correspond to the external world . In support of this perspective , John and
two colleagues, Nancy Katz and Erika Baker , wrote their 1974 article ,
" What 's in a Name ? A Study of How Children Learn Common and
Proper Names ," which reported an amazing finding . Their experiments
found that even 18-month -olds can use subtle linguistic cues, such as the
grammatical difference between " This is wug " and " This is a wug ," to
learn proper names and common nouns . These two articles have had a
substantial influence on the study of language , motivating considerable
research seeking to better clarify the relationship between children 's
understanding of meaning and their learning of basic grammar .
After the publication of these articles , most of John 's research into
language focused on the learning of words . This work was first discussed
at length in Names for Things and was subsequently elaborated in many
articles and in two further books , A Border Dispute .' The Place of Logic in
Psychology ( 1986) and The Logical Foundations of Cognition ( 1994, edited
in collaboration with Gonzalo Reyes) . This research, some in collabora -
tion with Gonza10 Reyes and Marie La Palme Reyes, focused on the
logical foundation of the learning of proper names and common nouns .
Consider again what a child must know in order to understand a proper
name . One aspect of this knowledge is that this word follows a single
object over time . Even if Freddie were to change color or lose a leg, he
would still be Freddie . On the other hand , if he were to have a twin

brother that looked just like him , this brother would not be Freddie . The
name " Freddie " does not pick out all objects with a certain appearance; it
picks out a unique dog , regardless of his appearance . The child must also
have some understanding of what changes dogs can go through and still
 ..
XII Jackendoff, Bloom, and Wynn

remain dogs, as well as what eventswould causesomethingto ceasebeing

a dog. More generally, John and his colleaguesarguedthat understanding
a proper name requiresan appreciation of the kind that the named object
belongsto.
At somelevel, this is all common sense.Who would doubt that Freddie
would still be Freddie if he lost a leg, or that the word " Freddie" refers to
the whole dog and not just to the head or the tail? This appreciation of the
logic of namescomesso naturally to us that it is easyto underestimatethe
richnessof the logical resourcesrequired for this understanding. John and
his colleagues developed an extensive theory of these resources and
argued further that much of the knowledge that underlies the learning of
proper namesand common nouns cannot itself be learned. Children must
be born with it .
This research has implications for a broader debate within develop-
mental psychology. Jean Piaget and his followers argued that the minds
of children are very different from those of adults. Whereas adults are
capable of complicated logical thought, children are limited to thinking
about specific events; they are incapable of abstraction. Adults can take
the perspectiveof others, but children are egocentricand believethat their
own perspectiveon the world is all there is. An adult is a moral being, but
a child is a little Attila the Hun, lacking any real understandingof fairness
or morality .
John was a central figure in this debate, arguing that this conception of
developmentis mistaken. This is not to deny that children know a great
deal lessthan adults or that they neednurturing and care in order for their
moral and intellectual capacitiesto thrive. But John argued that children
possessconsiderableresourcesto start with . To seethis, one must explore
(through logical analysis, supplementedwith careful experiments) the
sorts of things children are very good at. The learning of proper names
has beendiscussedabove; other examplesinclude children's knowledge of
logical terms like " some," their ability to understand and manipulate
small numbers, their understandingof mental notions like pretending and
forgetting, and the rich moral knowledge that is demonstrated in their
everydayinteractions with their parents and siblings.
F or each of thesedomains, John's researchposesthe samechallengeto
developmental psychology. If children are so limited in their capacities,
how is it that they know so much? If they have no logic, how can they
successfullylearn expressionsof language that depend so crucially on
logical notions? If they are egocentric, how is it that they learn the
 ...
Introduction Xlll

meanings of pronouns like " I " and " me," since learning these pronouns
requires taking the perspectivesof others? If they have no conception
of fairness or morality , how is it that they understand a fairy tale like
" Cinderella ," which is impossible to appreciate without understanding
that the stepmother is acting unfairly ? All the evidence, John argued ,
leads to the conclusion that children possess a rich set of cognitive , logi -
cal, and moral capacities from the very start .
John also presented more general arguments against the view that
children lack an appreciation of logic and morality . He noted that psy-
chologists fond of this view have never explained how children come
to acquire such capacities . Followers of Jean Piaget have appealed to
" assimilation " and " accommodation ," processes through which the
child 's primitive knowledge becomes more abstract through interaction
with the environment . But in a 1976 article entitled " Stomachs Assimilate
and Accommodate , Don 't They ?" , John argued that these notions are
empty metaphors that explain nothing . Along with the work of Jerry
F odor and N oam Chomsky , John 's empirical and theoretical defense of
the notion of innate ideas and capacities caused a major shift in the way
psychologiststhink about the mind and its development.
The final aspect of John 's work that we wish to mention is his proposal ,
first outlined in a 1990 article entitled " Ideals and Psychology ," that an
adequate psychology must explain the human capacity to understand
ideals . A good example of this , discussed first by Descartes, lies in the
domain of geometry . We all possess the notion of perfect geometric
forms , such as a perfect triangle or a perfect point , but no such forms exist
in nature and none ever will . Scientific progress is also based on ideals ;

physics would be nowhere without its frictionless plane and its perfect
vacuum . Similarly , we have personal ideals that we aspire to , notions such
as humility , friendship , and courage . As John put it , " Idealization is as
natural to the mind as breathing to the body ." But how is it that we come
to appreciate such ideals, given that , by their very nature , they are not to
be found in our experience?
In a 1991 article called " The Development of Moral Reasoning and the
Foundations of Geometry ," John provided a hint of what a psychology of
ideals would look like . The article begins with a critical discussion of a
well -known stage theory of moral development proposed by Jean Piaget
and extended by Lawrence Kohlberg . John argued that this theory seri-
ously underestimates children ' s knowledge , fails to describe adult moral
competence, and does not explain how notions of morality actually de-
 .
XIV Jackendoff, Bloom, and Wynn

velop and how supposedlyamoral children grow into moral adults. John
then proposed an alternative theory of moral development, based on an
intriguing analogy betweensystemsof morality and systemsof geometry.
In both domains, there exist ideal elements(in geometry, elementssuch as
point and line; in morality , elements such as fair and good) that are
related to one another through a system of axioms. John suggested
that there are extensive unlearned elements within both geometry and
morality . Children start off with an understanding of certain ideals in
both domains, and these are the notions that all people share. But psy-
chologists must also explain how new geometric and moral knowledge
emergesin adults and children. New geometries have been developed
by mathematicians, such as the non-Euclidean geometry proposed by
Riemann in the nineteenth century, and new moral ideals have been in-
troduced in the course of history, such as the ideal of chastity introduced
by the Christians. How are these learned and understood? The parallel
betweengeometry and morality placesthe study of moral developmentin
a striking new light . It might be that the child who is coming to grasp a
novel moral system is actually acquiring a coherent formal structure,
complete with ideals and axioms, in much the sameway he or shewould
learn a new systemof geometry, or mathematics, or physics.
This article is not as well known as many of John's other works dis-
cussedabove. (Perhaps the title scarespeople away!) But it is hard to
think of a better work in developmental psychology. Although it is
entirely accessiblein style and content, it doesnot shy away from the hard
questions. It is a significant intellectual accomplishment, but also some-
thing more, as it displays severalproperties typical of John's work . There
is creativity, intellectual courage, and a strong curiosity about the work-
ings of the world . Most of all, there is respectfor its subject matter: the
minds of young children and how they develop. It is more than an article
about morality ; it is itself a highly moral work .
For a closer look into John as a person, we quote from the eulogy by
John's friend and collaborator Gonzalo Reyes. The essencethat Gonzalo
so beautifully evokeshere strikes a familiar chord for anyone who knew
John.

I first met John ten yearsago whenhe gavea talk in the Mathematics
Departmentat McGill in October1985.He raisedthequestionof givinga
logicalaccountof thephrase" Freddieis a dog" that childrenlearnby the
Introduction xv

age of 18 months , claiming that logical theories were unable to account

for it . To tell you the truth , I thought that he was mad : how can there be
problems with that? But I found his talk so intriguing that I went to dis-
cuss it with him at the end. This was the beginning of a collaboration that
developed into a close friendship and resulted in one of the most beautiful
and rich periods of my life .
This talk at McGill was my first encounter with John 's approach to in -
tellectual matters : to ask basic simple questions to test theories or view -
points and to demand straightforward answers. Of course this Socratic
method enraged some people , but it delighted others , especially when he
used it masterfully in his talks . He had a robust , realistic , no -nonsense at-
titude and no patience for idealistic views that he felt did not go to the heart
of things . In this respect, he was like a brick , giving a senseof security to all
of us. In fact , John was a man of profound convictions : he was deeply
religious and deeply realist . Probably the thinker that was closest to his
heart was Aristotle . At the beginning of our collaboration , I asked what I
should read in psychology , and his answer was " Only De Anima ."
Another aspect of John 's approach to intellectual problems was the
unity of his thought . I discovered with bewilderment that behind the dog
Freddie stood none other than Plato , and behind the questions that
looked so trivial at the beginning stood in fact a formidable problem ,
Plato 's paradox of learning , which still haunts the venerable house of
philosophy . And the answer to these questions appeared to John as a
first step in creating a cognitive psychology that would be, as he put it ,
" worthy of its name ." His talks were fascinating , with flashes that allowed
us to catch a glimpse of this rich web of connections .
John died in a period that burst with creativity . He thought that the
time was ripe to study the relations between perception and cognition ,
and he had started to work on this subject : his last talk was on " The
Language of Vision ." It is a tragedy (for us) that he did not live to com -
plete at least part of this work . On the other hand , I do not believe that
this is necessarily a tragedy for John : he died with lots of projects to re-
alize and with a conviction of the importance of what he had done. He
worked until his mind could no longer accompany him . On one occasion,
shortly before his illness was discovered , he told me that people had asked
him what he would do if he knew that he had only a month to live , and
his answer was, " What I have always done ." One day before the opera-
tion , we had a working session in the hospital . And only one and a half
months before his death , John asked Marie Reyes to give him courses in
 .
XVI
Jackendoff, Bloom, and Wynn

logic , since he had missed so much of what was going on during his long
illness .

It is impossible to talk about John without mentioning the importance

that friendship had in his life. On one occasionI quoted a phrase of the
great physicist James Clerk Maxwell : " Work is good and reading is good ,
but friends are better." John approved wholeheartedly. He had a large
number of friends , as anybody who walked with him from the Faculty
Club to the Psychology Department trying desperately not to miss a lec-
ture can witness . Everybody in the street would stop and say, " Hi , John ,"
and some chat would result . But we would always arrive on time : punc -
tuality was a must with John .
If we ask why he attracted so many friends , part of the answer seems
clear : he would welcome anybody who needed a reference or some advice
on an academic subject or on a deep human problem . No subject was
considered taboo or improper . I cannot forget his words when receiving
somebody , either in his office or in his home , where he and Joyce would
exert their generous hospitality : " You are most welcome ." This renewed
reception made us feel wholly accepted once again and we could open our
hearts and laugh or cry according to the circumstances .
John was also a great entertainer . An amusing and compulsive talker ,
he would insert jokes of his own creation , some rather elaborate , about
Descartes and other historical figures . He was very proud of them and
would celebrate them with gusto , getting mad at the people who didn 't
laugh with him . He had a talent for reating limericks and he could play
the flute and the banjo . He had other talents as well , some quite unex-
pected. I remember that some friends in Belgium invited us for dinner
after John had given a beautiful lecture on Brentano and intentionality .
After the dinner our friends ' children asked John to play with them . John
proposed a game of catching pieces of bread in the mouth , the pieces
being thrown by a designated person . John 's ability was extraordinary .
Afterwards , he told us that he had once beaten a dog at this game (John
was rather competitive ) .
But I believe that part of the explanation for the love he inspired lies
also in the richness of his personality , which allowed different people to
find different Johns, to find " their John ." There was the religious man
who shared his faith with Joyce through common readings and travels
and by helping those in need. There was, most visibly , the academic John :
the university man , sometimes the Herr Professor . On occasions he would
take on a very solemn voice and finish off some suggestion of a student or
 ..
Introduction XVll

an alternative theory with the words " This simply cannot be done. . ." But
at the same time, this man who was a Fellow of the Royal Society of
Canada would sit down with undergraduatesin my course of elementary
logic, and would bring his homework to my teaching assistantto fight
with him for a better grade. Nothing is so touching for me as the sight of
this distinguishedman of over sixty going back to school to study logic
and category theory becausehe believedthat thesewere tools that a cog-
nitive psychologist should know. " My " John is also John the student. In
this context, I cannot forget his son Kieran's role as John's teacher: John
told me how he could appreciate films that were at first foreign to his
sensibility thanks to his son's efforts.
But I feel that I cannot enumerate, let alone do justice to, all the Johns
that so many peopleloved and found so fascinating. . . . John: we will miss
you . . .

The essaysin this volume are grouped roughly along thematic lines. We
begin with Richard Kearney's remarks expanding on John's early work
on languageand ethnicity. This is followed by a seriesof chapterson the
foundations of logic and conceptsby Anil Gupta, Michael Makkai , Ray
Jackendoff, Storrs McCall , David Olson, and SandeepPrasada. Then
comes an interlude of chapters on more philosophical topics by Steven
Davis and Leslie McPherson, after which we turn more directly to the
relation between language and conceptualization in chapters by Steven
Pinker and Alan Prince, Myrna Gopnik, Paul Bloom, SusanCarey and
Fei Xu , Geoffrey Hall , and Yuriko Oshima-Takane. We end with two
chapterson mathematical approachesto cognition, one by William Law-
vere and the other by John himself and his three dear colleaguesMarie La
Palme Reyes, Gonzalo Reyes, and Houman Zolfaghari.
This grouping, however, does not bring out the rich interconnections
among the chapters and their relation to John's work . We leave the
appreciation of thesedetails to the reader.
Paul Bloom Ray J ackendotI
Department of Psychology Program in Linguistics and
University of Arizona Cognitive Science
[email protected] Brandeis University
[email protected]
SusanCarey
Department of Psychology Richard Kearney
New York University Department of Philosophy
[email protected] University College Dublin
rkearney @ollamh .ucd .ie
StevenDavis
F . William Lawvere
Department of Philosophy
Simon Fraser University Department of Mathematics
[email protected] State University of New York at
Buffalo
Myrna Gopnik
wlawvere @acsu.buffalo .edu
Department of Linguistics
McGill University Michael Makkai

[email protected] Department of Mathematics

McGill University
Ani ! Gupta
makkai @triples .ma thomcgill .ca
Department of Philosophy
Indiana University Storrs McCall

[email protected] Department of Philosophy

McGill U niversi ty
D . Geoffrey Hall
mccall @dep.philo .mcgill .ca
Department of Psychology
University of British Columbia Leslie Margaret Perrin McPherson
[email protected] Cripple Creek , Colorado
lmpm @oldcolo .com
 ...
VIII Contributors

David Olson F ei Xu
Ontario Institute for Studiesin Department of Psychology
Education Northeastern University
University of Toronto fxu @neu.edu
[email protected]
Houman Zolfaghari
Yuriko Oshima-Takane Montreal , Quebec
Department of Psychology hzo If aghari @micro -intel .com
McGill University
yuriko @hebb.psych.mcgill.ca
StevenPinker
Department of Brain and
Cognitive Sciences
MassachusettsInstitute of
Technology
[email protected] .edu
SandeepPrasada
Department of Psychology
Dartmouth College
[email protected]
Alan Prince
Department of Linguistics
Rutgers University
[email protected]
Gonzalo E. Reyes
Department of Mathematics and
Statistics
University of Montreal
[email protected]
Marie La Palme Reyes
Montreal, Quebec
[email protected]
Karen Wynn
Department of Psychology
University of Arizona
[email protected]
Introduction Ray Jackendoff , Paul Bloom ,
and Karen Wynn

It is with a deep sense of gratitude and humility that we offer this collec -
tion of essaysin memory of John Macnamara by his students, colleagues,
and friends . The essays speak eloquently to the intellectual influence he
had on all of us, but it is seemly that we begin with John himself .
John grew up in County Limerick , Ireland , and spent the early part of
his adult life as a member of the Vincenti an community . He taught at
Castleknock and at St. Patrick 's College , Drumcondra . In 1963 he gained
a doctorate from the University of Edinburgh . His controversial research
findings cast doubt on the value of some of the educational priorities of
postindependenceIreland; they were published in a book entitled Bilin-
gualism and Primary Education . In 1966 John moved to Canada , where he
became Professor of Psychology at McGill University , a post he held till
the end of his life in 1996 .
John ' s intellectual impact in the period after he came to McGill was as
a scholar in the study of the mind and how it develops. Before reviewing
some of the contributions he made in different areas of psychology , it is
important to understand his broader intellectual commitments . These
pervade his work and are themselves interesting and important .
There is a tendency for psychologists to conclude that the human mind
is really quite simple . Psychologists influenced by the ideas of Watson and
Skinner have long argued that thinking and learning can be explained
through a few basic laws. Some reductionists believe that psychological
phenomena can be entirely accounted for in terms of principles of biology

We are grateful to John 's wife Joyce for her assistance in writing this introduction .
Some paragraphs are adapted from a previously published essay by Paul Bloom .
We also wish to thank Albert Bregman for his assistance and encouragement in
the initial stages of developing this volume .
 x
Jackendoff, Bloom, and Wynn

and chemistry . Many contemporary scholars are entranced with the idea
that the mind works like a computer (according to some, a rather un-
complicated computer ) . And surely , it is often said, we only need a simple
theory to account for the inner workings of a child - who is, after all , a
quite simple creature !
In John 's writings , one finds the sometimes unfashionable conviction
that the mind of the child is extraordinarily rich and complex . He begins
his 1982 book Names for Things by noting that psychologists typically
ignore the complexity of language learning - at the cost of not being able
to explain it . His favorite example , as will be seen throughout this vol -
ume, was that something as apparently simple as learning the name of a
dog requires considerable mental resources. For instance , children must
be able to appreciate the intentions of others . The 2-year -old must realize
that when an adult uses the name " Freddie " in certain contexts , he or she
intends it to refer to the dog and nothing else. Children must also possess
certain logical resources. The child has to understand how this new word
" Freddie " relates to other parts of speech, such as the common noun
" dog " and the adjective " brown ," and how it contributes to the meanings
of the sentencesthat it appearsin . Perhapsmost importantly , children can
use words to refer : to a real animal for a proper name like " Freddie , " to
an abstract kind for a common noun like " dog ," and to different kinds of
entities altogether when it comes to words like " two " and " Santa Claus ."
John argued that explaining the child 's learning and understanding of
language requires a psychology that includes notions such as inten -
tionality , reference, and truth . This means that psychology and philoso -
phy are more related than many would have thought . It also means that a
complete theory of the mind cannot be found in the fields of biology and
computer science , as such sciences cannot capture these essential proper -
ties of our mental life .

The implications of this perspective can be pursued through three re-

lated aspects of John 's research: his studies of language learning , his con -
tributions to the debate over the nature of child development , and his
work on a psychology of ideals- only a minority of his writings . In all the
best ways , John had the style of a nineteenth -century intellectual . He
wrote about an extraordinary variety of topics , from the nature of free
will to the demise of Freudian psychoanalysis to what formal logic has to
say' about the Holy Trinity . He also did not shy away from disagreement ,
and he was involved in productive and civil debate with scholars such as
the philosopher Mario Bunge and the linguist N oam Chomsky .
Introduction Xl.

Finally , he was one of the very few psychologistswho isn't a chore to

read . He wrote with style and wit , and most of his work was directed
to the educated public ; it was not crafted for a small group of fellow
scholars .

In 1972 John published a classic article with the title " The Cognitive
Basis of Language Learning in Infants ." This article presented a proposal
that has grown to be highly influential . John suggested that although
there are important regards in which language learning is special, distinct
from other sorts of learning that children succeed at, it also relies crucially
on the child 's general understanding of the situations in which sentences
are used. He proposed that the child 's learning of basic grammar (what is
a noun and what is a verb , how words are ordered within a sentence , and

so on) requires knowing what words and sentences mean , and how they
correspond to the external world . In support of this perspective , John and
two colleagues, Nancy Katz and Erika Baker , wrote their 1974 article ,
" What 's in a Name ? A Study of How Children Learn Common and
Proper Names ," which reported an amazing finding . Their experiments
found that even 18-month -olds can use subtle linguistic cues, such as the
grammatical difference between " This is wug " and " This is a wug ," to
learn proper names and common nouns . These two articles have had a
substantial influence on the study of language , motivating considerable
research seeking to better clarify the relationship between children 's
understanding of meaning and their learning of basic grammar .
After the publication of these articles , most of John 's research into
language focused on the learning of words . This work was first discussed
at length in Names for Things and was subsequently elaborated in many
articles and in two further books , A Border Dispute .' The Place of Logic in
Psychology ( 1986) and The Logical Foundations of Cognition ( 1994, edited
in collaboration with Gonzalo Reyes) . This research, some in collabora -
tion with Gonza10 Reyes and Marie La Palme Reyes, focused on the
logical foundation of the learning of proper names and common nouns .
Consider again what a child must know in order to understand a proper
name . One aspect of this knowledge is that this word follows a single
object over time . Even if Freddie were to change color or lose a leg, he
would still be Freddie . On the other hand , if he were to have a twin

brother that looked just like him , this brother would not be Freddie . The
name " Freddie " does not pick out all objects with a certain appearance; it
picks out a unique dog , regardless of his appearance . The child must also
have some understanding of what changes dogs can go through and still
 ..
XII Jackendoff, Bloom, and Wynn

remain dogs, as well as what eventswould causesomethingto ceasebeing

a dog. More generally, John and his colleaguesarguedthat understanding
a proper name requiresan appreciation of the kind that the named object
belongsto.
At somelevel, this is all common sense.Who would doubt that Freddie
would still be Freddie if he lost a leg, or that the word " Freddie" refers to
the whole dog and not just to the head or the tail? This appreciation of the
logic of namescomesso naturally to us that it is easyto underestimatethe
richnessof the logical resourcesrequired for this understanding. John and
his colleagues developed an extensive theory of these resources and
argued further that much of the knowledge that underlies the learning of
proper namesand common nouns cannot itself be learned. Children must
be born with it .
This research has implications for a broader debate within develop-
mental psychology. Jean Piaget and his followers argued that the minds
of children are very different from those of adults. Whereas adults are
capable of complicated logical thought, children are limited to thinking
about specific events; they are incapable of abstraction. Adults can take
the perspectiveof others, but children are egocentricand believethat their
own perspectiveon the world is all there is. An adult is a moral being, but
a child is a little Attila the Hun, lacking any real understandingof fairness
or morality .
John was a central figure in this debate, arguing that this conception of
developmentis mistaken. This is not to deny that children know a great
deal lessthan adults or that they neednurturing and care in order for their
moral and intellectual capacitiesto thrive. But John argued that children
possessconsiderableresourcesto start with . To seethis, one must explore
(through logical analysis, supplementedwith careful experiments) the
sorts of things children are very good at. The learning of proper names
has beendiscussedabove; other examplesinclude children's knowledge of
logical terms like " some," their ability to understand and manipulate
small numbers, their understandingof mental notions like pretending and
forgetting, and the rich moral knowledge that is demonstrated in their
everydayinteractions with their parents and siblings.
F or each of thesedomains, John's researchposesthe samechallengeto
developmental psychology. If children are so limited in their capacities,
how is it that they know so much? If they have no logic, how can they
successfullylearn expressionsof language that depend so crucially on
logical notions? If they are egocentric, how is it that they learn the
 ...
Introduction Xlll

meanings of pronouns like " I " and " me," since learning these pronouns
requires taking the perspectivesof others? If they have no conception
of fairness or morality , how is it that they understand a fairy tale like
" Cinderella ," which is impossible to appreciate without understanding
that the stepmother is acting unfairly ? All the evidence, John argued ,
leads to the conclusion that children possess a rich set of cognitive , logi -
cal, and moral capacities from the very start .
John also presented more general arguments against the view that
children lack an appreciation of logic and morality . He noted that psy-
chologists fond of this view have never explained how children come
to acquire such capacities . Followers of Jean Piaget have appealed to
" assimilation " and " accommodation ," processes through which the
child 's primitive knowledge becomes more abstract through interaction
with the environment . But in a 1976 article entitled " Stomachs Assimilate
and Accommodate , Don 't They ?" , John argued that these notions are
empty metaphors that explain nothing . Along with the work of Jerry
F odor and N oam Chomsky , John 's empirical and theoretical defense of
the notion of innate ideas and capacities caused a major shift in the way
psychologiststhink about the mind and its development.
The final aspect of John 's work that we wish to mention is his proposal ,
first outlined in a 1990 article entitled " Ideals and Psychology ," that an
adequate psychology must explain the human capacity to understand
ideals . A good example of this , discussed first by Descartes, lies in the
domain of geometry . We all possess the notion of perfect geometric
forms , such as a perfect triangle or a perfect point , but no such forms exist
in nature and none ever will . Scientific progress is also based on ideals ;

physics would be nowhere without its frictionless plane and its perfect
vacuum . Similarly , we have personal ideals that we aspire to , notions such
as humility , friendship , and courage . As John put it , " Idealization is as
natural to the mind as breathing to the body ." But how is it that we come
to appreciate such ideals, given that , by their very nature , they are not to
be found in our experience?
In a 1991 article called " The Development of Moral Reasoning and the
Foundations of Geometry ," John provided a hint of what a psychology of
ideals would look like . The article begins with a critical discussion of a
well -known stage theory of moral development proposed by Jean Piaget
and extended by Lawrence Kohlberg . John argued that this theory seri-
ously underestimates children ' s knowledge , fails to describe adult moral
competence, and does not explain how notions of morality actually de-
 .
XIV Jackendoff, Bloom, and Wynn

velop and how supposedlyamoral children grow into moral adults. John
then proposed an alternative theory of moral development, based on an
intriguing analogy betweensystemsof morality and systemsof geometry.
In both domains, there exist ideal elements(in geometry, elementssuch as
point and line; in morality , elements such as fair and good) that are
related to one another through a system of axioms. John suggested
that there are extensive unlearned elements within both geometry and
morality . Children start off with an understanding of certain ideals in
both domains, and these are the notions that all people share. But psy-
chologists must also explain how new geometric and moral knowledge
emergesin adults and children. New geometries have been developed
by mathematicians, such as the non-Euclidean geometry proposed by
Riemann in the nineteenth century, and new moral ideals have been in-
troduced in the course of history, such as the ideal of chastity introduced
by the Christians. How are these learned and understood? The parallel
betweengeometry and morality placesthe study of moral developmentin
a striking new light . It might be that the child who is coming to grasp a
novel moral system is actually acquiring a coherent formal structure,
complete with ideals and axioms, in much the sameway he or shewould
learn a new systemof geometry, or mathematics, or physics.
This article is not as well known as many of John's other works dis-
cussedabove. (Perhaps the title scarespeople away!) But it is hard to
think of a better work in developmental psychology. Although it is
entirely accessiblein style and content, it doesnot shy away from the hard
questions. It is a significant intellectual accomplishment, but also some-
thing more, as it displays severalproperties typical of John's work . There
is creativity, intellectual courage, and a strong curiosity about the work-
ings of the world . Most of all, there is respectfor its subject matter: the
minds of young children and how they develop. It is more than an article
about morality ; it is itself a highly moral work .
For a closer look into John as a person, we quote from the eulogy by
John's friend and collaborator Gonzalo Reyes. The essencethat Gonzalo
so beautifully evokeshere strikes a familiar chord for anyone who knew
John.

I first met John ten yearsago whenhe gavea talk in the Mathematics
Departmentat McGill in October1985.He raisedthequestionof givinga
logicalaccountof thephrase" Freddieis a dog" that childrenlearnby the
Introduction xv

age of 18 months , claiming that logical theories were unable to account

for it . To tell you the truth , I thought that he was mad : how can there be
problems with that? But I found his talk so intriguing that I went to dis-
cuss it with him at the end. This was the beginning of a collaboration that
developed into a close friendship and resulted in one of the most beautiful
and rich periods of my life .
This talk at McGill was my first encounter with John 's approach to in -
tellectual matters : to ask basic simple questions to test theories or view -
points and to demand straightforward answers. Of course this Socratic
method enraged some people , but it delighted others , especially when he
used it masterfully in his talks . He had a robust , realistic , no -nonsense at-
titude and no patience for idealistic views that he felt did not go to the heart
of things . In this respect, he was like a brick , giving a senseof security to all
of us. In fact , John was a man of profound convictions : he was deeply
religious and deeply realist . Probably the thinker that was closest to his
heart was Aristotle . At the beginning of our collaboration , I asked what I
should read in psychology , and his answer was " Only De Anima ."
Another aspect of John 's approach to intellectual problems was the
unity of his thought . I discovered with bewilderment that behind the dog
Freddie stood none other than Plato , and behind the questions that
looked so trivial at the beginning stood in fact a formidable problem ,
Plato 's paradox of learning , which still haunts the venerable house of
philosophy . And the answer to these questions appeared to John as a
first step in creating a cognitive psychology that would be, as he put it ,
" worthy of its name ." His talks were fascinating , with flashes that allowed
us to catch a glimpse of this rich web of connections .
John died in a period that burst with creativity . He thought that the
time was ripe to study the relations between perception and cognition ,
and he had started to work on this subject : his last talk was on " The
Language of Vision ." It is a tragedy (for us) that he did not live to com -
plete at least part of this work . On the other hand , I do not believe that
this is necessarily a tragedy for John : he died with lots of projects to re-
alize and with a conviction of the importance of what he had done. He
worked until his mind could no longer accompany him . On one occasion,
shortly before his illness was discovered , he told me that people had asked
him what he would do if he knew that he had only a month to live , and
his answer was, " What I have always done ." One day before the opera-
tion , we had a working session in the hospital . And only one and a half
months before his death , John asked Marie Reyes to give him courses in
 .
XVI
Jackendoff, Bloom, and Wynn

logic , since he had missed so much of what was going on during his long
illness .

It is impossible to talk about John without mentioning the importance

that friendship had in his life. On one occasionI quoted a phrase of the
great physicist James Clerk Maxwell : " Work is good and reading is good ,
but friends are better." John approved wholeheartedly. He had a large
number of friends , as anybody who walked with him from the Faculty
Club to the Psychology Department trying desperately not to miss a lec-
ture can witness . Everybody in the street would stop and say, " Hi , John ,"
and some chat would result . But we would always arrive on time : punc -
tuality was a must with John .
If we ask why he attracted so many friends , part of the answer seems
clear : he would welcome anybody who needed a reference or some advice
on an academic subject or on a deep human problem . No subject was
considered taboo or improper . I cannot forget his words when receiving
somebody , either in his office or in his home , where he and Joyce would
exert their generous hospitality : " You are most welcome ." This renewed
reception made us feel wholly accepted once again and we could open our
hearts and laugh or cry according to the circumstances .
John was also a great entertainer . An amusing and compulsive talker ,
he would insert jokes of his own creation , some rather elaborate , about
Descartes and other historical figures . He was very proud of them and
would celebrate them with gusto , getting mad at the people who didn 't
laugh with him . He had a talent for reating limericks and he could play
the flute and the banjo . He had other talents as well , some quite unex-
pected. I remember that some friends in Belgium invited us for dinner
after John had given a beautiful lecture on Brentano and intentionality .
After the dinner our friends ' children asked John to play with them . John
proposed a game of catching pieces of bread in the mouth , the pieces
being thrown by a designated person . John 's ability was extraordinary .
Afterwards , he told us that he had once beaten a dog at this game (John
was rather competitive ) .
But I believe that part of the explanation for the love he inspired lies
also in the richness of his personality , which allowed different people to
find different Johns, to find " their John ." There was the religious man
who shared his faith with Joyce through common readings and travels
and by helping those in need. There was, most visibly , the academic John :
the university man , sometimes the Herr Professor . On occasions he would
take on a very solemn voice and finish off some suggestion of a student or
 ..
Introduction XVll

an alternative theory with the words " This simply cannot be done. . ." But
at the same time, this man who was a Fellow of the Royal Society of
Canada would sit down with undergraduatesin my course of elementary
logic, and would bring his homework to my teaching assistantto fight
with him for a better grade. Nothing is so touching for me as the sight of
this distinguishedman of over sixty going back to school to study logic
and category theory becausehe believedthat thesewere tools that a cog-
nitive psychologist should know. " My " John is also John the student. In
this context, I cannot forget his son Kieran's role as John's teacher: John
told me how he could appreciate films that were at first foreign to his
sensibility thanks to his son's efforts.
But I feel that I cannot enumerate, let alone do justice to, all the Johns
that so many peopleloved and found so fascinating. . . . John: we will miss
you . . .

The essaysin this volume are grouped roughly along thematic lines. We
begin with Richard Kearney's remarks expanding on John's early work
on languageand ethnicity. This is followed by a seriesof chapterson the
foundations of logic and conceptsby Anil Gupta, Michael Makkai , Ray
Jackendoff, Storrs McCall , David Olson, and SandeepPrasada. Then
comes an interlude of chapters on more philosophical topics by Steven
Davis and Leslie McPherson, after which we turn more directly to the
relation between language and conceptualization in chapters by Steven
Pinker and Alan Prince, Myrna Gopnik, Paul Bloom, SusanCarey and
Fei Xu , Geoffrey Hall , and Yuriko Oshima-Takane. We end with two
chapterson mathematical approachesto cognition, one by William Law-
vere and the other by John himself and his three dear colleaguesMarie La
Palme Reyes, Gonzalo Reyes, and Houman Zolfaghari.
This grouping, however, does not bring out the rich interconnections
among the chapters and their relation to John's work . We leave the
appreciation of thesedetails to the reader.
Publications of
John Macnamara

Books

Macnamara , J. Bilingualism and primary education: A study of Irish

experience. Edinburgh : Edinburgh University Press, 1966.
Macnamara , J. (Ed .) . Problems in bilingualism . April issue of the Journal
of Social Issues, 1967, 23 .
Madaus , G . F ., and Macnamara , J. Public examinati ~ns: A study of
the Irish Leaving Certificate . Dublin : St. Patrick 's College , Educational
Research Centre , 1970 .
Macnamara , J. (Ed .) . Language learning and thought . New York : Aca -
demic Press , 1977 .
Macnamara , J. Names for things: A study of human learning . Cambridge ,
MA : MIT Press , 1982 .
Macnamara , J. A border dispute: The place of logic in psychologiv. Cam -
bridge , MA : MIT Press, 1986. Polish translation by Michal Zagrodzki .
Warsaw : Wydawnictwo Nankowe PWN , 1993.
Macnamara , J., and Reyes, G . E . (Eds.). The logical foundations of
cognition . New York : Oxford University Press, 1994.

Articles

Macnamara , J. The use of reason at the age of seven: Two contrasting

views . Irish Ecclesiastical Record , 1961 , 98 , 92 - 103 .

Macnamara , J. The report of the Commission on Irish : Psychological as-

pects . Studies , 1964 , 53 , 164- 173 .
Macnamara , J. Zero error and practice effects in Moray House English
Quotients . British Journal of Educational Psychology , 1964, 34, 315- 320.
Macnamara , J. The problem solving difficulties of bilingual children .
xx Publications

Paper read at A .G.M . of Northern Ireland Branch of the British Psy-

chological Society, January 1965. Bulletin B.P.S., 1965, 18, 60, 58- 59
(Abstract).
Macnamara, J. The bilingual's linguistic performance: A psychological
overview. Journal of Social Issues, 1967, 23, 58- 77.
Macnamara, J. The effectsof instruction in a weaker language. Journal of
Social Issues, 1967, 23, 121- 135.
Macnamara, J. The linguistic independenceof bilinguals. Journal of
Verbal Learning and Verbal Behavior, 1967, 6, 729- 736.
Kellaghan, T ., and Macnamara, J. Reading in a second language. In
M . D . Jenkinson (Ed.), Reading instruction: An international forum
(pp. 231- 240). Newark, DE : International Reading Association, 1967.
Macnamara, J., Feltin, M ., Hew, M ., and Klein , M . Analysis of diffi-
culties of reading in a weaker language. Irish Journal of Education, 1968,
2, 41- 53.
Macnamara, J., Krauthammer, M ., and Bolgar, M . Language switching
in bilinguals as a function of stimulus and responseuncertainty. Journal
of Experimental Psychology, 1968, 78, 208- 215.
Lambert, W . E., and Macnamara, J. Some cognitive consequencesof
following a first-grade curriculum in a secondlanguage. Journal of Edu-
cational Psychology, 1969, 60} 86- 96.
Macnamara, J., and Madaus, G. F. Marker reliability in the Irish Leav-
ing Certificate. Irish Journal of Education, 1969, 3, 5- 21.
Macnamara, J. How can one measurethe extent of a person's bilingual
capacity? In L . G. Kelly (Ed.), Description and measurementof bilin-
gualism (pp. 80- 97). Toronto : University of Toronto Press, 1969.
Kellaghan, T ., Neuman, E., and Macnamara, J. Teachers' assessments of
the scholasticprogressof pupils. Irish Journal of Education, 1969, 3, 95-
104.
Macnamara, J., and Madaus, G. F. The quality of the Irish Leaving
Certificate Examination. Irish Journal of Education, 1970, 4, 5- 18.
Macnamara, J. Bilingualism and thought. In J. E. Alatis (Ed.), George-
town University. 21st Annual Round Table, 1970, 23, 25- 40.
Macnamara, J. Comparative studies of reading and problem solving in
two languages. Tesol- Quarterly, 1970, 4(1- 2), 107- 116.
Macnamara, J. Parsimony and the lexicon. Language, 1971, 47, 359-
374.
Macnamara, J., and Kushnir, S. The linguistic independenceof bilin -
guals: The input switch. Journal of Verbal Learning and Verbal Behavior,
1971, 10, 480- 487.
 .
Publications XXI

Macnamara , J . Successes and failures in the movement for the restoration

of Irish . In J. Rubin and B . Jemudd (Eds.), Can language be planned ?

(pp . 65- 94). Honolulu : University of Hawaii Press, 1971.
Macnamara , J. The cognitive basis of language learning in infants . Psy-
chological Review, 1972, 79, 1- 13.
Macnamara , J. Perspectives on bilingual educa ti on in Canada . Canadian
Psychologist , 1972, 13, 341- 349.
Macnamara , J., Kellaghan , T ., and O 'Cleirigh , A . The structure of the
English lexicon : The simplest hypothesis . Language and Speech, 1972, 15,
141 - 148 .

Kellaghan , T ., and Macnamara , J. Family correlates of verbal reasoning

ability . Developmental Psychology , 1972, 7, 49- 53.
Macnamara , J. The objectives of bilingual education in Canada from an
English -speaking perspective . In M . Swain (Ed .), Bilingual schooling;
some experiences in Canada and the United States . A report on the Bilin -
gual Education Conference, Toronto , March 11- 13, 1971 (pp . 7- 10) .
Toronto : The Ontario Institute for Studies in Education , 1972 .
Macnamara , J. On learning each other 's languages. In P. Watson (Ed .),
Psychology and race (pp . 393- 402) . London : Penguin , 1973.
Macnamara , J. Nurseries , streets and classrooms : Some comparisons and
deductions . Modern Language Journal , 1973, 57, 250- 254.
Schwartz , J. A ., Singer, J., and Macnamara , J. An analytic comparison of
listening in two languages. Irish Journal of Education , 1973, 7, 40- 52.
Macnamara , J. Attitudes and learning a second language . In R . Shuyand
R . W . Fasold (Eds .), Language attitudes : Current trends and prospects
(pp . 36- 40) . Washington , DC : Georgetown University Press, 1973.
Katz , N ., Baker , E ., and Macnamara , J. What 's in a name? A study of
how children learn common and proper names. Child Development, 1974,
45 , 469 - 473 .

Macnamara , J. Nurseries as models for language classrooms . In S. T .

Carey (Ed .), Bilingualism , biculturalism and education ( pp. 91- 93).
Edmonton : University of Alberta , 1974.
Macnamara , J. The generalizability of results of studies of bilingual
education . In S. T . Carey (Ed .), Bilingualism , biculturalism and education
(pp . 95- 98). Edmonton : University of Alberta , 1974.
Macnamara , J. A note on Watt ' s economy criteria . Glossa , 1975 , 9 ,
107 - 113 .

Macnamara , J. A note on Piaget and number . Child Development, 1975,

46 , 424 - 429 .
 ..
XXII Publications

Mougeon , R ., and Macnamara , J. Language maintenance , bilingualism

and religion in Gaspe East . Cahier de Linguistique , 1975, 6, 171- 178.
Macnamara , J. A reply to Brainerd . Child Development, 1976, 47, 897-
898 .

Macnamara , J. First and second language learning : Same or different ?

Journal of Education , 1976, 158, 39- 54.
Macnamara , J. Comparison between first and second language learning .
Die Neueren Sprachen, 1976, 75, 175- 188.
Macnamara , J ., Baker , E ., and Olson , C . L . Four -year -olds ' understand -
ing of pretend, forget and know : Evidence for propositional operations .
Child Development, 1976, 47, 62- 70.
Macnamara , J. Stomachs assimilate and accommodate , don 't they?
Canadian Psychological Review, 1976, 17, 167- 173.
Macnamara , J., Svarc, J., and Horner , S. Attending a primary school of
the other language in Montreal . In A . Simoes (Ed .), The bilingual child
(pp . 113- 131). New York : Academic Press, 1976.
Macnamara , J. On the relation between language learning and thought .
In J. Macnamara (Ed .), Language learning and thought, (pp . 9- 11). New
York : Academic Press , 1977 .
Macnamara , J. From sign to lang\uage. In J. Macnamara (Ed .), Language
learning and thought (pp . 11- 35) . New York : Academic Press, 1977.
Macnamara , J. Problems with concepts. In J. Macnamara (Ed .), Lan -
guage learning and thought (pp . 141- 145). New York : Academic Press,
1977 .

Macnamara , J. Children 's command of the logic of conversation . In

J. Macnamara (Ed .) , Language learning and thought (pp . 261- 288). New
York : Academic Press , 1977 .
Macnamara , J. Cognitive strategies of language learning . In W . F .
Mackey and T . Andersson (Eds.), Bilingualism in early childhood ( pp. 19-
27). Rowley , MA : Newbury House , 1977.
Macnamara , J. How do babies learn grammatical categories? In D .
Sankoff (Ed .), Linguistic variation : Models and methods (pp . 197- 210).
New York : Academic Press , 1978 .
Macnamara , J. Another unaccommodating look at Piaget . Canadian
Psychological Re),iew, 1978, 19, 78- 81.
Macnamara , J. What happens to children whose home language is not
that of the school? In G . K . Verma and C . Bagley (Eds.), Race, education
and identity (pp . 95- 104). London : Macmillan , 1979.
Macnamara , J. Let 's see. Journal of Child Language, 1979, 6, 581- 584.
Publications xxiii

Macnamara , J. Doubts about the form of development . Behavioral and

Brain Sciences , 1979 , 2 , 393 - 394 .

Macnamara , J. Physical objects: The relationship between psychology

and philosophy . Canadian Psychologist , 1981, 22, 271- 281.
Macnamara , J. Meaning . In W . B . Weimer and D . S. Palermo (Eds.),
Cognition and the symbolic process (Vol . 2, pp . 35- 61). Hillsdale , NJ :
Erlbaum , 1982 .
Macnamara , J . Not the talk : It ' s what the talk is about . Alberta Modern
Language Journal , 1983, 22, 3- 9.
Macnamara , J . ANew Testament . Review article on J . Barwise and

J. Perry , Situations and attitudes . Journal of Applied Psycho linguistics ,

1985 , 5 , 254 - 265 .

Macnamara , J. Citation classic: Cognitive basis of language learning in

children . Psychological Re1,iew, 1972, 79, 1- 13. Current Contents, Social
and Behavioral Sciences, March 1985, 17( 11), 18.
Macnamara , J. But what is belief itself ? Annals of Theoretical Psychology ,
1986 , 4 , 357 - 364 .

Macnamara , J. 0 nazywaniu rzeczy: Uwagi na temat zwiazku miedzy

logika a psychologia (Naming things : Remarks on the connection between
logic and psychology ). In I . Kurcz , J. Bobryk , and D . Kadzielawa (Eds.),
Wiedza a jezyk : Tom 1. Ogolna psychologia jezyka i neurolingwistyka
(Knowledge and language: Vol . 1. General psychology of language and
neurolinguistics ). Wroclaw : Ossolineum , 1986.
Macnamara , J. Principles and parameters : A response to Chomsky . New
Ideas in Psychology , 1986, 4, 215- 222.
Macnamara , J. On naming things : Remarks on the relation between logic
and psychology . In I . Kurcz , G . W . Shugar , and J. H . Danks (Eds.),
Knowledge and language (pp . 3- 20). Amsterdam : North -Holland , 1986.
Macnamara , J ., Govitrikar , V . P ., and Doan , B . Actions , laws and
scientific psychology . Cognition , 1988, 29, 1- 27.
Niall , K ., and Macnamara , J. Projective invariance and visual shape
constancy . Acta Psychologica , 1989, 72, 65- 79.
Macnamara , J. More about principles and parameters : A reply to
Chomsky . New Ideas in Psychology , 1989, 7, 33- 40.
Macnamara , J. Ideals and psychology . Canadian Psychology , 1990, 31,
14 - 25 .

Niall , K . K ., and Macnamara , J. Projective invariance and picture per-

ception . Perception , 1990, 19, 637- 660.
Macnamara , J. Small numbers : A dialogue . Canadian Psych 0logiv, 1991,
32 , Ill - lIS .
 .
XXIV Publications

Macnamara , J. Understanding induction . Journal for the Philosophy of

Science , 1991 , 42 , 21 - 48 .
Macnamara , J. The development of moral reasoning and the foundations
of geometry . Journalfor the Theory of Social Behaviour , 1991, 21, 125-
150 .

Macnamara , J. Linguistic relativity revisited . In R . L . Cooper and

B. Spolsky (Eds.), The influence of language on culture and thought
( pp. 35- 147) . Berlin and New York : Mouton de Gruyter , 1991.
Macnamara , J. The takeover of psychology by biology or the devaluation
of reference in psychology . In M . Piitz (Ed .), Thirty years of linguistic
evolution : Studies in honour of Rene Dirven on the occasion of his sixtieth
birthday (pp . 545- 570). Amsterdam : John Benjamins , 1992.
Macnamara , J., and Austin , G . Physics and plasticine . Canadian Psy-
chology, 1993, 34, 225- 232.
Macnamara , J., and Austin , G . Response to critics . Canadian Psychology ,
1993 , 34 , 245 - 248 .
Macnamara , J. Cognitive psychology and the rejection of Brentano .
Journal for the Theory of Social Behaviour , 1993, 23, 117- 137.
La Palme Reyes, M ., Macnamara , J., Reyes, G . E ., and Zolfaghari , H .
Proper names and how they are learned . Memory , 1994, 1, 433- 455.
Macnamara , J. Father Sheehy' s flashlight . Canadian Psychology , 1994,
35 , 124 - 128 .
La Palme Reyes, M ., Macnamara , J., Reyes, G . E ., and Zolfaghari , H .
The non-Boolean logic of natural languagenegation. PhilosophiaMathe-
matica , 1994 , 3 , 45 - 68 .
La Palme Reyes, M ., Macnamara , J., and Reyes, G . E . Functoriality and
grammatical role in syllogisms . Notre Dame Journal of Formal Logic ,
1994 , 35 , 41 - 66 .
Macnamara , J. The foundations of logic and the foundations of cogni -
tion . In W . F . Overton and D . S. Palermo (Eds.), The nature and onto -
genesis of meaning ( pp. 145- 166). Hillsdale , NJ : Erlbaum , 1994.
Macnamara , J. The mind -body problem and contemporary psychology .
In R . Szubka (Ed .), The mind -body problem : A guide to the current debate
(pp . 165- 170). Oxford : Blackwell , 1994.
Macnamara , J. The Gestalt spirit in language and cognition : A critical
review of Language : A theory of structure and use by Per Saugstad. New
Ideas in Psychology , 1994, 12, 183- 191.
Macnamara , J., La Palme Reyes, M ., and Reyes, G . E . Logic and the
Trinity . Faith and Philosophy , 1994, 11, 3- 18.
Publications xxv

Macnamara , J ., and Reyes , G . E . Introduction . In J . Macnamara and

G . E . Reyes ( Eds .) , The logical foundations of cognition ( pp . 1 - 8 ) . New

York : Oxford University Press , 1994 .

Macnamara , J . Logic and cognition . In J . Macnamara and G . E . Reyes

( Eds .) , The logical foundations of cognition ( pp . 11 - 34 ) . New York :

Oxford University Press , 1994 .

La Palme Reyes , M . , Macnamara , J . , and Reyes , G . E . Reference , kinds

and predicates . In J . Macnamara and G . E . Reyes ( Eds .) , The logical

foundations of cognition ( pp . 90 - 143 ) . New York : Oxford University

Press , 1994 .

Macnamara , J . , and Reyes , G . E . Foundational issues in the learning of

proper names , count nouns and mass nouns . In J . Macnamara and G . E .

Reyes ( Eds .) , The logical foundations of cognition ( pp . 144 - 176 ) . New

York : Oxford University Press , 1994 .

Occasional Pieces

Macnamara , J . Tuirse sa scoil . Dochas " 1965 , 3 " 106 - 107 .

Macnamara , J ., and Edwards , J . Attitudes to learning French in the

English - speaking schools of Quebec . Study 8 prepared for the Commis -

sion of Inquiry on the Position of the French Language and on Language

Rights in Quebec . Quebec Official Publisher , 1973 .

Macnamara , J . The Irish language and nationalism . The Crane Bag ,

1977 , 1 , 40 - 44 .

Macnamara , J . La philosophie du LivTre Blanc . Le Devoir " 1977 , jeudi 5

mai , 5 .

Macnamara , J . Race and intelligence . Irish Review , 1988 , 3 , 55 - 60 .

Macnamara , J . Cinderella and the soul . New Oxford Review , 1990 , Nov .,

7 - 11 .

Macnamara , J . ( 1994 ) . Where Freud went wrong . Review article on Paul

Roazen : Meeting Freud ' s family . Literary Review of Canada , 3 , No . 11 ,

15 - 16 .

Reviews

Fletcher , B . A .: A philosophy for the teacher . In Philosophical Studies ,

1963 , 12 , 308 .
Young , M . : Innovation and research in education . In Irish Journal of
Education , 1967 , 1 , 150 - 152 .
 .
XXVI Publications

Morrison , D . F . : Multivariate statistical methods . In Irish Journal of

Education , 1969 , 3 , 71 - 73 .

Chomsky , C .: The acquisition of syntax in children from 5 to 10 . In

General Linguistics , 1970, 10, 164- 172.
Blumenthal, A . L .: Language and psychology (Review title : Discovering
respectable ancestors). In Contemporary Psychology , 1971, 16, 497- 498.
Painchaud , L .: Le bilinguisme a l 'universite . In International Review of
Applied Linguistics , 1971, 9, 72.
Balkan , L .: Les effets du bilinguisme franrais -anglais sur les aptitudes
intellectuelles . In Lingua , 1972, 28, 159- 162.
Herriott , P.: An introduction to the psychology of language. In Irish
Journal of Education , 1972, 5, 116- 118.
Steinberg, D . D ., and Jakobovits , L . A . (Eds.): Semantics: An inter -
disciplinary reader in philosophy, linguistics and psychology (Review title :
Family reunion ) . In Contemporary Psychology , 1972, 17, 428- 429.
Pearce, J.: School examinations . In Irish Times, Educational Supplement,
1973, April .
Beilin , H .: Studies in the cognitive basis of language development (Review
title : A general theory of cognitive and linguistic development ) . In Con-
temporary Psychology , 1976, 21, 649- 650.
Bates, E .: Language and context : The acquisition of pragmatics (Review
title : More talk about baby talk ) . In Contemporary Psychology , 1977, 22,
494 - 495 .

Maratsos , M . P.: The use of definite and indefinite reference in young chil -
dren: An experimental study in semantic acquisition . In Child Development
Abstracts and Bibliography, 1977, 51, 156.
Moerk , E . L .: Pragmatic and semantic aspects of early language develop-
ment. In Child DevelopmentAbstracts and Bibliography, 1978, 52, 85- 86.
Gelman , R ., and Gallistel , C . R .: The child 's understanding of number. In
Canadian Journal of Psychology , 1978, 32, 270- 271.
McLaughlin , B .: Second language acquisition in childhood . In Child
Development Abstracts and Bibliography , 1979, 53, 76- 77.
aleron , P.: L 'enfant et I 'acquisition du langage (Review title : Cognition in
language learning ) . In Contemporary Psychology , 1980, 25, 497- 498.
Lock , A .: The guided reinvention of language (Review title : Anthropology
in the nursery ). In Contemporary Psychology , 1981, 26, 542- 543.
Ginsburgh , H . P. (Ed .) : The development of mathematical thinking
(Review title : All about sums) . In Contemporary Psychology , 1984, 29,
136 - 137 .
 ..
Publications XXVII

Johnson -Laird , P. N .: Mental models : Towards a cognitive science of

language , inference , and consciousness (Review title : Faithful account of
a stewardship ) . In Journal of Mathematical Psychology , 1984, 28, 496-
503 .

Lycan , W . G .: Logical form in natural language. In Journal of Applied

Psy cholingu istics, 1986, 3, 182- 185.
Waldron , T . P.: Principles of language and mind . In Journal of Language
and Social Psychology , 1986, 5, 76- 80.
Wells , G . Language development in the preschool years. In Canadian Psy-
chology, 1987, 28, 304- 305.
Morgan , J. L . From simple input to complex grammar . In McGill Journal
of Education , 1987, 22, 165- 167.
Fine , K .: Reasoning with arbitrary objects. In Journal of Symbolic Logic ,
1987 , 2 , 305 - 306 .
Pateman , T .: Language in mind and language in society . In Mind and
Language , 1988 , 3, 71 - 74 .
Fodor , J. A .: Psychosemantics (Review title : Cognitive science in peril ) . In
Contemporary Psychology , 1989, 33, 1040- 1041.
Astington , J. W ., Harris , P. L ., and Olson , D . R . (Eds.): Developing
theories of mind (Review title : Children as commonsense psychologists ) .
In Canadian Journal of Psychology , 1989, 43, 426- 429.
Aitchinson , J .: Words in the mind : An introduction to the mental lexicon .
In Journal of Language and Social Ps} 'chology, 1989, 8, 355- 357.
Eco , U ., Santambrogio , M ., and Violi , P. (Eds.): Meaning and mental
representations. In Journal of Language and Social Psychology , 1989, 8,
349 - 353 .

Pinker , S.: The language instinct ,. and Jackendoff , R .: Patterns in the mind .
In Boston Globe , 13 March 1994 .
Cha pter 1

Language and Nationalism Richard Kearney

John Macnamara's writings on the psychology of language are inter-

nationally recognized. Lesswell known perhapsis his work on the politics
of language. His research on the relationship between nationality and
language- ranging from his doctoral thesis on bilingualism in Irish edu-
cation (1966) to his later essaysin the Irish journal , The CraneBag, in the
1970s- had a considerableimpact on the cultural identity debatesin his
native country. Macnamara's interventions were controversial. His first
book on bilingualism earnedhim the public censureof the then Minister
for Education in the Republic, who called for his resignation as President
of St. Patrick's Teacher Training College in Dublin . His spirited argu-
ment for the separation of racial identity and language- that is, of Irish
ethnicity and Gaelic speech- in an essayentitled " The Irish Language
and Nationalism" (1977) provoked equally vehementresponses .
John Macnamara challenged the race-language equation, especially
when narrowly applied to the politics of nationhood. He disputed, not the
fact that countries that speak different languagesare often different cul-
turally , but the belief that difference in language necessarilyimplies dif-
ferencein culture. In linguistic theory, Macnamara showed, this equation
of languageand national identity derived largely from German theorists
like Herder and Fichte in the nineteenth century and from Benjamin Lee
Whorf in the twentieth. In political history, the equation was identified
with a wide variety of nationalist movementsfrom Russia and Germany
to Hungary and Ireland. In his 1977essay, he concludedthat although the
Irish languageplayed a central role in establishingthe differencebetween
the Irish Free State and the rest of the British Empire in the early part of
this century, postindependenceIreland was in a position to acknowledge
that one could be equally Irish speakingGaelic or English.
2 Kearney

With political independence and a securelybasedtradition of Irish literaturein

English, a largepart of the programmethat motivatedthe settingup of the Free
Statehasbeenrealized , thoughnot in the form that the foundersforesaw . We no
longerneedargumentsto provewe are a distinctrace, not because we havesat-
isfiedtheworld that we are, but because to us the notionof raceis irrelevant.The
restof the English-speakingworld is not trying to annexus politically. . . . Every-
onetodayseesIrelandandIrish peopleashavinga flavourandstyleall their own.
In any case , to my mind it is just as falseto aim at beingdistinctiveas it is to
imitate. (Macnamara1977,40- 41)
In this chapter I would like to respondto Macnamara's questioning of
the criteria of nationalism and national identity by extending and supple-
menting his argument in somecritical directions. So doing, I wish to pay
homageto his timely and salutary intervention.

In current usage, a nation is often assumedto be a state, or a group of

people aspiring in common toward the condition of a state. As a state, the
modern nation is generally, in fact or in aspiration, a sovereignentity with
significant control over its government and identity, and recognized as
such by those residentwithin the nation and by other nations in the world
(the assemblyof nation-statesknown as the United Nations). As a legal
entity, the nation-state usually endorseswhat might be called " civic"
nationalism: the claim that the nation is composedof all those- regard-
lessof color, language, creed, or race- who subscribeto the nation's polit -
ical principles or constitution. The nation is thus defined as " civic" to the
extent that it is recognizedas a community of " equal, rights-bearing citi-
zens, united in patriotic attachment to a shared set of political practices
and values" (Ignatieff 1993, 6- 7). Britain is frequently cited as an example
of civic nationalism in its efforts, from the mid- eighteenthcentury onward,
to forge a nation-state composed of four different nations (Irish, Scots,
Welsh, and English), under a civic rather than ethnic definition- that is, a
sharedallegianceto a common Parliament, Crown, rule of law. Whether
Britain actually succeededin this, or simply masked " English national-
ism" behind the four-nation-state, is a moot point to which I will return
below. Less controversial examplesof the " civic" nation-state are to be
found in the American and French republics after their respectiverevo-
lutions of 1776 and 1789, though it must be rememberedthat even here
citizenship was still largely confined, for many decades, to the enfran-
chisedelite of white, propertied males and did not always include women,
slaves, or aboriginal peoples.!
 3
Language and Nationalism

A second common way of understanding nation is as territory . This

definition need not always involve a state as such, but it does, minimally ,
lay claim to a specificplace or land that comprisesthe so-called national
tel"ritory . To fully exist, the nation thus geographicallydefinedmust cover
or occupy the national territory - " the land which the nationalists believe
to be the nation's by right and (which) embracesall the members of the
putative nation therein domiciled" (Fennell 1989, 13).2 In the case of
Ireland, geography has been one of the most commonly invoked criteria
for justifying the aspiration to unity within the island and separationfrom
the neighboring island of Britain . The fact that Ireland is an integral
island, " surrounded by water," has been one of the mainstays of sepa-
ratist nationalism.3 The problem here is, of course, that both the Irish
Republic and Great Britain have consideredNorthern Ireland to be part
of their respective" national territories," though it has been reconfirmed
as British dejure sincethe 1920Government of Ireland Act . Many nations
have, of course, existed without a land (e.g., the Jews for centuries, the
Kurds, the Palestinians), but there are few if any examplesof nations with
no senseof territorial identification.
A third common senseof nation, critically discussedby John Macna-
mara, is that of ethnicity. Here the definition of belonging is one of blood
rather than law. The nation is defined, accordingly, in terms of a racially
homogeneous" people" that seeksout a state appropriate to its unique
identity. In contrast to most Western nation-states, which today define
their nationhood in terms of common citizenship (by birth , residence, or
allegiance), Germany is an example of a nation that still defines itself
ethnically. For the ethnic nation, a person's deepestidentity is inherited
rather than chosen; what unifies a society is common roots rather than
laws. " Regardlessof the real nature of the state-society, the nation-state
treats it as a mass of ethnically uniform individuals who exist to serveits
welfare and convenience. It is, after all, the nation" (Fennell 1989, 13). In
such a scenario, it is the national community- the Volk- that definesthe
individual citizens rather than the other way around. The People before
persons. We thus find severalGerman Romantics of the nineteenth cen-
tury repudiating the Enlightenment idea that the state createdthe nation,
and mobilizing German patriotic resistanceto Napoleon's invasion of
the German principalities around a specifically German- as opposedto
French revolutionary- understanding of the nation-state. The extraordi-
nary appeal of this ethnic model of nationalism for emergentnations in
the nineteenth century is describedby Michael Ignatieff as follows:
4
Kearney

What gave unity to the nation , what made it a home , a place of passionate at-
tachment , was not the cold contrivance of shared rights but the people 's preexist -
ing ethnic characteristics : their language , religion , customs and traditions . The
nation as Volk had begun its long and troubling career in European thought . All
the peoples of nineteenth -century Europe under imperial subjection - the Poles
and Baltic peoples under the Russian yoke , the Serbs under Turkish rule , the
Croats under the Hapsburgs - looked to the German ideal of ethnic nationalism
when articulating their right to self-determination . ( 1993, 7)

As a result , when Gefll1any was finally unified under Bismarck in 1871

and assumed world -power status, it served as exemplar of the success of
ethnic nationalism for the " captive nations " of imperial Europe - and
beyond .4
As Macnamara pointed out , certain instances of Irish nationalism also
display an ethnic feature based on alleged differences between Gaelic and
non -Gaelic . This is, admittedly , uncharacteristic from a historic point of
view . Ideological geneticism or Darwinian biology were rarely invoked
in the Irish separatist cause. Probably the closest that one gets to some
kind of racist inference was Douglas Hyde 's odd , and largely innocuous ,
theory about natural selection and language . " As President of the lin -
guistic revivalist Gaelic League , [Hyde ] actually argued in the mid - 1890s
that Irish people had been speaking Gaelic for so long that their mouths
were unfitted physically for English " (Garvin 1994, 84).5 Even the most
zealous nationalists of the time took little heed. Less impervious , alas,
were certain members of the Irish Ministry of Justice in the 1940s who
refused to shelter Jews fleeing from Nazi persecution on the grounds that
" they do not assimilate with our own people but remain a sort of colony
of a worldwide Jewish community . This makes them a potential irritant in
the body politic and has led to disastrous results from time to time in
other countries ." 6 Racism may have been rare in Ireland ; it was not
nonexisten t.

A fourth , more generous, understanding of the nation came under the

rubric of the " migrant nation " - or the nation as " extended family ." Here
the definition of the nation remained at least partially ethnic but was
enlarged to embrace all those emigrants and exiles who live beyond the
territory of the nation -state per se. The most celebrated example of this is
no doubt the Jewish diaspora ; but it also pertains to many other emigrant
nationalities , not least the Irish . If over 70 million people in the world
today claim to be of Irish descent, it is evident that this definition of
nationality , or at least of national genealogy , extends far beyond the
borders of a state or territory . Irish Americans or Irish Australians or
Language and Nationalism
5

Irish Britons , for example , can affirm a strong senseof national allegiance
to their " land of origin " even though they may be three or four gen-
erations away from that land and frequently of mixed ethnicity . With the
formulation of a " new nationalism " in the Forum for a New Ireland
( 1984- 85), the then Foreign Minister , Peter Barry , included the Irish em-
igrant population in his definition of the nation (appealing , for example ,
to " Irish nationalists North and South and everywhere else in the world " ),
as did President Mary Robinson when she came to office in 1990 . Similar
appeals are heard whenever the international community is solicited to
support peace in Northern Ireland (the international Ireland Funds , etc.),
or when it is a question of debating emigrant voting rights in national
elections and selecting national football teams.
In short , although the " migrant nation " is still largely ethnically based,
it is far more inclusive than the ethnic nation model in that it embraces

the exiled along with the indigenous .7 By implication , this means accept-
ing that the Irish diaspora comprises not only different Irish peoples
(" Ulster Scots" as well as " native Gaels" ) but , through intermarriage in
the melting pots of overseas continents , different racial confections as
well . One wonders how many of the Irish Irelanders - who sought to
confine the nation to those of Gaelic extraction - would have believed

that by the 1994 World Cup the Irish team would be over half British -
born with a center forward named Cascarino .8

A fifth , and final , variable that merits attention here is what might
broadly be called the " nation as culture ." This is not unrelated to the two
preceding definitions but differs in this crucial respect: the cultural para -
digm of national identity is not reducible to any specific race. National
culture can include many things besides ethnicity : for example , religion ,
language , art , sport , dance, music , cuisine , clothes, literature , philosophy ,
even (some believe) economics . As such, it is by nature far more pluralist
than any strictly racial understanding of nationhood . For its part , the
Irish nation includes several different religions (as Wolfe Tone 's reference
to " Catholic , Protestant , and Dissenter " recognized), two different lan -
guages (Gaelic and English ) , and a rich variety of philosophical tradi -
tions . Indeed , even when cultural nationalism does refer to ethnicity , it is
obliged - if it is to be consistent - to invoke several different ethnic lega-
cies. In the case of the Irish cultural nation this requires acknowledgment
of the Viking , Norman , Scots, and Anglo -Saxon contributions alongside
those of the " ancient Celtic race." Thus , we might say that the exclusivist
equation of " Irish Irish " with Gaelic and Catholic , by D . P. Moran and
 6 Kearney

other fundamentalists in the first decades of this century , was in fact a

betrayal of the full complexity of Irish culture .

Although Gaelic is the official " first language " of the Republic , and an

invaluable resource to the nation , it is not the first tongue spoken by the

vast majority of the population . Nor is it the tongue , as Macnamara re -

minds us without revisionist rancor , responsible for making Irish writing

one of the world ' s great modern literatures . Joyce , Yeats , Wilde , Kava -

nagh , Heaney may well be haunted by Gaelic as their souls " fret in the

shadow of the ( English ) language " - but it is English that they speak and

write , albeit in a singularly Irish way . The language question is even more

vexed , of course , when it comes to politics . Elements in the Gaelic League

did , for a time , pursue the quasi - Herderian thesis that the collective genius

of the race found its true expression in the Irish tongue . Hence the view

that Gaelic equaled rurality , purity , and authenticity whereas English

equaled urbanity , decadence , and vice . The implication here was , of

course , that because Ireland had a unique and ancient language , it had a

natural right to separate political status .

However , several recent commentators , following in the wake of John

Macnamara ' s critical intervention , argue differently . Historian Tom

Garvin notes the irony that the more the Irish embraced English , the more

nationalist they became !

English has always been the language of Irish political separatism , in an extra -

ordinary succession from Jonathan Swift and Clement Molyneux in the early

eighteenth century , through Theobald Wolfe Tone and William Drennan in the

1790s , to John Mitchel , Arthur Griffith , James Connolly and Patrick Pearse in the

nineteenth and twentieth centuries . Political sentiment in the Irish Gaelic language

tended to be rather pre - political , rarely getting further than a sentimental and

rather rhetorical Jacobitism . ( 1994 , 86 )

It certainly seems the case that it was through English that the republican

principles of the French and American revolutions spread through Ireland

in the 1790s , leading eventually to the United Irishmen rebellion in 1798 .

Thus , it has been remarked that the acquisition of English , " far from

making the Irish assimilate to English attitudes and loyalties , rather gave

the weapons of European Enlightenment to a previously impotent tradi -

tion " ( Garvin 1994 , 88 ) . 9 By such accounts , the historical relationship

between language and Irish nationality is far from evident .

The same could be said about religion and Irish nationalism . Although

Catholicism is the majority creed of Ireland , it is by no means the only

one , and besides carries a strong " internationalist " component that COll -
 7
Language and Nationalism

trasts with the notion of a closed insular Volk (e.g., the missionary tradi-
tion of diaspora throughout Europe, Africa , Asia, and the Americas).
That said, one should not underestimatethe powerful role religion has
played in Irish nationalism, and nationalism worldwide. The fact that
nationalism, particularly in its modern republican guises, is an offspring
of Enlightenment secularismhas not prevented its more popular mani-
festations from being fueled by the collectivist, and often atavistic, en-
ergies of traditional religions. Though this connection has been greatly
underrated in much academic literature, recent research by Marianne
Elliott (1996) shows the pivotal symbiosisbetweenIrish nationalism and
Catholicism in our own century and argues that the Ulster Protestants'
rejection of Catholicism dominates their rejection of nationalism.10 Tom
Garvin makes a similar point : although he concedesthat the founders of
the separatistIrish tradition were Protestant rather than Catholic (largely
due to the Protestants' near-monopoly of higher education in Ireland up
to the twentieth century), he observesthat the nationalist rank and file
have been overwhelmingly Catholic since the 1790sand the nationalist
leadership so since the 1900s. " Irish nationalism," he concludes, " has
always been rhetorically anti-sectarian, but is tempted into sectarianism
becauseof its anthropological rooting in the Catholic majority and in the
ideology of dispossessionand repossession " (1994, 89).11 The fact that
Protestantismwas the religion of the dominant Ascendancyclass, which
disenfranchisedthe Catholic majority , was an additional factor in the
equation.
Here once again, history is vexed. One of the main reasonsthe Catholic
hierarchy was not officially allied to Irish nationalism during the eigh-
teenth and nineteenth centuries was that it feared that the nationalist-
republican ideasbeing imported into Ireland from the French Revolution
were anti-Catholic. The fact that theseideaswere also anti-British meant,
logically, that a tacit alliance of interests bound the Catholic hierarchy
and Westminster together: the Catholic hierarchy actually approved the
abolition of the Irish parliament and union with Britain in 1800, and the
English government financed the establishmentof the Catholic seminary
Maynooth in 1795. After the fall of Parnell and 1916, however, it became
clear to the Church that the soul of the Irish nation was up for grabs and
that the need for a unifying collective identity for the newly emerging
state could best be provided by a form of Catholic nationalism that
allowed (in Joyce's words) " Christ and Caesargo hand in hand." Indeed,
 8 Kearney

the 1937 Constitution of Dail Eireann came close, at times, to ratifying

the equation of Catholic, Gaelic, and Irish.
The Irish Republic was not, of course, unique in its early tendency
to accommodate religion as a badge of national identity . North of the
border, one found a " Protestant Parliament for a Protestant People"
emerging, while on the neighboring island the Church of England was still
a primary means of asserting English national identity. And those who
consider the link betweennation and religion to be now no more than an
anachronism need look not only to the continuing sectarianismon both
sides in Ulster, but also to the more global fate of Yugoslavia and the
former Soviet Union .
Finally , although some cultural nationalists of the Sinn Fein school
argued for an " economic nationalism" in the 1930sand 1940s, this ulti -
mately came to grief. Independent Ireland's original policy of national
capitalism- economic self-suffIciency- foundered because Ireland (1)
lacked a sufficient body of industrial entrepreneurs, (2) inherited a bu-
reaucratic and centralized state apparatus from the British, and (3) con-
tinued to look to Britain for ultimate solutions. As Desmond Fennell
remarks in his analysis of economic nationalism, " The state was changed
only to make it more what it was- more centralized, interventionist, un-
coordinated, and uncontrolled by citizens' decisions" (1989, 67).12 Eco-
nomic nationalism also failed, of course, becausethe world economy was
becoming such that nation-stateswere no longer the determining units in
the regulation of wealth but were increasingly subject to a transnational
network of global banks, markets, and industries. The notion of a quin-
tessentiallyGaelic-Catholic-nationalist economy was never more than an
ideological illusion.
Whether it was predicated upon religious, linguistic, or economic
grounds, Irish cultural nationalism has always displayed an unusual
capacity, pace the stereotypes, to change its mind about itself. This is
largely due to the fact that the tradition is extremely complex and its
legacies multifaceted. As testified in the old quip (Garvin 1993, 63):
wheneverthe English thought they'd found the answer, the Irish changed
the question! 13
The one aspect of Irish cultural nationality that I believe John Mac-
namara would have approvedis preciselythis multiple nature of identity: a
wonderfully confusedkaleidoscopeof possibilities that, at best, kept Irish
people at home or abroad always on their toes and always open to others.
Languageand Nationalism 9

Notes

1. On this tension between civic universalism and state nationalism , see Ingram
1995 , Garvin 1993 , and Ricoeur 1995b . As Ricoeur observes , one of the main
dangers of equating nation with state is the eclipse of society behind the state.
Thus we see, for example , how the French Revolution apportioned political
sovereignty to all levels of the community , from the government at the top to the
individuals at the bottom , but allowed the state to become omnipresent in the
process, reducing the citizen to a mere fragment of the state.
2. Fennell develops his analysis of territorial nationalism as follows :
The definition of the nation as a community determined by a territory - the so-called
national territory - was typical of much European nationalism . . . . As a matter of history ,
most European nation -states were constructed by using force , administrative pressures, and
schooling to convert this theory into fact . . . . The primary aim of Irish nationalism , a united ,
self-governing Ireland , was also typically nationalist : the nation inhabiting the national ter -
ritory , so the theory ran , had the right to political autonomy under a single government .
(1989, 20- 21)

3. For example :
The image of the Emerald Isle alone and inviolate in the ocean has itself become a cultural
symbol , a mute argument for the unity of Ireland and its necessary separation from Britain .
This argument from geography has commonly impressed not only Irish separatists them -
selves, but also many British observers and perhaps most third -country observers of the tor -
tured and tortuous relationship , one of mutual love and mutual hatred , between the two
islands . Anti -partitionists in particular commonly justify the Republic 's constitutional claim
to Northern Ireland by claiming that what God and History have united , no man , and cer-
tainly no Englishman , can sunder . (Garvin 1994, 84)
This debate has direct relevance to current discussions of a revised Irish constitu -
tion . De Valera ' s constitution of 1937 described the " Nation " in terms of three
articles . The Irish nation is asserted (axiomatically ) rather than defined in Article
1. It is described in terms of national territory encompassing the entire island in
Article 2 ; and the divided condition of the island is referred to as a temporary
matter , " pending the re-integration of the national territory ," in Article 3. The
subsequent descriptions of the " State" that follow - Articles 4 to 9- are equally
nebulous and ambiguous . Hence the need for a revised constitution setting out a
clear definition of citizenship and nationality , and taking into consideration the
new status of Northern Ireland as a region relating concentrically to the island of
Ireland , the British archipelago , the European Union , and the United Nations . I
am indebted to Philip McGuinness for several of the above comments .
4. See Sluga 1993, 60, 91, 103- 108, on the Romantic philosophy of Germany as
an ethnic nation . In his Addresses to the German Nation ( 1807), Fichte called on
the spirit of the German people to remember " its own true nature " and to realize
that the battle against other nations cannot be " fought without one's household
gods, one' s mythic roots , without a true 'recovery ' of all things German " (sec. 23) .
It was not a huge leap from these musings of the German Romantics to the
argument of a Nazi idealogue like Rosenberg , who , in his speech " The Crisis and
Construction of Europe " ( 1934), called for a rejection of the Enlightenment uni -
versalism that had dominated European thought and politics , concluding , " The
 10 Kearney

point , the idea , the fact from which we must start today is the fact of the nation "

( cited in Sluga 1993 , 60 ) . For further discussion of German nationalism , particu -

larly in the contemporary context where the traditional equation of " people , "

" nation , " and " state " is becoming increasingly confused and questioned , see

Hobsbawm 1990 , 180 - 181 .

The antisemitism , racism , social Darwinism , and antipathy to all forms of

internationalism that typified National Socialism were not , it is fair to say , qual -

ities shared by Irish , British , or French nationalists in any general sense . Although

British and French nationalisms are in principle civic rather than ethnic , recent

responses to the " immigration problem " have revived some concept of the latter ,

albeit frequently at a cultural level - for example , the British Heritage Industry or

the French neoconservative revivalism .

On this vexed distinction between civic and ethnic nationalism , see Ricoeur ' s

pertinent remarks :

The notion of ' people ' according to the French Constitution is not ethnic . Its citizenship is

defined by the fact that somebody is born on . the territory of France . For example , the son or

daughter of an immigrant is French because he or she was born on this territory . So the rule

of membership has nothing to do with ethnic origin . This is why it was impossible to define

the Corsican people , because we had to rely on criteria other than citizenship , on ethnic cri -

teria . . . . The Corsican people are also members of the French people . Here we have two

meanings of the term ' people ' . On the one hand , people means to be a citizen in a state , so

it ' s not an ethnic concept . But on the other hand , Corsica is a people in an ethnic sense -

within the French people which is not an ethnic concept . I think it ' s an example of what is

happening throughout Europe now . ( 1995b , 35 - 36 )

A further conflict of interpretationswith regard to French nationality/ Francophone

identity/ " Frenchness" applies to de Gaulle's famous Cabinet Statement of July
1967, where he spoke of the Quebecpeople as a " population descendedfrom her
own (French) people and admirably faithful to their country of origin." This
effectively contravened the traditional civic, nonlinguistic, and nonethnic notion
of Frenchnessas defined by the French revolutionary constitution. Hobsbawm
(1990, 181) notes the emergenceof similar tensionsin the British definition of cit-
izenship and national identity: " Until the 1960s'Britishness', in terms of law and
administration, was a simple matter of being born to British parents or on British
soil, marrying a British citizen, or being naturalized. It is a far from simple matter
today." For a more detailed discussionof the tension between ethnic and civic
nationalism, seeGreenfeld 1992, 6- 13.
5. Garvin acknowledgesthat the use of race theory has been rare and untypical
" on the part of both separatistsand anti-separatistsin the British/Irish relation-
ship" (1994, 83).
6. Seereport by Andy Pollack, Irish Times, April 15, 1995, basedon researchby
ProfessorDermot Keogh into governmentpapersof the prewar and war period.
7. On this question of the " migrant nation," see contributions by Liam Ryan,
Maurice Hayes, Graeme Kirkham , Kirby Miller , J. J. Lee, George Quigley, and
SeamusHeaney et at. in Kearney 1990. See also Benedict Anderson's lectures
titled " Exodus, Exile and the Nation -State" at Dublin City University and Trinity
College Dublin , December 1, 1995. The notion of a " national people" or popu-
I-,anguageand Nationalism 11

lation is becoming increasingly porous, if not outdated, in many countries today

where three -generation families are a rarity and multiculturalism , pluralism , and
mass migration are eroding the old connection between people and land (as in
Bossuet's seventeenth-century phrase about those who " live and die in the land of
their birth " ) . Increasingly , nationality is becoming something people carry with
them rather than a fixed attachment of a specific place of birth . When one con -
siders, moreover , that some 17 million people crossed the Atlantic to North
America bet \\,Teen the seventeenth and nineteenth centuries , and that millions of
" nonresident aliens " lived and worked temporarily in the United States in the
1980s, with no allegiances as " national citizens " (regarding defense, elections , or
taxation ), one begins to grasp how attenuated the notions of national population
and original homeland are becoming . Understanding nationalism to be one of the
greatest means of commanding the loyalty and allegiance of people , the United
States has deployed various means to " renationalize " the landscape (e.g., the
frontier myth of the " Western " ) and the emigrant " melting pot " population (e.g.,
through a them /us strategy of war , from the old Revolutionary and Mexican Wars
to the two world wars this century and subsequent campaigns in Korea , Vietnam ,
the Middle East , and Latin America ).
8 . An additional variation on the post -nation -state model is what Benedict
Anderson calls " long -distance nationalism ." His basic hypothesis here is that
several forms of contemporary nationalism are the products of exile and exodus -
expatriate populations looking back nostalgically on the lost homeland - rather
than expressions of atavistic / nativist tradition , as is commonly supposed. The
cliche of the " sea-divided Gael " has more significance than one might think . In
this respect, one might cite how it is often expatriate Irish -Americans (frequently
second-generation or more ) who cling to the most " traditionalist " forms of na-
tionalism , as witnessed in NORAID support for the IRA and the controversies
over the participation of Irish gay and lesbian groups in the New York and
Boston St. Patrick ' s Day parades (or , one might add , expatriate funding for the
1995 Anti -Divorce Campaign ). But the nostalgia neurosis that " invents " outdated
models of nationality jnationalism retrospectively can equally be a feature of a
modern nation ' s internal sense of exile . Nationalism thus also reveals itself as a

response to an internal experience of loss- for example , a lost language (Gaelic )

or ancient homeland (the four green fields) . Indeed , the cultural critic Luke
Gibbons goes so far as to argue that the ideal of authentic territorially based na-
tionality is often a fiction of advanced , secular, urban centers. Thus , for example ,
the famous St. Patrick 's Day speech by De Valera in 1943 was, Gibbons suggests,
a traditionalist portrait of the old , pristine nation delivered from the threshold of
modernity . In this sense, notions of national tradition frequently turn out to be an
invention of modernity , as in the American fantasy of the Wild West or the Irish
fiction of the West of Ireland (see Luke Gibbons and Fintan O 'Toole ) . Or to put
this paradox in another way , it is precisely when the old securities of the classic
nation -state- fixed territorial boundaries , contractual citizenship , economic -
political autonomy , civic allegiance , and cultural identity - are threatened , that
we find the need to retrieve and revive so-called primordial nationalism , namely ,
the need to combine a sense of homeland with a sense of accountability . This
12 Kearney

phenomenon of " late nationalism ," or what we might call retro -nationalism , is a
phenomenon that Anderson links with the dislocationary processes of late capi -
talism . The more people feel surrounded by a disenchanted universe - an alien -
ated, simulated , mass-produced world - the greater their need for a return to
a reenchanted universe . Neonationalism seems a response to such a need (as
witnessed in the doubling of newly formed nation -states since World War I ,
accelerated by the breakup of the " Communist bloc " ) . Yet another feature of
" long -distance nationalism " is the internationalization of national politics as evi -
denced in the growing number of expatriate or foreign citizens - usually American
or Canadian - entering the presidential campaigns of post -Communist nation -
states (e.g., Croatia , Poland , Estonia ) . Nor should we underestimate the influential
role played by American -Irish politicians , especially the Kennedys and President
Clinton , in the Northern Ireland peace process ( 1994- 1996) .
9. It was through the use of the English language , as Garvin points out , that " the
idea of an Irish Republic , free of aristocratic privilege , monarchical trappings and
religious distinctions , was transmitted to a popular culture " ( 1994, 86). In this re-
spect language proved to be a far more malleable badge of national identity than ,
say, race or religion . " Language has the great advantage from the point of view of
nation -builders , of being learnable ; unlike one's religion , there is no spiritual price
to pay for 'changing sides' " (Garvin 1993, 72) . See also Macnamara 's ( 1977) more
psychological -philosophical -linguistic approach to this subject .
The complex relationship between language and nation extends beyond the
Irish case and has many repercussions in the emerging European context . As Paul
Ricoeur puts it :

There is no political distribution of borders which is adequate to the distributions of lan -

guages and cultures , so there is no political solution at the level of the nation -state . This is
the real irritator of the 19th century (legacy ), the dream of a perfect equation between state
and nation . That has failed so we have to look for something else. But there is a political
problem here. Is the project of European federalism to be a confederation of regions , or of
nations ? This is something without a precedent . Modem history has been made by nation -
states. But there are problems of size. We have five or six nation -states in Europe of major
size , but we also have micro -nations which cannot become micro -states in the same way as
national states have done . ( 1995b, 34- 35)

This problem of size relates as much to regions as to nations , as Ascherson (1995,

20fT.) points out . Regions in the European context , for example , range from a
German Land like Bavaria to the smaller regions of Italy or Spain where , in some
instances like Catalonia or the Basque region , considerable cultural -linguistic -
educational autonomy and even fiscal -economic devolution is allowed .

10. See also O 'Brien 's ( 1994) trenchant critique of what he calls " sacral national -
ism " ; and see Kearney 1984, Murphy 1988, and Keogh 1988. The relationship
between nationalism and Catholicism resurfaced during the heated debates on
abortion information and divorce in the 1990s, particularly as they affected the
Irish constitution . The narrow passing of the controversial Divorce Referendum in
November 1995 was a particularly significant moment in this ongoing process of
Irish " self-identification ." The reverse side of the Catholic -nationalist equation is
the Protestant -unionist ; on this relationship between Protestantism and Orangeism
see, for example , Moloney 1980.
Language and Nationalism 13

11. Seealso Garvin 1990, Kiberd 1995. An additional reasonfor the reversion of
Irish nationalism to confessionalallegiancemay have something to do with the
more general issue of new nation-states' experiencing a foundational crisis of
legitimation. As Paul Ricoeur points out (1995a, 201): " A la racine du politique, a
son fondement, il y a l'enigmede l'origine de l'autorite. D 'ou vient-elle? C'est une
chosequi est toujours non reglee, et qui fait que I' ombre ou Ie fantome du theo-
logique continue de roder autour du politique" ('The enigma of the sourceof au-
thority lies at the very root and foundation of politics. From what does it derive?
This is something never fully resolved, which means that the political is always
haunted by a theological ghost or shadow').
12. Seealso Neary 1984and Daly 1988.
13. Among these contradictions and complexities, Garvin (1993) points to the
tension betweenthe Anglo-Irish and Gaelic traditions that Yeats sought to com-
bine in a single narrative.
Yeats, himselfa Protestant
, wishedto devisean Irish nationalmyth that providedroomfor
peopleof his own'Anglo-Irish' tradition. Thetroublewasthat thematerialshewishedto use
weremutuallyincompatible , beingderivedfrom Anglo-Irish andIrish traditions, whichhad
historicallyalwaysbeenat loggerheads and, worse,werestill knownto havebeenso. (p. 64)
Another, more general contradiction involves the conflict between modern-
secularistand tribal -traditionalist aspectsof Irish nationalism.
The modernself-contradictory , attemptedderivationsof Irish nationality from Gaelic,
Catholic, or evenAnglo-Irish aristocraticidentitieswould havehorrified and bewildered
Wolfe Toneand the otherFrance-inspiredIrish Jacobinsof the 1790s . The Enlightenment
conceptof 'citizenship
' independentof one's traditionalidentificationhassurvivedin Ireland,
but it hasfacedveryseverecompetitionfrom, in particular, religion-basedidentities. (p. 65)
While conceding that modern political nationalism is a child of Enlightenment
universalism, in principle, Garvin arguesthat nationalism as a cultural phenome-
non (with political consequences ) is far more ancient, and more atavistic, than the
modern universalizedversion.
In Ireland, land hunger, a quasi-racistsenseof common'stock', memoriesof confiscation
anda strongCatholicism , all disguised
ineffectively
in theragsof thedeadGaelicidentity, lie
behindthe theoreticallysecularmodernnationalidentityof English-speakingIreland. The
inheritanceis ambivalent , andhasseveralideologicalrepertoires that may be drawnupon.
(p. 66)

References
Ascherson, N . (1995). Nations and regions. In R . Kearney (Ed.), Statesof mind:
Dialogues with contemporarythinkers on the Europeanmind. Manchester: Man-
chesterUniversity Press, and New York : New York University Press.
Daly , M . (1988). The impact of economic development on national identity. In
lrishness in a changing society (Princess Grace Irish Library 2). Totowa, NJ:
Barnesand Noble.
Elliott , M . (1996). Religion and identity in Northern Ireland. In W . A . Van Horn
(Ed.), Global convulsions : Race, ethnicity and nationalism at the end of the twen-
tieth century'. Albany , NY : SUNY Press.
 14
Kearney

Fennell , D. (1989 ). The revision ofIrishnationalism . Dublin : Open Air.

Garvin , T. (1990 ). Thereturn ofhistory : Collective myths andmodern national -
ism. IrishReview , no.9(Autumn , 1990 ), 16 - 20.
Garvin , T. (1993 ). Ethnic markers , modern nationalisms , andthenightmare of
history . InP. Kruger (Ed.), Ethnicity and nationalism . Marburg : Hitzeroth .
Garvin , T. (1994 ). Nationalism andseparatism inIreland , 1760 -1993 : A com -
parative perspective . InJ. Bermanendi , R. Maiz , andX. Nunez (Eds .), National -
isminEurope past and future . Universidade deSantiago deCompostela .
Greenfeld , L. (1992 ). Nationalism : Fiveroads tomodernity . Cambridge , MA:
Harvard University Press .
Hobsbawm , E. (1990 ). Nations andnationalism since 1780 . Cambridge : Cam -
bridge University Press .
Ignatieff , M. (1993 ). Blood and belonging : Journeys intothenew nationalism . New
York : Farrar , Straus &Giroux .
Ingram , A. (1995 ). Constitutional patriotism . Paper presented at17th IVRWorld
Congress , Bologna .
Kearney , R. (1984 ). Faith andfatherland . The Crane Bag, 8(1), 55-68.
Kearney , R. (Ed.). (1990 ). Migrations : TheIrishathome andabroad . Dublin :
Wolfhound Press .
Keogh , D. (1988 ). Catholicism andformation ofthemodern Irishsociety . In
lrishness ina changing society (Princess Grace IrishLibrary 2). Totowa , NJ:
Barnes andNoble .
Kiberd , D. (1995 ). Inventing Ireland . London : Jonathan Cape .
Macnamara , J. (1966 ). Bilingualism and primary education : Asurvey afIrish ex-
perience . Edinburgh : Edinburgh University Press .
Macnamara , J. (1977 ). TheIrishlanguage andnationalism . TheCrane Bag , 1,
40-44.Reprinted in M.-P. Hederman andR. Kearney (Eds .), TheCrane Bag
book ofIrishstudies : vol . I (1977 -81). Dublin : Blackwater Press .
Moloney , E. (1980 ). Paisley . The Crane Bag , 4(2), 23-28.
Murphy , J.A. (1988 ).Religion and Irishidentity .InIrishness inachanging society
(Princess Grace IrishLibrary 2). Totowa , NJ:Barnes andNoble .
Neary ,P.(1984 ).The failure ofeconomic nationalism . The Crane Bag , 8(1),68-81.
O'Brien , C. C. (1994 ). Ancestral voices : Religion andnationalism in Ireland .
Dublin : Poolbeg Press .
Ricoeur , P. (1995a ). Lacritique etlaconviction . Paris : Calmann -Levy .
Ricoeur ,P.(1995b ).Universality and thepower ofdifference .InR.Kearney (Ed.),
States ofmind : Dialogues withcontemporary thinkers ontheEuropean mind . Man -
chester : Manchester University Press , and New York :New YorkUniversity Press .
Sluga, H. (1993 ). Heidegger 's crisis : Philosophy andpolitics inNaziGermany .
Cambridge , MA:Harvard University Press .
Chapter
- 2
Meaning and Misconceptions Anil Gupta

How should we think about the meaning of sentencesin discoursesthat

contain fundamental misconceptions? Reflection on this question reveals,
I shall argue, someflaws in our current ideas about meaning (sections2.2
and 2.3). And it motivates some modifications of theseideas that, I be-
lieve, are intuitively plausible and theoretically fruitful (sections2.4- 2.6).

2.1 An Example

Let us begin with an example of a discoursethat embodiesa fundamental

misconception: discourseabout " up" in a community that believesthere

This chapteris dedicatedto the memoryof my friendJohn Macnamaraand his

loveof philosophy .
I havebeenthinking aboutthe problemsdiscussed in this chapterfor a long
time, andovertheyearsI haveaccumulated a largedebt. My mostsignificantdebt
is to my friend and formercolleagueProfessorMark Wilson. Mark and I have
had numerousdiscussions on the problemsof meaningand truth, and thesecon-
versationshavebeenof greatvalueto me. Also of valueto me weresomeearly
discussions I had with Eric Dalton, Adam Kovach, and especiallyJerryKapus. I
presented someof theideasin this chapterin a graduatecourseon thephilosophy
of languagegivenat IndianaUniversityin spring1997.I alsopresented theseideas
in talks at MIT , National ChungChengUniversity(Taiwan), and Notre Dame
University. I wishto thankmy auditorsfor the feedbacktheygaveme. I alsowish
to thankthefollowingpeoplefor theircomments andsuggestions : AndreChapuis ,
William Demopoulos , MichaelFriedman, JamesHardy, Allen Hazen, Christo-
pherHill , Petervan Inwagen , Ray Jackendoff
, HansKim, RobertKraut, Byeong
Deok Lee, Vann McGee, GeorgeN akhnikian, Alvin Plantinga , Jerry Seligman,
Robert Stalnaker , and BosukYoon. Finally, I wish to thank the National En-
dowmentfor the Humanitiesfor support. An NEH Fellowshipfor University
Teachersenabledme to devotea part of the academicyear 1995 - 1996to this
chapter.
 16 Gupta

Figure2.1

is an absolute distinguisheddirection in spacecalled 'up'. It will simplify

our discussionif we make the following assumptionsabout the commu-
nity 's use of 'up'. (In other respects, let the use of the term be much like
ours .)
(i) Let us assumethat two criteria govern the use of 'up' in the com-
munity's language. One criterion is perceptual: members of the commu-
nity recognizethe assertionof
(1) a is up above b
to be warranted in certain perceptually distinguishable situations. The
other criterion is conceptual: members of the community recognize an
assertionof (1) to be warranted when it is basedon the premises
(2) c is up above d
and

(3) The directionof the ray dc is the sameasthat of the ray b71
.
(Seefigure2.1.) Furthermore, they recognizethe denialof (1) to be war-
rantedwhenbasedon (2) and the denialof (3). Thesetwo criteria, let us
assume , are of equalimportancein the usesof 'up'. Both criteria come
Meaning and Misconceptions 17

Standard Up
= the direction of

Satellite
s theraypi

Mountain
peak p

Figure2.2

into play to an equal degree, and neither can be given priority over the
other.
(ii) Let us also assumethat the community usescertain objects to de-
fine the " Standard Up " direction. The community, let us assume, defines
the Standard Up to be the direction of the ray jiSdetermined by a moun-
tain peakp and a natural satellite s located in a geostationaryorbit above
p . (Seefigure 2.2.) So, the community's conceptual criterion can also be
reformulated as follows:
-+
a is up above b iff the direction of the ray ba is the Standard Up .
This criterion, let us imagine, is so strongly embeddedin the community
that the two sidesof the biconditional are regardedas equivalent ways of
saying the samething.
(iii ) Let us assumethat the community engagesonly in Plain Speech.
There is in its discourseno subtle exploitation of conversationalmaxims
to communicate one thing while saying another.1 The content of, for ex-
ample, an assertionmade by using 'a is up above b' is preciselywhat the
conventional meaning of the sentencewould dictate it to be. The as-
sumption of Plain Speechwill enableus to focus on the main issuesbefore
us, freeing us from irrelevant distractions.
 18 Gupta

(iv) Finally , let us assumethat the main conventions and facts about
" up" and other related subjectsare common knowledge. All membersof
the community are (and are recognizedto be) authorities of equal stand-
ing on these subjects.2 This assumption, like the previous one, removes
merely extraneousfactors from our deliberations. We lose nothing essen-
tial by making it .
The question before us is how to think about the meaningof sentences
such as 'a is up above b' in the community's language. It will be necessary
to address also the parallel question about the contents of speechacts
(e.g., assertions) and of attitudes (e.g., beliefs) that the speechactsexpress.
The meaning of a sentencewill be assumedto be fixed; it will not vary
from context to context. But the content that a sentenceis used to express
cannot be assumedto be fixed, and I shall explore both possibilities with
respectto it . In one group of theories I shall consider- namely, the abso-
lute theories- content will remain fixed through all contexts. Thesetheo-
ries will identify meaning with content. In the other group of theories I
shall consider- namely, the relativistic theories- content will vary with
context. These theories will draw a categorial distinction betweenmean-
ing and content.
The concept of meaning is called upon in current philosophy of lan-
guage and mind to servemany functions. Thesefunctions are so diverse
that one may be excusedfor thinking that no concept can servethem all,
that different notions of meaning are neededfor different kinds of ends.
The notion- or aspect- of meaning of interest here, I should stress, is
one that yields a true/false distinction (or something similar such as the
warranted/unwarranted distinction). The account of meaning (and con-
tent) should yield a satisfactory assessment of speechacts and of practices
found in the community. The account should provide a way of separating
those assertionsthat are true (or warranted) from those that are false (or
unwarranted). And it should provide a way of separatingthose inferential
practices that are sound (or adequate) from those that are unsound (or
inadequate).

2.2 Conceptual
-RoleSemantics

Philosophyof languageofferstwo broadapproaches to gainingan under-

standingof meaning : the representationalapproachand the conceptual -
roleapproach. Therepresentational approachattemptsto explainmeaning
by invokinglanguage -world relations. On this approach, the meaningof
Meaning and Misconceptions 19

a linguistic element is- or is constructed out of- what the item repre-
sentsor would representin various contextsand possiblesituations. Thus,
on this approach, the meaning of a proper name (e.g., 'Socrates') might be
identified with the referent of the name (the man Socrates)3 and the
meaning of a predicate F might be identified with the property F repre-
sents. The vocabulary this approach favors for the assessmentof speech
acts is that of the true and the false. An assertion of 'Socratesis F ', for
example, is evaluated as true if Socrateshas the property representedby
F; otherwise, the assertionis evaluatedas false.
The second, conceptual-role, approach attempts to explain meaning via
the rules governing the proper use of language. On this approach, the
meaning of a linguistic element is given by the rules that specify the ele-
ment's conceptual role in the linguistic system. Thus, on this approach,
the meaning of a predicate F might be given by rules that state the con-
ditions under which predications of F are warranted (" F-introduction "
rules) and by rules that state the conclusions, both discursiveand practi-
cal, that may be basedon thesepredications (" F-elimination" rules).4 The
vocabulary this approach favors for the assessmentof speechacts is that
of the warranted and the unwarranted. An assertion of 'Socratesis F ',
for example, is evaluatedas warranted, if the conditions under which the
assertion is made are of the kind laid out in the F-introduction rules;
otherwise, the assertionis evaluatedas unwarranted.s
The two approachesto meaning rest on vastly different pictures of lan-
guage. Still, the assessmentof speechacts they yield are similar: the true/
false distinction coincides with the warranted/unwarranted distinction
over a large domain. This is not surprising, for both the language-
world relations of the representational approach and the introduction-
elimination rules of the conceptual-role approach have their foundations
in the actual use of language. Over " decidable" assertions(i.e., assertions
whosetruth or falsity can be settledby the usersof the language), the two
approachesaim to yield the same verdicts. The debate betweenthe two
approachescenters on the " undecidable" assertions. Here the language-
world relations of the representationalapproach may yield an assessment
whereasthe introduction-elimination rules of the conceptual-role approach
may fail to do 80.6
Although they illuminate much about language, the two approachesdo
not yield a solution to the problem of meaning before us. This is easiestto
show for the conceptual-role approach. This approach takes the meaning
of 'up' to be given by the rules governing its use- rules such as the per-
 20 Gupta

ceptual and conceptual criteria introduced above . The difficulty is that

theserules do not yield a satisfactory assessmentof assertions. For many
pairs of objects a and b, the rules dictate that the assertion of 'a is up
above b' is warranted and also that its denial is warranted . Suppose a and
~

b are objects such that the direction of the ray ba is epistemically accessi
-
ble to the community . The near spherical shape of the earth ensures that
~

objects c and d can be found (or placed) so that the direction of the ray dc
~

is the same as that of the ray ba and , furthermore , the perceptual criterion
for 'c is up above d ' is satisfied. The conceptual criterion now yields that
the assertion of 'a is up above b' is warranted . Similarly , the conceptual
criterion can be made to yield that the denial of 'a is up above b' is war - ~

ranted : objects e andfcan be found so that the direction of the rayfe is

~

not the same as that of the ray ba and , furthennore , the assertion of ' e is

up above f ' is deemed warranted by the perceptual criterion (see figure

2.3) .
This argument , it should be observed , reads the criteria for ' up ' in a
strong way : as laying down what assertions, actual and possible, would be
warranted in various situations irrespective of the presence of a warrant -
supplying intelligence . Thus , it reads the perceptual criterion as stating

Figure2.3
Meaning and Misconceptions 21

that in certain perceptually distinguishablesituations ' a is up above b' is

warranted; whether someoneactually experiencesthe situation and makes
the assertionis irrelevant for the application of the criterion. This reading
of the criterion is essentialif the conceptual-role approach is to yield, even
in unproblematic cases, a satisfactory demarcation of speechacts. On a
weaker reading of the perceptual criterion (e.g., one that requires the sit-
uation to be actually experiencedfor ' a is up above b' to be warranted),
the warranted/unwarranted distinction diverges from the true/false dis-
tinction in completely unacceptableways.
The argument assumes , I should note, that the community has the
ability to determine the identity and distinctnessof the directions of cer-
tain rays. This, it seemsto me, is a harmless assumption and changes
nothing essentialin the example. In particular, it does not erasethe pos-
sibility of the misconception about 'up': the community may harbor its
misconception, for it may simply not have undertakenthe expeditionsand
the experimentsnecessaryto exposethe misconception.
The conceptual-role approach, then, assesses far too many assertionsof
the form " x is up above y" in the sameway. It attributes to them an in-
coherent content, a content determined by empirically incoherent con-
ceptual rules. But, as the following examples illustrate, there can be a
significant true/false distinction (or warranted/unwarranted distinction)
among speechacts even in the presenceof a fundamental misconception.
The conceptual-role approach provides no resourcesfor making the dis-
tinction , however.
The lamp example Supposethat two roommatesA and B in our imagined
community are debating which of the two lamps in their kitchen needs
repair. A says that it is the one up above the stove, and B denies this.
Supposethat as a matter of fact the broken lamp is the one used to illu -
minate the stove. Now we should assessA 's assertionto be warranted (or
true) and B's to be unwarranted (or false).7
The Vishnuexample Supposethat the community has a primitive obser-
vatory that studies astronomical phenomena. The astronomers at the
observatory record and predict- as far as they can- positions of astro-
nomical objects. Their predictions are circulated in the local community
and are usedby the more curious membersin their own study of the night
sky. The astronomers, let us suppose, use their earlier observations to
predict the relative positions of a bright astronomical object that the
community calls 'Chandra' and another faint one the community calls
'Vishnu'. Their prediction " Vishnu will be directly up above Chandra at
 22 Gupta

9 :00 p .m . today " is distributed , and is used by some astronomically in -

clined individuals to locate Vishnu in the night sky . Now , if the direction

of the ray from Chandra to Vishnu is in fact the same as that of the local

up direction - that is , if the astronomers ' prediction is a good guide for

the community members - we should assess the astronomers ' claim to be

warranted ( or true ).

Incoherence in conceptual rules , then , does not erase the true / false

( or the warranted / unwarranted ) distinction between speech acts . The

problem before us is to find an account of meaning and content that will

enable us to draw this distinction even when the discourse contains radical

misconceptions and the resulting incoherence .

2 .3 Representational Semantics

Let us now turn to the representational approach . The key question here

- one to which it is difficult to find a good answer - is what relation ' up '

represents in the community ' s language . Let us consider some possible

answers .

(i ) Suppose it is said that ' up ' represents the relation R , where

x bears R to y iff the direction of the ray jiX is the Standard Up .8

This suggestion respects the conceptual criterion , but it neglects the per -

ceptual criterion . As a result , it yields wrong assessments of many asser -

tions ; for instance , it yields that A ' s assertion in the lamp example is false
and B ' s assertion true . 9

( ii ) Suppose it is said instead that ' up ' represents S , where

x bears S to y iff the straight line joining x and y passes through the

center of the earth 0 and the ray Y; points away from 0 ( equivalently ,

iff the directions of the rays Y; and 0; are the same ).

This suggestion encounters problems similar to those facing the previous

proposal . It respects some of the applications of the perceptual criterion ,

but it neglects the conceptual criterion . It also yields wrong assessments of

many assertions ( e .g . , the astronomers ' prediction in the Vishnu example ).

One response to these difficulties is to insist that ' up ' represents Sand

that the troublesome phenomena can be explained away . But the explan -

atory burdens that the insistence entails are , in my view , unsustainable :

( 1) An explanation has to be provided why it is S - as opposed to Rand

various other relations - that constitutes the proper semantics for ' up ' .
Meaning and Misconceptions
23

There must be something about the use of 'up ' that makes S the proper
interpretation , but our example provides nothing to distinguish it as
proper . (2) An explanation has to be provided how certain sentences
evaluated as false by the proposed semantics (e.g., the astronomers ' pre-
diction ) are nonetheless good guides for action - whereas their true neg-
ations are poor guides. (3) This explanation will have to be quite different
from the explanation given for those more straightforward cases in which
sentences deemed true by the semantics are good guides for action . An
account has to be given of this difference in the two explanations . Why
are certain instancesof successfullinguistic behavior explainedin one way
and others in a completely cliffe' ent way ?
(iii ) Suppose it is said that '1I.p ' represents the conjunctive relation R &
S , where

x bears R & S to y iff x bears R to y and x bears S to y .

This suggestion lands us in an " error theory " for 'up' : nearly all assertions
of the form " x is up above y " are evaluated as false; only if a and bare
collinear with the mountain peak p and the geostationary satellite s (see
figure 2.2) can the assertion " a is up above b" be true . Thus , the theory
fails in the same way that the conceptual -role approach failed : it does not
yield a significant true / false distinction . (This kind of difficulty attaches, it
seems to me , to all error theories . Hartry Field and J . L . Mackie are well
known for their advocacy of error theories for , respectively , mathematics
and ethics (see Field 1980; Mackie 1977) . Field regards mathematical
statements to be untrue and Mackie takes the same view of ethical claims .
However , even if all the premises on which Field and Mackie base their
views are granted - principally , that mathematical statements are bur -
dened with an unacceptable Platonism and ethical statements with an
unacceptable claim to objectivity - their error theories will remain un -
acceptable until these theories are shown to yield a significant true / false
distinction .)
(iv ) Suppose it is said that 'up ' is ambiguous between Rand S. This is
also unsatisfactory . First , the postulated ambiguity is not discoverable by
the community members through reflection on their language . Second, R
and S are plausible candidates for the semantics of 'up ' in only a few of its
uses; other uses demand other relations . Hence, a manyfold ambiguity in
'up ' will have to be postulated - something that is plainly unattractive .
(v) Suppose the idea of indeterminacy is tried in place of ambiguity .
The suggestion now is that the semantics of 'up ' is indeterminate , that 'up '
 24 Gupta

is to be assigned a class of interpretations including R , S , and others . But

this is not an improvement over earlier ideas . It implies that virtually all

assertions of the form " x is up above y " are neither true nor false . We are

landed again in an error theory .

Let us note , finally , that this and the previous difficulties are not over -

come by the idea that we assign truth - conditions to ' a is up above b '

directly , bypassing the specification of a relation for ' up ' .

So far we have considered only absolute theories . These theories take

all uses of ' a is up above b ' to have the same content irrespective of the

context of use . Let us now consider whether the difficulties we have en -

countered can be overcome if we allow content to vary with context .

Perhaps the most natural way of relativizing content here is to view ' up '

as being implicitly indexical - an indexical like ' I ' whose interpretation

varies systematically with context . We can spell out this variation as

follows . Suppose we view the context as supplying the location of dis -

course , much as it supplies such elements as speaker and time for the

interpretations of ' I ' and ' now ' . We can use the location to define " the

standard up direction relative to the context c " : this is the direction of

- +

the ray ol determined by the center of the earth 0 and the location l supplied

by the context c . And we can say that , in a context c , ' up ' represents the

relation T c , where

x bears T c to y iff the direction of the ray yX is the standard up relative

to c .

On this view , the relation represented by ' up ' is determined by both the

perceptual and the conceptual criteria governing ' up ' , but the application

of the perceptual criterion is limited to the location of the discourse . Each

location yields its own standard up direction and its own distinctive

ordering of objects as constituting the relation " up above . "

In contrast to the absolute theories , the relativistic theories make a

categorial distinction between the contents expressed using ' a is up above

b ' and the meaning of ' a is up above b ' . The content expressed by ' a is up

above b ' varies from context to context , but the meaning of ' a is up above

b ' does not vary . Further , the content can - but the meaning cannot - be

evaluated as true or false . One can think of the content expressed by ' a is

up above b ' in a context c as a structured item built out of the denotations

of ' a ' and ' b ' and the relation represented by ' up ' in the context c . And ,

following David Kaplan , one can think of the meaning of ' a is up above

b ' as a rule ( or function ) that , given a context c , yields the content ex -

pressed by ' a is up above b ' in c .

Meaning and Misconceptions
25

The indexical view has the virtue that it yields a nonvacuoustrue/false

distinction for assertions. But the view faces severalrelated- and, in my
opinion, overwhelming- difficulties. First, the view does not draw the
true/false distinction in the right way. In the lamp example, if the debate
betweenA and B is conductedin the vicinity of the kitchen, then the view
does yield the right truth -values for A 's and B's assertions- namely, the
value " true" for A 's assertionthat the broken lamp is up above the stove,
and the value " false" for B's opposite claim. However, if the debate is
conducted away from the kitchen, then the view yields assessmentsex-
actly opposite to the proper one: A 's assertion is evaluated as false and
B's assertion as true. For, now, the standard up relative to the debate's
location differs from the standard up relative to the kitchen. And relative
to the debate's location, the broken lamp does not lie in the standard up
direction above the stove.
Theseconsiderationspoint to a second, related difficulty . The indexical
view makes assessmentof A 's and B's assertionsdependenton the loca-
tion of their debate. As a consequence , the view yields a negative assess
-
ment of somepractices- for example, reassertionand appeal to authority
- that are fruitful and proper. It is easy to imagine that the practice of
reassertingsentencessuchas 'The broken lamp is up above the stove', and
of passingthem from one mouth to another (regardlessof location), can
do useful work in the community. It may help, for example, in guiding a
third party to repair the right lamp. According to the indexical view,
however, reassertionof 'a is up above b' in a different location is always
as fallacious as the reassertionof 'I am hungry' on the basis of another's
assertion of the sentence. The indexical view explains the successof the
practice of reassertionas due to a lucky accident: a seriesof fallacious
moves combine to yield a happy result. Such an explanation is plainly
unattractive. No fallacy is committed by A (or by B), if he reiterateshis
claim (possibly in a different location), or if others do so on the basis of
his authority . 10 An adequate account of meaning should respect and
explain this phenomenon. The successof the practice of reassertionis too
ubiquitous and commonplaceto warrant an imputation of error.
There is a third difficulty that highlights yet another aspectof the pre-
vious problem. Sentencesof the form ' X believesthat a is up above b',
,X desiresthat a be up above b', and the like are complete as they stand
(neglecting tense as before). In this respect, they differ markedly from
sentencescontaining genuineindexicals, such as 'X believesthat sheis not
trustworthy' and ' X desiresthat Fred sit over there' . The latter sentences
 26 Gupta
are incomplete : unless semantic values are specified for indexical terms in

them , these sentences cannot be interpreted as making a definite claim or

as expressing a proposition . The indexical view predicts that sentences

such as 'A believes that the broken lamp is up above the stove ' are simi -

larly incomplete . But this plainly does not fit the facts . It is a plausible

principle governing belief that the content of a belief is the content of any

sincere expression ( actual or potential ) of it . A 's belief receives sincere

expression in his assertions of ' The broken lamp is up above the stove '.

So the indexical view implies that the content of A 's belief changes as

A moves from location to location . This yields the undesirable result that

the persistence of A ' s belief through his meanderings is as erroneous as the

persistence of A ' s belief " Here is Mooresville " when he has long left the
town behind on Route 67 . 11

The essential problem with the indexical view , then , is the same as that

with the absolute views considered earlier . The indexical view finds only

error where there is in fact much truth , and only fallacy where there is in

fact essentially sound practice .

2 .4 Conceptual Disengagement

Let us gather together what we should accept from the theories consid -

ered above . First , we should accept the absolutist idea that 'a is up above

b ' expresses a proposition . F or , as just observed , ascriptions of proposi -

tional attitudes such as 'X believes that a is up above b ' and 'X desires that

a be up above b ' are complete as they stand . Their complement clauses

express definite contents without needing values for any parameters . It

follows that we should accept also that ' up above ' expresses a binary

concept ( neglecting time as before ) . This concept takes as arguments two

individual concepts - or individuals , if one prefers a Russellian semantics

for names over the Fregean - and yields a proposition . 12 Second , we

should concede to the absolutist that , in one sense , the proposition " a is

up above b " is not true or even perhaps truth - apt . 13 That is , in one sense ,

all assertions of the form " x is up above y " are infected with error . Third ,

notwithstanding this concession , we have to recognize that there are im -

portant differences among these assertions , and we need to find a way of

making sense of the differences . To do so , we shall have to accommodate

relativistic ideas .

The difficulties we have had in making sense of the " up / down " dis -

course have their source , it seems to me , in the idea that the elements of
Meaning and Misconceptions
27

our language have fixed interlocking conceptual connections with each

other - connections that are all invoked in the uses of the elements . It is
this idea, irrespective of whether we follow the lead of the absolute or of
the relativistic theories , that blocks our way to making sense of the " upj
down " discourse . If we follow the absolute theories , the idea makes it

impossible to see how an assertion such as " The broken lamp is up above
the stove" could be true . For the assertion conceptually implies numerous
claims that are plainly false- for example , " The direction of the ray that
begins at the stove and goes through the broken lamp is the Standard
Up .,,14 The same holds of assertions of the form " x is up above y " gen-
erally . So the absolute approach lands us in an error theory for the dis-
course. We can avoid the error theory by shifting to the relativistic
approach . But matters do not improve if we stick with the idea of rigid
conceptual connections . For we are now forced to deny that sentences
containing 'up ' have problematic conceptual connections and , thus , that
there is conceptual incoherence in " up ." The indexical theory considered
above goes so far in this direction as to say that the sense of ' up ' , like that
of 'he' , is not rich enough to yield assertible contents ; some contextual
information must be supplied before sentences containing 'up ' express
complete propositions . But to say this is to recoil from the phenomenon
confronting us to the opposite extreme . The fact is that the sense of 'up ' is
not at all poor like that of the indexicals . The sense is actually far too
rich - so rich that it constitutes a problem . If we stick with the idea of
rigid conceptual connections , we are confined to unpalatable choices:
error theory and the denial of the phenomena at hand .
Not all conceptual connections are relevant - not all are invoked , not
all come into play - in every use of language . This is the lesson we should
draw from the lamp example . Given the way the assertion " The broken
lamp is up above the stove" is used, and is expected to be used, the rele-
vant conceptual component is that provided by the perceptual criterion
for 'up ' . The noffi1al use of the sentence might be something like this : the
sentence is passed from person to person till it reaches a repairman , who
then uses it to perceptually locate the lamp he will repair . Given that this
is how the sentence is used, assertions of 'The broken lamp is up above
the stove' are good and fruitful guides to action - they are true . Asser-
tions of the negation are poor and misleading guides to action - they are
false. This assessment of truth -value is one at which the community
members themselves will arrive once they know the relevant facts . And
the assessment would survive any unmasking of their misconception .
 28 Gupta

Even a god with a sweeping and eternal view of the world would concur
with their assessment. In short , the assessment is neither parochial , nor
fleeting , nor unstable . It is one that an account of meaning needs to
respect and accommodate .
Every assertion of the form " x is up above y " contains error . The as-
sertion of the sentence 'The broken lamp is up above the stove' can be
used, it has to be conceded, in ways that conform to conceptual rules but
lead to error . One , less serious , kind of error occurs in idle uses - when ,
for example , one derives from the lamp sentence the false claim that
the direction of the ray that goes from the stove to the broken lamp is the
Standard Up . If this derivation does not materially affect the use of the
lamp sentence, then it is idle .15 A more serious kind of error is also pos-
sible, however . Suppose, for example , that the repairman were to use the
assertion in the following way : The repairman goes on an expedition to
find the direction called 'the Standard Up ' . Then through elaborate
engineering he somehow constructs a pointer at the stove that points in this
direction . Finally , he uses the pointer to locate the lamp he will repair . If
he were to do all this , he might uncover a deep misconception in his
community but he would not find his way to the right lamp . He would
have been misled by the sign. This hypothetical way of using the sign is
utterly eccentric , however .16 And its existence should not call into ques-
tion the idea that the assertion " The broken lamp is up above the stove"
is , in one sense , true .

An analogy will help make this point clear . Suppose we have a flawed
map of a city . The map , let us say, accurately depicts the streets in the
center of the city , but is inaccurate regarding some streets that lie on the
periphery . Plainly , despite the errors , the map remains a good and true
instrument for navigating through the city center . This is so even though
the map could mislead one if it is used to plan an eccentric roundabout
route from one city -center street to another . Similarly , many conceptual
paths issue from 'The broken lamp is up above the stove'; one can use this
sign in many ways to guide one's actions . But some ways of using it are
(in certain situations ) salient and ordinary , whereas some other ways
- of
using it are extraordinary and eccentric . The eccentric uses are valid and ,
indeed, important : they can expose fundamental flaws in our conceptions .
Nonetheless , their possibility leaves intact the need for a distinction be-
tween truth and falsehood that respects actual practice .! 7
As noted above , not all conceptual connections come into play in all
uses of language . This conceptual disengagement or isolation may occur
Meaning and Misconceptions
29

even when we are not explicit- or even clear- about its extent or exis-
tence. Sometimes, however, conceptual disengagementresults becauseof
an explicit stipulation on our part. Consider, for example, our use of the
notion of the celestial sphere- a notion that is a product of an ingenious
synthesisof agesof ancient astronomical observations. We find this notion
useful- even indispensablein some contexts- though we recognizeit to
rest on a deeperror. Our use of the notion is in someways very similar to
that of the ancients: we follow them in speaking of certain parts of the
sphereas constituting the constellationsof the zodiac; we speakof the sun
as being in Capricorn one month and Aquarius another; we infer the on-
set of spring when we seethat the sun has a certain position in the zodiac;
and so on. But we disengageall this talk from various other parts of our
discourse: we no longer infer motion from variation of position; we no
longer derive the true distance betweenstars from their (apparent) loca-
tion on the sphere; we no longer think of the sphereas having a definite
volume; and so on. A farmer in ancient times might have looked at the
starsin the westernnight sky somechilly night and have said, " The sun is
in Pisces; spring will be here soon." A modern child might do the same.
The farmer's assertionis in one senseerroneous- the sensein which it is
seenas fully engagedwith the ancients' conception of the zodiac. But in
another sense - one in which we are concernedwith how the assertionis
used in the farming community- the assertion might be as true as the
child's. The child is explicit about the disengagement ; the farmer is not.
Nonetheless, the farmer's assertion, like the child's, can be assessedas
true and can report a hard fact about the world .
Our language is not a rigid system. Its terms do not stand in fixed
interlocking conceptual connectionswith each other. Engagementof our
conceptsis important for both theoretical and practical purposes. But so
also, I want to stress, is disengagement. Only if we keep thesedual possi-
bilities- the possibility of engagementand of disengagement - firmly in
mind can we make senseof how our languagecan function evenwhen it is
infected with deepmisconceptions.

2.5 Frames

A familiar and good way of thinking about meaning is this. A declarative

sentence,understood as a linguistic type, has meaning. When the sentence
is used (e.g., to make an assertion), the speechact has a particular con-
tent- a content that is fixed by the meaning of the sentenceand by certain
 30 Gupta

features of context. The content, in turn, fixes, in light of the facts, the
truth-value of the assertion. Thus, for example, the sentence'She is in
Indianapolis' has a certain meaning. SupposeI use the sentenceto make
an assertion and that the context of use suppliesa person- say, MJ- as
the denotation of 'she' and 24 January 1997as the time referencefor the
present tense. This contextual information together with the meaning of
'She is in Indianapolis' fixes the content of my assertionto be that MJ is
in Indianapolis on 24 January 1997. In view of the fact that MJ is actually
in Bloomington on 24 January 1997, the content fixes the truth -value of
my assertion to be the False. We can schematizethis familiar way of
thinking about meaning thus:
(Meaning + Context) Fixes =;.. Content;
(Content+ The World) Fixes =}- Assessmentof truth-value.
Theories of meaning often use these schematain reverseto " solve" for
meaning. Thus, the secondschemais used to derive the idea that content
is a function that maps worlds to truth-values. And the first schemais
used to derive the idea that meaning is a function that maps (relevant)
contextual information to content.
This familiar way of thinking about meaning needsa little fine-tuning if
it is to fit the phenomenahighlighted above. We need to distinguish first
of all two types of assessment , which we shall call 'assessmentof absolute
truth-value' and 'assessmentof effective truth -value' . We encountered
instancesof these types in the previous section. The assessmentof A 's
assertion, " The broken lamp is up above the stove," as erroneous is an
assessmentof its absolutetruth-value. The other assessment , namely, that
A 's assertionis true, is an assessmentof its effectivetruth -value. We need
to distinguish also two types of content that are correlated with these
assessments : absolute content and effective content. Absolute content is
simply the old and familiar content under a new name. This content takes
into account all the conceptual connectionsof an assertionand provides
the basis for the assessmentof absolute truth -value. Effective content, on
the other hand, takes into account the conceptual engagementsand dis-
engagementsthat are in effect and provides the basisfor an assessmentof
effectivetruth -value. 18Absolute content captureswhat an act of assertion
is committed to; effectivecontent capturesthe content that is in play .
The two types of assessmenthave their own distinctive meaning-truth
schemata. The version for absolute assessmentparallels the original ver-
sion closely:
Meaning and Misconceptions 31

( Meaning + Context ) Fixes = } Absolute content ;

( Absolute content + The World ) Fixes = } Assessment of absolute truth -

value .

The version of the schemata for effective assessment has to be a little more

complicated because content that determines effective truth - value ( i . e . ,

effective content ) is not fixed by meaning and context - at least , not if

context is understood in the traditional way . Something else is needed ;

and this something else I shall call ' frame ' . Under this terminology , the

schemata for effective assessment are as follows :

( Meaning + Context + Frame ) Fixes * Effective content ;

( Effective content + The World ) Fixes * Assessment of effective truth -

value .

Let us clarify the notion of frame by reflecting on how contexts and

frames differ .

First , let us note that contexts are local but frames are not . Contexts

carry information such as denotations of personal pronouns and demon -

stratives , senses in which ambiguous words are to be understood , and

degrees of vagueness to be tolerated in certain predicates - information

that is provided by the local speech situation . 19 But this local sort of

information will not suffice to fix effective content . For this we need ,

in addition , information about normal or standard uses of sentences -

information that is not localized to any particular speech situation but is

spread across uses of language . Frames , then , are not local in that they

carry this nonlocal sort of information . There is another respect in which

frames fail to be local : frames stay constant through a substantial range of

speech situations . Contexts , on the other hand , tend to be highly variable ,

for the denotations , say , of the personal pronouns are liable to shift

from speech situation to speech situation . Frames thus occupy a place

intermediate between context and language . Language can remain fixed

through variation in frame , and frame through variation in context .

Context is highly local , language is global , and frame lies in between the

two .

Second , contexts and frames have different functions in the fixing of

content . Context often serves , as remarked earlier , to fill holes in content .

Without contextual information , an assertion of , say , ' He is hungry ' is

incomplete . The function of frame is altogether different . Frame helps

determine the content that is in play . So , although context often adds

 32 Gupta

to content, frame often subtracts from it . This difference in function is

reflected in a differencein relationship to the rules of language. Rules of
languagemake referenceto, and partly determine, what constitutes con-
text. The rules governing 'I ', for example, determine that the identity of
the speakeris relevant contextual infonnation , and the rules refer to this
information in specifying the proper use of sentencescontaining '1'.
Frames, on the other hand, do not figure in the statementof the rules of
language. Nor do theserules fix the character of frames. Frames concern
how the rules of language are applied or implemented. Perhapsthis con-
trast can usefully be put as follows: contexts are in part internal to the
rules of language, but frames are wholly external.
Third , context and frame have different connectionswith understand-
ing. Information carried by the context is necessaryfor understanding
what a speaker has said; not so for information carried by the frame.
If one does not know the intended denotations of indexicals used in an
assertion, then one does not understand what has been said. But lack of
information carried by the frame doesnot imply lack of understanding.20
Frames contain information that accountsfor effective usesof language,
and this information is not alwaysnecessaryfor usinglanguage. The point
here is general: knowledge neededfor using an instrument effectively is
not necessarilythe knowledge neededfor understanding the effectiveness
of the instrument.21
Fourth , frames do not always exist. Successfulcommunication implies
the existenceof a sufficiently rich context, but it does not imply the exis-
tence of a frame. Existenceof a frame requires not only successfulcom-
munication but also successfulpractice. Let a speakerin the community
imagined earlier say, 'Up above every star there is another star'. The
speakerexpressesa definite thought that can be grasped, and debated, by
her neighbors. But there may well not be systematicand successfulprac-
tice connectedwith her assertion(and related assertions) to allow a sepa-
ration of absolute content from effective content; there may be no frame
associatedwith her assertion. Existence of requisite frames is a feature
that separatesthe ordinary and the practical from the theoretical and the
speculativein the use of language.
Let these observations suffice for now as an explanation of frame.
Much more can be said- and needsto be said- on the subject. But let
us postpone this to another occasion and turn to some philosophical
consequencesand applications.
MeaningandMisconceptions 33

2.6 Realism and Antirealism

Metaphysical debates about the reality of objects and properties (e.g.,

numbers, theoretical posits, everyday objects, and goodness ) typically
transfoffil into debatesabout the nature of the discoursesabout the re-
spective objects and properties (arithmetic, science, ordinary discourse,
and normative discourse).22Two contrasting pictures dominate the latter
debates. One picture, favored by those who espouserealism, presentsthe
discourseas representational : the elementsof the discourse" hook up with
the world " ; the namesand predicatesin the discourserefer to objects and
properties (including relations) in the world .23 On this picture, the dis-
course engageswith the world: our use of the discourseis seenas guided
and constrained by, and as being responsible to, how things are. The
other picture, favored by those who espouseantirealism, presentsthe dis-
course as nonrepresentational, as not hooking up with the world . This
picture builds on the idea that not all our discoursesserve one and the
same function: reporting how things are. Different discourses- for ex-
ample, mathematics, physics, and everyday discourse- serve, it is sug-
gested, different functions: the point of mathematical discourseis not to
report on how things are in someotherworldly mathematical reality, but a
more practical one of enabling shorter deductions; the point of physical
discourseis not to report on some mysterious invisible universe, but to
enable us to cope with everyday needsand problems in an efficient way;
and eveneverydaydiscourse, it has beensuggested , doesnot aim to report
on an inaccessibleexternal world , but to bring order and manageability
into our chaotic senseimpressions. On the antirealist picture, discourse
does not engagewith the world ; it is not constrained by how things are.
But this doesnot mean that the discourseis completely unconstrainedand
freewheeling. According to the antirealist, the distinctive function of the
discourseprovides a strong constraint on it .
This is, I should stress, a bare sketch of the dominant pictures under-
lying the current debate. The debateitself provides a wealth of alternative
ways of developing, modifying, and defending the pictures. I will not
enter into the details here, for my aim is to point out a problem in the
dominant pictures. This problem in turn suggestsa synthesisthat may be
useful for understandingsome(not all!) of our conceptual activity .24
The realist picture, we have seen, subscribesto the ideas of repre-
sentationalism (R) and engagement -with-the-world (E); the antirealist
picture subscribesto the ideas of antirepresentationalism (not-R) and
 34 Gupta

disengagement -from-the-world (not-E). The typical realist argument (e.g.,

for physics) is based on success : if the objects physics speaks about
(gravitational action at a distance, curved space-time, and such) do not
exist, then- in the words of Hilary Putnam-
it is a miraclethat a theory \\7hichspeaksof gravitationalaction at a distance
successfullypredictsphenomena ; it is a miraclethat a theory which speaksof
curvedspace -time successfullypredictsphenomena . (1978
, 19)
The strength of this argument lies in the transition from successto
engagement -with -the-world (E)- it would indeed be a miracle if physics
yielded helpful predictions and were at the same time disengagedfrom
the world . The rest of the realist argument consists of a move from
engagement -with-the-world (E) to representationalism(R). The typical
antirealist argument, on the other hand, rests on the " queerness " (to use
an expressionof J. L . Mackie's) of the objects, and of the " referring" or
" picturing" relation, required by representationalism. The antirealist thus
rejects representationalismand moves from there to embrace disengage -
ment (not-E).
The realist argument moves from engagement(E) to representational-
ism (R); the antirealist argument from antirepresentationalism(not-R) to
disengagement(not-E). The two arguments share an important assump-
tion (namely, if E then R)- and one that is problematic. The fact that a
discoursegenuinely engageswith the world does not imply that it " pic-
tures" the world . Contrapositively, the fact that a discoursefails to picture
the world doesnot imply that it is disengagedfrom the world . The prob-
lem with the realist and antirealist pictures is that they contain an illegit-
imate link betweenengagementand representationalism.
A useful synthesisof the realist and antirealist pictures is possible. :This
synthesisviews the discourseas engaging with the world and as serving
the function of informing us how things are- in this respect it follows
realism- but it adopts the antirealist idea that the discourseis not repre-
sentational. This possibility is illustrated by our imagined community and
its talk about " up." A representationalsemanticsfor 'up', we have seen, is
not acceptable. But the community's discourseis fully engagedwith the
world . The primary point of the " up" talk is to help community members
navigate their way around the world by making them sensitive to how
things are (or are desiredto be).
The gap between representationalismand engagementcan be made
intelligible through the notion of " frame" (section2.5). Consideragain the
Meaning and Misconceptions 35

example of the repairman and how he locates the broken lamp. Room-
mate A determines the position of the broken lamp perceptually and
arrives at the judgment that the broken lamp is up above the stove. He
utters the words 'The broken lamp is up above the stove' to inform others
of the location of the lamp. These words (or their variants) are passed
from mouth to mouth till they reach the repairman, who uses them to
perceptually locate the lamp to be repaired. The sentence'The broken
lamp is up abovethe stove', we have seen, cannot be assigneddeterminate
truth-conditions. However, relative to the pattern of useexemplified, rela-
tive to the frame of use, we may be able to assign 'up above' a definite
relation- perhapsthe relation S defined in section2.3- and the sentence
determinatetruth -conditions. Once we keep the frame fixed, we can rea-
sonably view A 's perceptual interaction with the world as providing A
with information about the relative positions of the lamp and the stove,
information that he passes to the repairman via the various inter-
mediaries. Absolutely speaking, 'up' doesnot representany relation, but
relative to a frame, when the use of 'up' is disengagedfrom some of the
conceptual elements, it may well representa definite relation. Disengage-
ment within the conceptual systemcan help bring about engagementwith
the world .
The successof our linguistic and conceptual practices proves at most
that our assertionshave true effectivecontents. That is, it proves at most
that within their respectiveframes our assertionshook on to the world .
But from this we cannot conclude that our assertionshave true absolute
contents and that our discourse pictures the world . Such a conclusion
would follow only if our linguistic and conceptual practicesconstituted a
monolithic whole, unfragmented into multiple frames. Mere successof
our practicesdoesnot establishthe requisite unity in our thought.
The proposal, then, is to combine nonrepresentationalismand engage-
ment in the following way. We allow a discourse to be nonrepre-
sentational in that the absolute contents expressedby its statementsare
problematic- perhapsthey all contain error; perhapsthey are ill defined.
But we allow the discourseto engagewith the world in that much (per-
haps most) of the discourseis viewed as separableunder severalframes.
Relative to theseframes, the statementsof the discoursehave (effective)
contentsthat can be assessed for truth and falsehood.
This combination of nonrepresentationalismand engagementavoids
some of the weaknessesof both realism and antirealism. Antirealism is
prone to view our theory construction as being so much storytelling;
 36 Gupta
realism , on the other hand , is prone to view it as so much revealed truth .
Neither tendency is acceptable- except insofar as it provides a counter to
the other . Realists are correct to insist that there is a vast difference
between fictional discourse and theoretical discourse . Fictional discourse

does not - but theoretical discourse does - engage with the world .25
Unlike fictional discourse, a discarded scientific theory - even one that
proves to be deeply erroneous , and one for which we are unable to pro -
vide a representational semantics- generates useful frames and true effec-
tive contents . ( Example : " Don 't touch that . It has high caloric content ." )26
Indeed , sometimes a discarded theory proves such a convenience within
certain frames that we allow its continued use, with the proviso that
the use shall not extend beyond those frames (as in the celestial sphere
example given above). I suspect that this kind of fragmented use is more
prevalent in our language than we realize .
Antirealists , on the other hand , are correct to insist that our present
theoretical terms do not have a special status over those of the earlier
scientists, prophets , and myth -makers : there is little reason to count as
representational our present-day theoretical terms while denying this
status to the terms of the earlier , discarded theories . Our evidence and our
situation in the world are not essentially different from those of our an-
cestors. Our evidence is not so comprehensive , nor our understanding so
deep and clear , that we can rule out the possibility that we ourselves suffer
from deep misconceptions . The perennial problems of philosophy and the
irresistible force of the skeptical arguments stand witness to the fragmen -
tation in our understanding and the narrowness of our evidence. 27
Conceptual disengagement and fragmentation of thought are essential
for creatures such as we are. Had God (or evolution ) equipped us with the
right set of concepts once and for all , had the overall structure of the
world been somehow revealed to us (as Plato and Descartes thought ), had
our inquiries only the simple goal of filling in the mere details in a given , a
priori , picture of the world - had our epistemic position been so fortu -
nate, disengagement and fragmentation might have had no place in the
workings of our language and thought . But , unfortunately , our epistemic
position is a precarious one. We do not approach the task of under -
standing the world with perfect ready -made tools in our hands . We
approach the task bare-handed; we need to make the tools, including the
tools needed to make the tools themselves. We need to make (or discover )
the system of concepts needed to conceptualize the detailed facts about
the world , facts that are essential to our well -being . However , we cannot
Meaning and Misconceptions 37

arrive at the right systemof conceptswithout an adequatefund of facts,

and we cannot arrive at an adequatefund of facts without the right sys-
tem of concepts. If we had the right systemof concepts, we could securely
go about collecting and cataloging facts. If we had a fund of solid facts,
we could securelygo about improving our systemof concepts. However,
we begin our inquiry into the world having neither- neither the right
systemof conceptsnor a fund of solid facts. We need somehow, through
our inquiry , to work our way to both. It is inevitable that in the process
we need to work with imperfect systemsof concepts- systemsthat em-
body deepmisconceptions- and with the resultant distortions of fact. It is
this epistemic environment that makes conceptual disengagementand
fragmentation of thought inevitable and important .

Notes
1. The importance of conversationalmaxims for the study of languagewas first
observedby Paul Grice (seeGrice 1989).
2. So I wish to set aside what Hilary Putnam calls 'the division of linguistic labor '
(see Putnam 1975 ) .
3. And it is so identified by a currently popular theory .
4. The version of conceptual -role semantics I shall be working with is due to
Wilfrid Sellars and Robert Brandom (see Sellars 1953, 1974; Brandom 1994) . Two
distinctive features of the Sellars-Brandom semantics are worth noting . First , it
explains meaning and content in terms of inferential role , as opposed to , say,
functional role in the user's psychology (cf . Harman 1982) . Second, it understands
inferential role to include role in substantive , material inferences, not merely for -
mal inferences .

5. Note that the conceptual -role approach uses an absolute notion of " warrant ."
This notion is connected to , but it is not identical with , the notion " warranted

given that the language user is in such-and-such epistemic situation ." The con -
nection between the two notions is roughly as follows : P is warranted in the
absolute sense if and only if P would be warranted by the rules of language under
idealized epistemic conditions .
6. I follow Michael Dummett jn seeing the primary disagreement between the
representational approach and its rivals to be over undecidable sentences (see
Dummett 1978, 1993) .
7 . Let me stress that the sense in which ' ~ Tarrant ' is used here is not that of per -
sonal warrant . It does not mean " being warranted given one' s epistemic situa -
tion . " In this latter sense , it could well be that both the assertions of A and Bare
warranted - or that neither of them is .

8. Here and below , I make several simplifying assumptions - for example , that
objects can be treated as if they were points and that 'up ' is not vague . Further , I
38 Gupta

suppress the relativity of 'up ' to time . (Some of the complexities of our actual uses
of 'up ' are detailed in Jackendoff 1996.)
9. Assuming , as I shall , that A 's and B 's kitchen is not on the mountain peak p
(figure 2.2) .

10. I am assuming , of course, that there are no other reasons to call into question
A ' s authority on this matter .
11. Another way of putting the difficulty is this : the wish of a person , X , that a be
up above b can be fulfilled only by adjusting the positions of a and b, not by
moving X to a place where the standard up is the direction of ~ !
12. The idea that 'a is up above b' expresses a proposition and 'up above ' a binary
concept will meet resistance from the Platonist and from the Nominalist . These
new entities will seem to the Platonist to be unworthy of a place in the realm of
Platonic Heaven ; to the Nominalist they will seem unworthy of a place even in the
realm of ordinary existence. In response, it may be observed to the Platonist that
the new entities exemplify an important virtue : they do real work in our concep-
tual scheme . Furthermore , whatever flaws may be found in these entities , the same
flaws (and many more ) are found reflected in Truth , whose place in Platonic
Heaven is unquestioned . To the Nominalist it may be observed that the new enti -
ties are not going to make their programs any more difficult to execute. Any
scheme they devise that succeeds in eliminating the familiar abstract entities will
succeed in eliminating the new , unfamiliar ones as well .
13. Propositions are defined in the literature in two nonequivalent ways . Some-
times they are defined as the objects of attitudes such as belief and desire;
sometimes they are defined as objects that are truth -apt . The first definition puts
propositions closer to the representings ; the second puts them closer to the repre-
senteds. I have chosen above to follow the first definition , but the choice is purely
terminological . The important point is that the two definitions define distinct
notions . Which notion ends up winning the label 'proposition ' is not at all
important .

14. Recall we are assuming that the kitchen is not located on the mountain peak p
(see figure 2.2).
15. Another sort of idle use is worth noting . Suppose roommate A informs the
repairman of the location of the lamp by saying ,
(i) The direction of the ray that goes from the stove to the broken lamp is the
Standard Up .
The repairman naturally uses (i) to infer that
(ii ) The broken lamp is up above the stove,
and proceeds in his usual way to locate the lamp to be repaired . The use of (i) is
not now idle , for it informs the repairman of the lamp 's location . But the detour
via directions contained in (i) is idle . For A arrives at (i) through (ii ), and he relies
on the repairman to return to (ii ) from the assertion of (i) . ( This sort of idle detour
via the theoretical is also witnessed in a college cafeteria when a student employee
marks the water container 'H2O '.)
Meaning and Misconceptions
39

16. More eccentric than if at a dinner party a guest should pullout a microscope
to examine the contents of her plate before declaring that the host had served peas.
Deep errors in our botanical theories might be revealed by this chance examina -
tion, but a dinner party is not the time and place to explore the possibility.
17. This conclusion is bolstered by the fact that we, who are aware of the mis -
conception , are willing to ascribe knowledge that the broken lamp is up above the
stove. We would explain some of the successes of the community members in
terms of their possession, and some of their failures in terms of their lack , of this
knowledge . Since knowledge implies truth , it follows that we recognize that there
is a sense in which the assertion that the broken lamp is up above the stove is true .
18 . I use the word ' effective ' here to indicate that the notions of truth - value and
content under discussion serve to explain the practical effectiveness of language .
This is not the place to develop a precise account of effective contents . For the
purposes of this chapter , effective contents can be thought of as being of the same
generaltype as absolute contents. Whether one favors the idea that absolute con-
tents are conceptualroles, or the idea that they are setsof possibleworlds, or that
they are Russellian structured entities , one can take a parallel view of effective
contents . Note that absolute and effective contents , though they can formally be
similar , are, in general , materially different . For example , if content is understood
as conceptual role , effective content will be restricted conceptual role . For another
example , if content is understood as a set of possible worlds , the effective and
absolute contents of an assertion will often be given by different sets of possible
worlds .

19. For a good account of information that might be supplied by context , and
how it might evolve in the course of conversation , see Lewis 1979.
20. Hence , the meaning -truth schemata for effective truth should not be used to
" solve" for meaning , if we wish to preserve a connection between meaning and
understanding .
21. The importance of separating factors that help explain understanding from
those that help explain the workings of language has been stressed by Mark Wilson
( see Wilson 1994, forthcoming a,b) .
22. The literature generated by these debates is vast (see Field 1980; Mackie
1977; Leplin 1984; Sayre-McCord 1988; Devitt 1991 and the works cited there) .
Jackendoff 1991 is a valuable contribution to the realism debate for intentional
discourse . Maddy 1990 and Wright 1992 contain illuminating discussions of sev-
eral aspects of the debates .
23. This claim needs qualification . First , it should be understood to be restricted
to names and predicates that are essential to the discourse . A realism with respect
to physics may be willing to allow that some of the names and predicates found in
physical discourse (e.g., arithmetical ones) fail to refer . Second, the names and
predicates that are essential may not be discernible simply from the surface
grammar of the discourse ; it may be necessary to reveal the deeper logical form of
the discourse . Third , on some varieties of realism , language -world relations count
as representationalif statementsare correlated with states of affairs, bypassing
40 Gupta

worldly assignmentsfor namesand predicates. For example, if a statementsuchas

" The United Statesgovernment has reduced all tariffs on imports" is correlated
with a complicated worldly state of affairs (including the actions of the president
and the various members of Congress), the correlation will count as representa-
tional, even though the terms and predicates in the sentence(and in its logical
analysis) are not correlated with worldly items.
I will continue to usethe simpler formulations of " representationalism" given in
the text, leaving the qualifications implicit .
24. In contemporary philosophy, one finds the dialectic of thesis and antithesis
followed often, not by a useful synthesis, but by a sterile super-antithesis, one
which assertsthat the earlier dialectic is meaningless
. This super-antithesisinvites,
of course, a super-thesis that the dialectic is indeed meaningful. And the focus
shifts to a new debate at a higher level- " higher" not in the senseof nobler or
better but in the senseof being more detachedfrom down-to-earth concernsthat
really matter. The entire foundation for the super-antithesis (with respectto the
realism/ antirealism debate) rests, in my opinion, on misconceptions about the
conceptsof truth and meaning. In essentialrespects , the super-antithesis mirrors,
in content and in the ways it goeswrong, the logical positivist rejection of meta-
physicson the basis of the verification theory of meaning. (The proponents of the
super-antithesis are some of the most vehementcritics of logical positivism. But
they seemto me to exemplify many of the intellectual vices of the positivists, and
few of their virtues. In philosophy wrong attitudes are often more pernicious than
wrong doctrines.)
25. I am setting asidefigurative usesof fictional terms.
26. A distinction is worth noting here: a complete rejection of the vocabulary of
the theory and a complete rejection of the statementsof the theory. The former
does not imply the latter. We reject the use of 'caloric' and 'witches' in our de-
scriptions of the world, but this doesnot require us to reject completely all earlier
statementscontaining these terms. We can, and should, attribute true effective
contents to many of thesestatements.
27. I do not mean to suggestthat nonrepresentationalismis plausible for all the-
oretical terms. I do think that it is attractive to treat all of our theories, both the
earlier rejectedonesand the currently acceptedones, as semanticallypretty much
on a par. All engagewith the world, though somemore fruitfully than others. All
fail to be representational, though again somemore than others.

References

Brandom, R . (1994). Making it explicit: Reasoning

, representing
, and discursive
commitment. Cambridge, MA : Harvard University Press.
Devitt , M . (1991). Realismand truth (2nd ed.). Oxford: Blackwell.
Dummett, M . (1978). Truth and other enigmas. London: Duckworth .
Dummett, M . (1993). The seasof language. Oxford: Clarendon Press.
Field, H . H . (1980). Sciencewithout numbers: A defenseof nominalism. Princeton,
NJ: Princeton University Press.
Meaning and Misconceptions
41

Grice, P. (1989
). Studiesin thewayof words.Cambridge
, MA : HarvardUniversity
Press
.
Harman, G. (1982 ). Conceptualrole semantics . Notre DameJournalof Formal
Logic, 23, 242- 256.
Jackendoff , R. (1991). Theproblemof reality. Noils, 25, 411- 433.
Jackendoff , R. (1996 ). The architectureof the linguistic-spatial interface
. In
P. Bloom, M. Peterson , L. Nadel, and M. Garrett (Eds.), Languageand space
(pp. 1- 30). Cambridge , MA : MIT Press .
Leplin, J. (Ed.). (1984). Scientificrealism
. BerkeleyandLosAngeles : Universityof
CaliforniaPress .
Lewis, D. (1979 ). Scorekeeping in a languagegame. Journal of Philosophical
Logic, 8, 339- 359.
Mackie, J. L. (1977 ). Ethics: Inventingright andwrong. London: Penguin .
Maddy, P. (1990 ). Realismin mathematics . Oxford: ClarendonPress.
Putnam, H. (1975 ). Mind, language , and reality: Philosophicalpapers, vol. 2.
Cambridge : CambridgeUniversityPress .
Putnam, H. (1978 ). Meaningandthemoralsciences . London: Routledge& Kegan
Paul.
Sayre-McCord, G. (Ed.). (1988 ). Essayson moral realism. Ithaca, NY: Cornell
UniversityPress .
Sellars, W. (1953 ). Inferenceandmeaning . Mind, 62, 313- 338.
Sellars, W. (1974 ). Meaningandfunctionalclassification . Synthese
, 27, 417- 437.
Wilson, M . (1994 ). Canwetrust logicalform? Journalof Philosophy , 91, 519- 544.
Wilson, M. (Forthcominga). Inferenceandcorrelationaltruth.
Wilson, M. (Forthcomingb). Physicastint, non leguntur.
Wright, C. (1992 ). Truth and objectivity
. Cambridge , MA : Harvard University
Press.
Chapter .L 3

On Structuralism in Michael Makkai

Mathematics

For some years , I have been pursuing a program I call the structuralist

foundation of mathematics ( SFM ) ( Makkai 1996 , 1997 , 1998 , to appear ) .

SFM is based on category theory ~ a branch of mathematics founded by

Samuel Eilenberg and Saunders Mac Lane about fifty years ago ( Eilen -

berg and Mac Lane 1945 ) . William Lawvere ' s work ( e . g . , Lawvere 1969 )

has made category theory the basis of a new foundational approach to

mathematics and logic . Lawvere ' s categorical logic ( e . g . , Lawvere 1970 )

and Lawvere ' s and Myles Tierney ' s top os theory ( Lawvere 1971 ) , subjects

that have been extensively developed , are integral parts of SFM .

In relation to other approaches to foundations within category theory ,

the distinguishing features of SFM are , first , that its aims are global , en -

compassing the totality of mathematics , and , second , that it is committed

to a total linguistic articulation in the spirit of Gottlob Frege . Today , we

express the latter commitment by saying that a foundational proposal

has to be presented in the form of a formal theory . At the same time ,

SFM intends to articulate , intuitively and philosophically , a universe of

The author ' s research is supported by Canada ' s Natural Sciences and Engineering

Research Council and Quebec ' s Fonds pour la formation de chercheurs et l ' aide a

la recherche .

I want to thank Bill Boshuck , Michael Hallett , Dusko Pavlovic , Gonzalo

Reyes , and Marek Zawadowski for the conversations they had with me about the

subject of this chapter . The chapter is based on a talk that I gave in the Depart -

ment of Philosophy at McGill University in the fall of 1994 . I also thank Ray

Jackendoff for reading the chapter and for valuable suggestions .

John Macnamara was , for many years , a constant source of inspiration and

encouragement for me . His interests and insights in philosophy have influenced

my thinking profoundly . This chapter is dedicated to his memory .

 44
Makkai

discourse behind the foffi1al theory . It turns out that the latter articulation
involves a fairly novel mathematical metatheory .
There is a structuralist philosophy of mathematics behind the program
of SFM . In this chapter , my aim is to make an initial contribution to the
statement of this philosophy . This structuralism , it seems to me, is quite
different from the structuralism found in the literature of the philosophy
of mathematics . The basis of this difference is , of course . the almost total

lack of attention paid by the philosophical literature to category theory .

Charles Parsons has argued , somewhat tentatively , that structuralism may
be used as a philosophical justification of the classical Cantorian iterative
set theory (Parsons 1990) . The structuralism I have in mind involves a
radically different " set theory ," a new conception of mathematical total -
ities , the formal explication of which involves concepts of category theory ,
(some of which - higher dimensional categories- are being developed at
present). I see mathematical structuralism as identical to a specific con-
ception concerning mathematical totalities - not something that can exist
independently of any such conception and can then be applied to deal
with already given such conceptions . According to a pervasive complaint
(see Parsons 1990) about structuralism , it cannot be an independent
foundation , since it relies on the concept of structure , which , in the final
analysis , is a set-theoretical concept . This complaint also assumes that the
nature of set theory is given before we can discuss what structuralism is.
In this chapter , I will argue for the desirability of a comprehensive
language of mathematics in which only structural properties of mathe -
matical objects can be expressed. The discussion intends to lead to the
conclusion that " structural properties " are identical to " mathematically
meaningful properties ." Paul Benaceraff 's classic paper , " What Numbers
Could Not Be" ( Benaceraff 1965), is in fact a sustained argument to the
same end: it is a plea for a language in which one talks about numbers
only as numbers . Section 3.1 of the present chapter is a commentary on
the first two sections of Benaceraff ' s paper . What separates my views
from Benaceraff ' s is that I believe that the articulation of the desired

structural language is possible . As a matter of fact , what Benaceraff says

in his third section entitled " Way Out ," especially the subsection " Iden -
tity ," points importantly to what I consider to be the solution (compare
the references to my work ) . However , in his subsequent paper " Mathe -
matical Truth " ( Benaceraff 1973), Benaceraff seems to have lost faith in a
structuralist program , at least in the radical form I am advocating . In
section 3.3 of the present chapter , I will try to show that the " standard
Structuralism in Mathematics 45

semantical picture of the world ," to which Benaceraff is (essentially)

committed in his 1973 paper , is in fact fictitious . This conclusion will
involve the thesis that the nature of mathematical objects is not funda -
mentally different from the nature of myriad nonmathematical everyday
abstract objects. The reality of abstract objects is grounded in a structural
language that, although referring ostensibly to concrete, representing
carriers of the abstractions , manages to unambiguously refer to the
abstract objects, by the discipline of staying within the framework of
concepts that are meaningful for the abstract objects as abstract objects.
In this case, reality is grounded in language.
Makkai 1998 is a semi technical introduction to the mathematical work ;
it also contributes to explaining the philosophy of structuralism as I
understand this term . However , there is much more to be said about this

subject, to which I hope to return another time .

3.1 Reality through Language

In his famous 1965 paper , " What Numbers Could Not Be ," Paul
Benaceraff describes two children , Ernie and Johnny , " sons of militant
logicists ," who , before learning about numbers , how to count , and so
on, had been instructed in set theory in the standard Zermelo - F raenkel
formulation . After that instruction , the parents " needed only to point out
what aspect or part of what the children already knew , under other
names, was what ordinary people called numbers " (p . 48). Benaceraff
goes on to describe how this pointing out took place in the case of Ernie ;
but he temporarily and cleverly conceals what actually took place in the
process. The story looks like the familiar one, the one of defining the
natural numbers as the finite von Neumann ordinals , and defining zero
and the successor operation in the familiar ways- the one we teach now -
adays in set theory courses. As Benaceraff puts it :
To recapitulate : It was necessary [for a proper education in arithmetic ] ( 1) to give
definitions of " 1 " , " number " , and " successor " , and " + " , " x " , and so forth , on
the basis of which the laws of arithmetic could be derived ; (2) explain the " extra -
mathematical " uses of numbers , the principal one being counting - thereby in -
troducing the concept of cardinality and cardinal number . I trust that both were
done satisfactorily , that the preceding [description of Ernie ' s education ] contains
all the elements of a correct account , albeit somewhat incompletely . (p . 54)

Then we learn that Johnny underwent a similar education .

 46 Makkai

Delighted with what they had learned , they started proving theorems about num -

bers . Comparing notes , they soon became aware that something was wrong , for a

dispute immediately ensued about whether or not 3 belonged to 17 . Ernie said that

it did , Johnny that it did not . Attempts to settle this by asking ordinary folk ( who

had been dealing with numbers as numbers for a long time ) understandably
brought only blank stares . ( p . 54 )

At this point in the story , we learn what exactly happened in the respective

educations . Indeed , as we thought , Ernie was taught that the natural

numbers are ( Johnny ) von Neumann ordinals . Johnny , on the other hand ,

was told that they are the ( Ernie ) Zermelo numerals ; that is , zero is the

empty set , and the successor of n is { n } , the singleton whose unique ele -
ment is n .

After reflecting on the dilemma , Benaceraff arrives at the conclusion

that

3 and its fellow numbers could not be sets at all . ( p . 62 )

Along the way , he says :

But if , as I think we agreed , the account of the previous section [ in which the set -

theoretic reconstruction of the concept of number was related , albeit with what the

numbers and their operations actually were left unspecified ] was correct - not only
as far as it went but correct in that it contained conditions which were both nec -

essary and sufficient for any correct account of the phenomena under discussion ,
then the fact that they disagree which particular sets the numbers are is fatal to the

view that each number is some particular set . ( pp . 55 , 56 )

Later :

Furthermore , in Fregeanterminology , each[of the two] account

(s] fixesthesense
of thewordswhoseanalysisit provides . Eachaccountmustalso, therefore , fix the
referenceof theseexpressions . Yet, as we have seen , one way in which these
accountsdiffer is in thereferentsassigned to the termsunderanalysis . (p. 56)
And :

Therefore, exactlyoneis correct, or noneis. But thenthecorrectonemustbe the

onethat picks out which setof setsis in fact the numbers . We are facedwith a
crucialproblem: if thereexistssucha "correct" account , do therealsoexistargu-
mentswhichwill showit to be thecorrectone? (p. 57)
Casting his net wider, he then probes the Fregean account according to
which numbersare predicatesor classes , and finds it wanting, for expected
reasonsrelated to the paradoxes, but also becausethere is no grammatical
evidencethat seventeenis a predicate of classes(" has seventeenmembers"
Structuralism in Mathematics 47

is obviously such a predicate , but that is something else than to say that
seventeen is such a predicate ) . He then summarizes :
If numbers are sets, then they must be particular sets, for each set is somepartic-
ular set. But if the number 3 is really one set rather than another , it must be pos-
sible to give some cogent reason for thinking so; for the position that this is an
unknowable truth is hardly tenable . . . . There is no way connected with the refer -
ence of number words which will allow us to choose among them , for the accounts
differ at places where there is no connection whatever between features of the
accounts and our uses of the words in question. . . . [ AJnyfeature of an account that
identifies 3 with a set is a superfluous one, and therefore , 3 and its fellow numbers
could not be sets at all . ( p . 62)

Benaceraff emphasizes that for the purposes of explication , one may very
well wish to temporarily identify numbers with particular sets and show
that we can do with the sets what we now do with numbers . However , as

he says :

It is . . . obvious that to discover that a system of objects will do cannot be to dis-

cover which objects the numbers are. ( p . 68)

Benaceraff mentions a result due to G . Takeuti according to which set

theory is in a strong sense reducible to the theory of ordinal numbers .
No wonder numbers are sets; sets are really (ordinal ) numbers , after all . But now,
which is reall ~v which ? ( p . 68)

What are we to make of Benaceraff ' s argument ?

The first reaction may be to say that the argument invalidates the usual
set-theoretic treatment of numbers . I want to argue that this is not so-
that in fact , the set-theoretic Platonist engaged in the foundations of
mathematics and Benaceraff can agree without compromising their posi-
tions in any essential way . A perfectly satisfactory position the standard
Platonist may adopt seems to be that , indeed, set theory is incapable of
telling us what the numbers are; but in fact , it does not aim at doing that
in the first place. What set theory wants to do , and does, is explain what
\ve mean when we talk about numbers. It gives us a systematic way of
translating " ordinary " mathematical statements involving numbers into
set-theoretic statements, and it provides reliable methods for proving and
refuting such statements. Indeed , and this is an important additional fact
contributing to the explanatory power of set theory , it does its work in a
remarkably simple particular manner , by identifying numbers and other
mathematical objects as certain sets, such that the original statement -
say, about numbers - is translated into its set-theoretical version by sim-
 48 Makkai

ply replacing reference to the mathematician 's numbers by reference to

their set-theoretical surrogates . But note , set theory does this without
creating any prejudice concerning the issue of what the numbers really are,
or , for that matter , whether they exist at all ! Further , standard set theory
can do more than that : it can give a theory of what constitutes a satisfac-
tory surrogate for the notion of number . This exemplifies a very impor -
tant point about a good foundation : namely , the requirement that it should
have the tools to reflect about itself to the largest extent possible . How set
theory does this in relation to the concept of natural number will be dis-
cussed below , although the facts are familiar ; I have in mind the (second-
order ) Peano axioms for the natural numbers , and their categoricity .
I conclude that from the set-theoretic point of view , Benaceraff 's argu -
ment is not a problem . In fact , it may be interpreted as a support of the
standard foundations , arguing that set theory is needed precisely to pro -
vide the right (because true) paraphrase of the pretheoretical language
concerning numbers , and other dubious entities like them , a paraphrase
that ensures the correctness of the language globally , without committing
its user to any ontology - in fact , freeing the user from all commitment to
any ontology !- concerning the " ordinary mathematical objects" such as
numbers . Of course, it does this at the obvious price of exacting the
ontological commitment to sets themselves; but who is unwilling to submit
to such a discipline when the resulting advantages are as great as they in
fact are ? The set theorist treats numbers as second - class existents , the first -
class existents being the sets; numbers are the results of conventions ,
which , by their being conventions , are arbitrary .
Incidentally , I find that the response to " Benaceraff 's challenge " in
Penelope Maddy 's book Realism in Mathematics (Maddy 1992), pressing
as it does the interpretation that numbers are properties (see p . 86), is on
the wrong track . If set theory does anything well , then that is the job of
interpreting mathematical entities as objects of the theory , that is, sets.
Interpreting entities as metatheoretical ones such as properties is in -
herently inferior to interpretation -as-objects . Also , I do not think that
Maddy is answering Benaceraff 's arguments against precisely this move .
What , then , is the force of Benaceraff ' s argument ?
As I said , the set theorist and Benaceraff may agree on the status of
numbers to the effect that numbers as first -class objects either do not
exist at all , or they do , and they are not sets; in short , they may agree on
Benaceraff 's main conclusion that numbers are not sets. But it is equally
clear that- they do not agree with respect to their overall positions . The
Structuralismin Mathematics 49

differenceis Benaceraffs implicitcommitmentto the thesis that there

mustbe an accountof numbersthat treatsthemasfirst-classentities,that
treats numbers as numbers.
Admittedly,
thisthesisis to a largeextentnormative;
it maybeignored
bythosewhodonotcarefortheunderlying imperative.
Thesettheorist
mayjust say,Go ahead,try to findyourpreciousaccountof numbers,
butI thinkyouarewastingyourtime;I personally
donotcarea bitforit,
even if it can be found, becausemy accountof number-talkis perfectly
sufficient. Is there anythingmore compellinghere than a possiblyinter-
esting,butultimately unnecessary,philosophical
suggestion?
I believe
the
answeris yes;it is therealityofthelanguage
ofmathematics
aswefindit
in practice.
There is a conflictbetweenthe languageof mathematicalpractice(the
coarsetalk)and the languageof settheoryas it appliesto numbers(the
refinedtalk).Interestingly
enough,thisconflict is not that settheoryis
unableto expressthingsthat the practitionerswant to, and can, express,
which would be the natural thing to expect in a relationship between
an existingpracticeand an a posteriorifoundational reconstructionof
thatpractice.Theconflictis of an oppositenature:thepractitioners,
after
acknowledging that the refinedtalk handlesnicelywhat the coarsetalk
did maybenot so nicely,complainthat the refinedtalk alsointroduces
nonsense,phraseslike3 belongsto 17, 3 equalsthe setwhoseelements
are the emptyset,the singleton of the emptyset,andthe setwhoseele-
ments are the latter two sets, and the like, which clutters up the lan-
guage.Thepractitioner
immediatelyrecognizes
whatis andwhatis not
meaningful
talk aboutnumbers;therecannotbe any doubtaboutthe
objectiverealityof the distinction.And the practitionerdemands,or at
leastdesires,that a foundationbe providedthat observesthis distinction.
Benaceraffdoes not say what I just said;maybehe findsit unnecessary
and/orunjustifiable
to sayit. ButI findplacesin hisnarrativewherehe
touchesthe groundof mathematicalpractice.In one of the quotations
above,the ordinaryfolkare saidto be dealingwithnumbersas numbers.
Thisplacein Benaceraffs
textindicates
theacceptance
of thepracticeof
mathematics,the locus of numbersas such,to be the final arbiter con-
cerningtheissuewhena proposed
language
ofmathematics
is right.The
little word as is the Archimedean fixed point on which the argument
pivots.Anotherof thequotationscontainsthereferenceto our usesof
the [number]words; theseareclearlytheestablished
usesin thepractice
that excludesthe nonsense I mentionedabove. I feel that these in-
 50 Makkai

stances of reliance on mathematical practice are the source of the strength

of Benaceraff ' s argument .

It should be clear by now that I want to turn the issue at hand into one

concerning language , from the ostensible one that is about ontology . True ;

I think the issue is not what numbers are , but what the nature of numbers

is , nature that can only be revealed in the language of the practice dealing

with numbers . The question " What are numbers ? " is inherently defective .

It accepts only a certain kind of answer , one that may be impossible to

provide ; it presumes that we can point out a domain of things that are

already familiar and can proceed to narrow down that domain to arrive at

the precise domain of the numbers . The question " What are numbers ? "

has a reductionist bias . We will not likely get to know the numbers by

coming to see that they are just like some other things that we already

know ; it is quite possible that they are unique in their essential nature .

What that nature is , is the good question . And , we then must realize , that

nature is to be found in the only tangible reality that contains the num -

bers , the language of the practice dealing with numbers .

There is natural language as it is found in practice ; and then there is

formal language that is the result of philosophical reflection on natural

language and of a subsequent deliberate act of articulation .

Our problem , then , is this : is it possible to articulate a language of

mathematics in which we talk about numbers as numbers ?

3 .2 Structures

Let us get closer to the way mathematicians talk about , and use , numbers ;

in this instance , by " number " I mean " natural number " ( any of 0 , 1 , 2 , 3 ,

etc .) . I will invoke the " Ideal Mathematician " ( I . M .) , a creation of Philip

J . Davis and Reuben Hersh ( 1980 ) ( a delightful concept indeed !) ; I will

construct the " ideal belief " ( not necessarily the best belief ; see Davis and

Hersh 1980 ) , I . Mis belief , concerning numbers . I . M . says :

1 M All I need is a system ( set ) N of entities called henceforth the

( natural ) numbers , with a distinguished element called zero ( 0 ) , a dis -

tinguished unary operation giving the successor S ( n ) of any number n ,

such that 0 is not the successor of anything , S is a one - to - one function ,

and the Principle of Mathematical Induction is valid : given any property

P ( ) of numbers ( equivalently , any subset P of N ) , if P ( O ) holds ( 0 E F ) ,

and \ in E N . (P (n ) - t P ( S ( n ) ) ) , then all numbers have the property P

( \ in E N . P ( n ) ) .
Structuralism in Mathematics 51

I . M . is familiar with logical notation . Our undergraduate students make

mistakeswith the quantifiers, but they know them! Frege's reform of the
language of mathematics has percolated thoroughly through everyday
usagein mathematics. To continue the imagined quotation from I . M .:
1 M. There is an axiom of set theory asserting that such a system exists.
Given such a system, I can do everything you have ever thought of or will
ever think of doing with numbers . One such thing is defining by recursion
a function one (or more ) of whose variables is ranging over numbers . I
mentally fix such a system, and will always refer to this one when I think
of numbers . That ' s all there is to natural numbers .

Above , I . M . described a more or less typical kind of structure ; in this

case, an entity of the kind , a structure of the given kind , is a set with a
distinguished element and a distinguished unary operation , satisfying
certain definite conditions . I . M . claims that all he needs is a Peano sys -
tem, as the structures of the described kind are called (although it was
Richard Dedekind who invented them ) .
Let me say something that may be partly new for I . M . The definition
of " Peano system" may be given equivalently in the following way .
(N , 0, S) , where N is a set, 0 EN , and S : N ~ N (I : A - + B means that
1 is a function with domain equal to the set A , and range contained in the
set B ; I . M . is familiar with this notation ) , is a Peano system if and only if
a certain condition holds . Before I give it , for the uniformity of notation ,
instead of 0 EN , I will write 0 : t ~ N . Here t is any fixed one - element

set, (say, t == { Q5} , but it does not have to be that); note that a mapping
from t to any set A is really the same as an element of that set, the element
which is the value at the unique element of t. The condition for (N , 0, S)
to be a Peano system is this : given any (A , a : t ~ A , 1 : A ~ A ) (thus , so
far , we have something that is " like " our system (N , 0, S)),
there is a unique function g : N - + A such that the diagram
0 s
t 1 N 1 N

i 0 190 19
t , A , A
a f

commutes .

For example , when we first apply S, and then g (on the right ), the effect is
the same as when we first apply g, and then f (this is the commutativity of
 52 Makkai

the right - hand square ) ; also , the left vertical i denotes the one possible
function from t to t , the identity function . When we rewrite this in " alge -
braic notation , " we get this :

g (O) == a (here , I revert to 0 EN , a E A from 0 : t ~ N , a : t ~ A)

g ( S (n ) ) == f (g (n ) ) ( n EN ) .

We realize that , in case (N , 0 , S ) is the Peano system that I . M . fixed in his

mind , g (n ) is equal to

f (f (f ( . . . f ( a ) . . .) ) == In ( a ) ,

i i
nil
n - l

the result of applying f to a n - times . Really , g is defined by a very special

recursion , iteration . The fact that we thus have an equivalent definition is
to say that the Principle of Iteration is equivalent to the Principle of
Induction when the latter is bolstered by the first two Peano axioms .
This alternative definition of " Peano system " is due to F . W . Lawvere .

I . M . should now sit down and prove to his satisfaction that indeed , we
have an equivalent definition . Of course , he is allowed to use set theory ; it
is unlikely that he is unfamiliar with the needed tools .

This definition is nice for instance because it is more compact than the
traditional one . There is no talk about S being one - to - one and missing 0
in its range ; also , it is unifoffi1 in the sense of being " diagrammatic ,"
using sets and functions , but not properties ( subsets ) . But really , the nicest
thing about it is that it has a pattern . It looks like we can talk about a

morphism between any two " such things ," (B , b : t - + B , h : B - + B ) and

(A , a : t - + A , f : A - + A ) ( let us call " such things " pre - Peano systems ) , as
being a map g : B - + A making the diagram
b h
t ~ B ~ B

0
19 19
A
0

~ A
a
f

commute; and thenwe can saythat a Peanosystemis distinguished

amongpre-Peanosystemsby thefactthatit hasexactlyonemorvhism to
vL
anypre-Peanosystem
. (An "impredicative . was
" definitionif thereever
one!)
Structuralism in Mathematics 53

Let us probe I . Mis convictions about the notion of number .

Q. You said that you fixed any Peano system for the purposes of
" number ." Does it not matter which one you have chosen ? To put it dif -
ferently, suppose you are probing some property P(n) of the natural
numbers , and you are not sure whether the property is universally true or
not of numbers . You discover that it is indeed true in your chosen system .
How do you know that it would have turned out to be true if you had
picked another Peano system as your numbers ?
L M . Aha ! Good question . Well , the answer is simple . If I have dis-
covered that the natural numbers have that property you have in mind ,
then I must have inferred this on the basis of the general principles of
mathematics (set theory if you wish ), and on the basis of the definition of
" Peano system" ; as I say, I make a point when I reason about numbers of
not using anything else than the defining properties of " Peano system."
Q. But wait ; does everybody make such a commitment ? Maybe I ex-
plicitly refuse to make such a commitment ; I simply take my system of
natural numbers , a Peano system , and by direct examination , mixed with
ingenuity , I see that in my system P (n) in fact holds for all n. I tell you
about this ; I describe to you how I directly examined my Peano system.
You said that you can do anything that I can do with numbers . Are you
sure that you can now also show that all n satisfy P (n) , given that in effect
you have committed yourself to using only methods of proof that apply to
all Peano systems ?
L M. How clever- that is a good question indeed . But there is an
important fact here: actually , any property that is verified in one Peano
system is going to be true in any other . More than that : any property
of any Peano system as a whole will also be' shared by any other Peano
system. Take the example that the underlying set of a Peano system
(the set of natural numbers , in anyone interpretation ) can be mapped in
one - to - one fashion into any infinite set . This is a fact ; but even if I did not
know that , but knew that it was true for one Peano system , I would know
that it was true for any other . The reason is that any two Peano systems
are isomorphic ; there is a one-to -one and onto mapping from one to the
other taking the zero of one to the zero of the other , and taking a pair that
is in the relation of one term 's being a successor of the other term into a
pair in the same relation in the other . Isomorphic structures share all
conceivable properties .
Q. That is indeed interesting . I can actually see why any two Peano
systems are isomorphic ; this is quite easy using Lawvere 's definition . We
 54 Makkai

have a morphism from one to the other , and from the other to the one;
moreover , their composites are self-morphisms of the systems, and thus
they must be identities ; I really have an isomorphism , because I have an
invertible morphism ; great ! But how do you know that isomorphic struc-
tures share all properties ? I can see this holds for the one you just quoted ;
this is quite obvious . But why so in case of any conceivable property ?
Anyway , you cannot be quite right . Look : suppose you take the property
that the empty set Q5is an element of the Peano system. This is clearly not
going to be shared by all Peano systems; I can willfully make this both
true and false by simply exchanging elements, clearly not disturbing the
fact that I have a Peano system! Something is wrong .
1 M . No , no ! Of course, when I said " all conceivable properties ," I
meant " all meaningful properties ," " all properties of Peano systems as
Peano systems." Q5, being an element of the Peano system, is not a
meaningful property of Peano systems at all ; no sane person would con-
template such a property .

It is clear that communication is breaking down at this stage. I . M .

cannot really say what he means by a meaningful property of Peano
systems, although he can unerringly say of any particular proposed
property whether it is meaningful or not . Of course, he could say that
a meaningful property is, by definition , one that is invariant under iso-
morphism : if it holds of one structure , it holds of any other that is iso-
morphic to the first . But that begs the question ; how is he so sure, and so
quickly , when he is presented with a particular property ? It is important
to see that he can make the judgment in the cases when he has no idea
whether the property in question does in fact hold in Peano systems.
There is, in fact , a more basic problem . Defining " meaningful " by refer -
ring to the condition of invariance under isomorphism , one would be
making an inadmissible move . The meaningfulness of a phrase should be
a matter of grammar , not of a question of a possibly difficult -to -verify
mathematical fact. In fact, it is demonstrably undecidable whether a
property formulated in set theory of Peano systems is invariant under
isomorphism.
I . Mis colleague , I . L ., a student of Alfred Tarski , overhears the dis-
cussion and joins in .

I L. This is a question of logic . The meaningful properties of structures

are the logical properties . Yau have in fact several precisely and explicitly
defined logical languages in which you can formulate properties of struc-
 55
Structuralism in Mathematics

tures, and you can rigorously show that these properties are invariant
under isomorphism. The most basic of theselanguagesis first-order logic.
As Tarski has shown, we have a precisenotion of truth for sentencesof
first-order logic in structuresof the right kind (interpreting all the symbols
in the sentencein question), defined globally for the whole language of
first-order logic at once. When for instancewe were asking whether P(n)
held for all n, we A were askingAwhether the Peanosystem(N , 0, S) satisfied
.
a sentenceVnP(n), where P(n) is a first-order formula expressmgthe
property P(- ), of courseprovided that such a formula is available. As a
matter of fact, in most casesof interesting Ps it is not available; in most
cases, it would become available if we added new operations such as
addition and multiplication to the primitives of the structure (N , 0, S).
Now , the important point is that, with the help of the Tarski truth-
definition, there is a rigorous proof of the fact that properties given by
first-order sentencesare invariant under isomorphism; it is a proof by
induction on the complexity of the sentencein question. Thus, there is a
partial answerto the question " What are I . Mis meaningful properties?" ;
they include the first-order properties. It is true, however, that most
mathematically interesting properties of structures- and now I am talk-
ing about not just Peano systemsbut all the various structuresmathema-
ticians use: groups, rings, topological spaces, and so on- are in fact not
expressiblein first-order logic. The way out is that there are other, more
expressivelanguagesfor which the Tarskian way of defining truth is also
available; now the truth-definition may be more conspicuouslydependent
on set theory, but in fact, it was so dependentalready in the caseof first-
order logic. These languages include second-order logic, higher-order
logic, infinitary logics, logics with generalizedquantifiers. It turns out that
the proof that a property expressiblein any of these languagesis iso-
morphism invariant is a very straightforward structural induction on the
complexity of the expressioninvolved.
Q. May we then say that the idea of a " meaningful property of a struc-
ture" is ultimately an open-ended one; there is no uniform syntactical
criterion that describesall the meaningful properties that mathematicians
contemplate?
1 L . Yes, this is correct; there are large classesof explicitly described
properties that are meaningful, for which we indeed have a proof in
advancethat they are isomorphism invariant, but it seemsrather impos-
sible to give such a classthat would be all-encompassing .
 56 Makkai

1 M . Mind you, this is not a problem in practice. I am working'in group

theory, and I have never yet asked myself whether something somebody
proposed or asked about groups made sense; it simply always did when
the person was mathematically competent, even on a minimal level. And
when somebodyproposessomething like " Is the monster-group equal to
the first prime number greater than 10100 ?" , then I know that this is
meaningless , you are comparing applesand oranges. Mind you, when you
ask " Is the monster-group the sameas the largestsporadic simple group?" ,
you are not asking about a literal equality of two objects; you are asking
whether those two things are isomorphic. You see, provided you have
identified the monster-group as a specific set, the set-theoretic statement
" The monster-group is equal to the first prime number greater than
10100 " is meaningful as a set-theoretic statement, but not as a mathemat-
ical statement.

I will summarizeby saying that we have a tantalizing two-faced situa-

tion within the set-theoretical foundation. On the one hand, through the
processof self-reflection it provides glimpses, collectively called " logic,"
of large connectedparts of a languageof mathematicsthat treats numbers
as numbers, and in general, structuresas structuresinasmuch as it allows
only meaningful propositions about those things. On the other hand, it
does not give an articulated and complete statementof what such a lan-
guage, Logic, is; rather, it makes it look likely that such Logic does not
exist. I would like to obtain a foundational language in which all well-
formed propositions are meaningful.

3.3 TheNatureof Things

I . M . was urging the view that one does not have to assumethat terms
like 3, 17, 1010have absolutely fixed denotations; the latter can be made
dependent on a free choice of a structure of natural numbers, a Peano
system, in which the denotations then becomedetermined. It also seems
that this attitude toward denotation is the prevailing one in mathematics,
dealing as it does with definite descriptions of various entities- " the
monster-group," " the field of the real numbers," and so on- in a sys-
tematically ambiguousmanner. This may be so in mathematical practice,
but it is entirely possiblethat this way of behaving is necessarilya meta-
phorical one, and when it comesto the crunch of articulating a founda-
tion for mathematics, this ambiguity has to be given up. This would not
Structuralism in Mathematics 57

necessarily mean that one would have to adopt the precise foundations of
Zermelo -Fraenkel set theory , but it would show that something like that
system, with a single domain of well -individuated objects at its base, is
necessary, after all ; and in particular , it seems to me, it would answer our
question " Is it possible to articulate a language of mathematics in which
we talk about numbers as numbers ?" in the negative .
And indeed, there are many signs in mathematical practice suggesting
that ultimately it may be necessary to fix reference of terms unambiguously .
In " Complete Functors in Homology I " Max Kelly writes :
(The objects of a category have but a tenuous individuality ; 'the group 7L of
integers ' is not the same set of elements to different writers , or even to the same
writer in different contexts . In a sense it is only the isomorphism class that counts ;
and yet we must at any instant be considering a definite set of elements, in order
that we may sensibly talk of homomorphisms : a group is not the same thing as a
group type , because there are non -identical automorphisms . . . .) ( 1964, 722)

This partially quoted parenthetical remark is in a section called " Gen -

eralities on Functors ." It confirms the view that the systematic ambiguity
of mathematical objects discussed above reigns within a category'. How -
ever, when it comes to a functor F : X - + A , that is, a mapping con-
necting two categories, it seems that the very notion denies systematic
ambiguity ; it assigns a definite object F (X ) as value to any argument
object X in the domain category X . In other words , when we step out of
the context of a single category , where the precise identity of objects
did not matter (see Kelly 's " tenuous individuality " ), and want to bridge
the worlds of two categories in a global context , we seem to be forced to
use tools (the functor , for instance) that do refer definitely (to the object
F (X ) in this case), and not systematicallyambiguously.
And indeed, there is an entrenched view in philosophy that truth in
mathematics must be explained by a standardly referential semantical
theory . Benaceraff ( 1973) argues to this effect, and further that this cir -
cum stance leads to an inevitable conflict between the demands of ontol -

ogy and epistemology of mathematics . Without committing myself on

Benaceraff 's conclusions , I want to call into question the part of his basic
position that concerns the necessity of a standardly referential semantical
theory of mathematical truth . Along the way , I want to open up the pos-
sibility of a nonstandardly referential theory of truth , in which a system-
atic ambiguity of terms reigns, and thus to make it plausible that it is
possible to deal with numbers as numbers in an articulate manner . Inci -
58 Makkai

dentally, there is a specificparallel technical issue, in the context of func-

tors; one needs to show that " functors are not necessary ," and I have
made an effort to this effect in Makkai 1996. It is interesting to point out
that two parenthetical paragraphsin Kelly 1964contain a partial formu-
lation of the notion that should replace that of functor (the concept of
" anafunctor," as I call it ; Kelly 's description is partial sincethe necessary
condition of " saturation" is not mentioned), though without a firm sug-
gestion that this notion could be used to systematically reintroduce
ambiguity of objects in categories.
I want to challengeBenaceraff' s (1973) basic commitment to
(1) the concernfor havinga homogeneoussemanticaltheoryin which semantics
for the propositionsof mathematicsparallel the semanticsfor the rest of the
language.. . (p. 661)
The quotation is footnoted as follows:
I am indulginghere in the fiction that we havea semanticsfor "the rest of
the language " , or moreprecisely
, that the proponentsof the viewsthat taketheir
impetusfrom this concernoftenthink of themselves ashavingsuchsemantics, at
leastfor philosophicallyimportantsegments of the language. (p. 661)
The footnote shows that Benaceraffis trying to leave the door open for
himself to abandon the commitment to the above concern. Even if he did
not himself have this concern, his overall argument to the effect that
almostall accountsof the conceptof mathematicaltruth can be identifiedwith
servingoneor anotherof thesemasters[the othermasteris " (2) the concernthat
the conceptof mathematicaltruth meshwith a reasonableepistemology "] at the
expenseof theother(p. 661)
would stand up. On the other hand, the rest of the paper showsthat he is,
after all, committed to the stated concern. For instance, he writes:
Some(includingone of my past and presentselves[and herehe refersto Bena-
ceraff 1965]), reluctantto facethe consequencesof combiningwhat I shalldub
sucha " standard " semanticalaccountwith a platonisticview of the natureof
numbers , haveshiedawayfrom supposing that numeralsarenames . . . (p. 664)
Thus, to repeat, I am not really arguing with Benaceraff' s main thesis.
What I am arguing is on a more basic level. I am going to deny that
one can maintain (1); I will assert that one cannot have a homogeneous
semanticaltheory of the world as we know it, evenif we disregardmathe-
matics proper; I will assert that the " fiction" Benaceraff refers to is a
fiction properly, and it is to be discarded.
Structuralismin Mathematics 59

To fix ideas, let us look further at Benaceraff1973:

Considerthe followingtwo sentences
:
(1) Thereareat leastthreelargecitiesolderthanNew York.
(2) Thereareat leastthreeperfectnumbersgreaterthan 17.
Do theyhavethe samelogicogrammatical
fonn? More specifically
, aretheyboth
of the fonn
(3) Thereareat leastthreeFG's that bearR to a.
. . . ?What arethetruth-conditionsfor (1) and(2)?Are theyrelevantlyparallel?. . .
[I]t seemsclearthat (3) accuratelyreflectstheform of (1) andthusthat (1) will be
true if and only if the thing namedby the expression replacing'a' ('New York')
bearsthe relationdesignated by the expression replacing'R' (CDis olderthan (2))
to at leastthreeelements(of the domainof discourseof the quantifiers ), which
satisfythepredicates replacing'F' and'G' ('large' and'city', respectively
). . . . But
what of (2)? May we use(3) in the sameway asa matrix in spellingout the con-
ditionsof its truth? That soundslike a silly questionto whichthe obviousanswer
is "Of course " . (p. 663)
Thus, for Benaceraff the homogeneity of our semantical theory means
that it is like the account of (1) given abovevia (3). 1 claim that we do not
have this kind of account of our reference, and truth-determination, even
in discourseabout certain entities that we habitually encounterin real life.
To clarify the respectivepositions further, before I get to the claim
itself, let me point out that we see here the causal theory of reference
at work . " I believe in addition in a causal theory of reference.. . ," writes
Benaceraff (1973, 671). The causal theory of referenceis described, for
example, by Maddy (1992, 38- 41), who attributes it mainly to Saul
Kripke in Naming and Necessity(Kripke 1972). According to this, the
naming that Benaceraffrefersto when he talks about the " thing named by
the expression'New York ' " involves a causal chain of eventsthat starts
with an act of " initial baptism," the dubbing of the thing New York with
the name 'New York ,' an event in which New York and the dubber had
to be both physically present.
I believe in a fundamental inhomogeneityin the kinds of things that
make up the world . On the one hand, we have the " medium-sizedphysical
objects" to which the causaltheory of referenceand of knowledgeapplies;
but we also have, in a way parasitically living on thoseobjects, things that
1 call representationalobjects: things that exist only in ambiguous repre-
sentation by physical objects, but neverthelesshave an irreducible pres-
encein our world . We do not refer to representationalobjectsdirectly; we
 60 Makkai

refer to them through representationsof them. Before I say what these

representationalobjects are, I want to emphasizethat I do not claim to
have classifiedall things into the two groups I mentioned; in fact, I think ,
there are further kinds of things that are neither " medium-sizedphysical
objects" nor representationalobjects. However, I also believethat mathe-
matical objectshave much to do with representationalobjects, evenif I am
not preparedto fully identify the former with someof the latter. Thus, my
casefor representationalobjects is but a preparation of a casefor mathe-
matical objects as such.
Representationalobjectsare the onesthat are symbolically represented ;
they are invariably artificial , human artifacts. Take William Blake's poem
" The Chimney Sweeper." You probably know this poem; if you don't, I
will tell you how you may find it , although this will not be the only pos-
sible way. The referenceis Blake's Poemsand Prophecies, edited by Max
Plowman, in Everyman's Library (London: Dent, New York : Dutton ),
1972 reprint. The poem is found on page 11. Actually , I did not in this
way give you a unique way of finding the poem. In referring to Blake's
Poemsand Prophecies, I gave you only a type, not a token; here, in my
hand, I have a particular copy of the book, and I can show you the poem
on the 11th page. Thus, the book I describedis another example of a
representationalobject! This immediately tells you something about the
hierarchical interrelatednessof representationalobjects.
Can I doubt the existenceof Blake's poem " The Chimney Sweeper" ?
Does Blake's poem " The Chimney Sweeper" exist? Is this question a
meaningful one? It seemsthat the question may have two different mean-
ings. The " nonphilosophical" meaning is the question asked by someone
having a possibly superficialbut referentially still firm idea about Blake to
the effect of whether Blake has indeed written such a poem (we hope he
did not write two with this sametitle. . .). This is an unproblematic ques-
tion from the point of view of the theory of truth and reference; the ways
of answering it are, in principle, satisfactorily circumscribed by the
" causal theory of referenceand knowledge." But there is also the " philo-
sophical" meaning of the samequestion; it is in force when we ask, " Does
Blake's 'The Chimney Sweeper' exist as a poem?"
You may maintain, " No , there are no poems as such; what there are,
are the ordered setsof inscriptions that run as follows:
" The Chimney Sweeper"
When my mother died I was very young ,
(etc.)
Structuralism
in Mathematics 61

Well, what about an electronically stored copy of Blake's poem? You say,
" Of course, I meant that also in an extended senseof 'ordered set of
inscriptions.' " Can you be more explicit? You may resort to the formula
on the inside coversof books (" No part of this book shall be reproduced,
stored in a retrieval system, or transmitted by any means, electronic,
mechanical, photocopying, recording, or otherwise. . ." ), in order to exhaust
all the possiblerepresentationsof the poem. But note the indefinitenessin
the italicized phrase " or otherwise" ; the practical people know that it is
impossible to foreseeall possibleways one may representthe poem, and
they want to forestall all (lucrative) representations.Trying to say that the
poem's essenceis somehow synonymous with the totality of its repre-
sentationsis wrong-headed, for two reasons, at least. One is that we know
perfectly now what the poem is, but we cannot know now all its repre-
sentations, not even the possibletypes of its representations. The other is
that one representationis perfectly enough for knowing what the poem is!
Then again, you will now say that referring to Blake's poem is just a
fat;on de parler; you can eliminate it from any context, in exchangefor
referencesto first-class objects. I doubt that; how do you do that when
I say, " I like Blake's poem 'The Chimney Sweeper'" ? Here we enter an
infinite, and familiar , controversy, involving behaviorism and the like.
Sufficeit to say that the correspondingquestion that we really care about,
the one concerning objects in mathematics, is answereddefinitively: we
know that, for instance, the presenceor absenceof the notion of natural
number doesmake a differencein our ability to prove theorems(a form of
G6del's incompleteness ). In other words, we cannot eliminate the notion
of natural number from contexts.
In the last two paragraphsI tried to show a bit what it would be like to
deny the existenceof " The Chimney Sweeper" as a poem. I think , how-
ever, the best reaction to the " philosophical question" of the existenceof
" The Chimney Sweeper" as a poem is that it is meaningless . There is no
mystery, we have everything in front of us; the question cannot seriously
imply a searchfor something temporarily unknown answeringa descrip-
tion, and it seemsthat how we answerthe question is largely arbitrary .
Before we probe matters along theselines a bit more, I want to suggest
that the question of the existenceof the system of the natural numbers
may be to a large extent similar and, as a consequence , meaningless .
Here you will quickly object, " No ; whereasin the caseof 'The Chimney
Sweeper' the question of the existenceof a representationis not in ques-
tion, now, with the system of the natural numbers, this is precisely the
 62 Makkai

important question." To which I will reply, " Certainly, if by a represen-

tation of the systemof the natural numberswe mean somethinginvolving
an actual aggregate(set) of elements; but I do not think we have to, or
evenshould, do so." Note also that the existenceof a Peanosystemwithin
a formal systemof set theory, say, is of course meaningful; in fact, it is
expressedby a proposition of the theory, which mayor may not be
provable in the theory.
The being of the poem " The Chimney Sweeper" as a poem is exhausted
by our ability to talk about it as a poem. Talking about it as a poem is to
be referring to its representations, possibly simultaneouslymore than one
(for instance, the ones that you and I have in our minds), and to make
sensewhile doing so. For instance, " in the second line of the poem"
makes senseas talk about the poem, even though, strictly speaking, it is
eachof the representationsthat has a secondline. On the other hand, " the
typeface of the poem" does not make senseas talk about the poem as
such; rather, with that talk we immediately know that we are talking, not
about the poem as such, but about one of its representations. In natural
language, we do not want to rigidly separatethe higher-level talk about
the poem as such and the lower-level talk about its representations; it is
the very spirit of natural languagethat we want to be able to jump from
one context to the other instantaneously. On the other hand, the legend-
ary purity of mathematicsis rooted in an opposite kind of requirement; in
mathematics, we want to be able to talk about structures as structures,
and more generally about mathematical objects as such, in a permanent
manner. It is this requirement that compels us to articulate formal lan-
guagesfor mathematics.
We should add that talk about the poem as such grows out of talk
about its representations. Continuing the work of the poet, we bring the
poem into existenceas a poem by articulating our talk about it as such.
This is another reason why the two types of talk are not, and cannot be,
rigidly separated. In fact, the processof creating the poem is an ongoing
one; it has not beenquite completed. In mathematics, we have completed
our talk about our abstract objects to a greater extent, although the dif-
ferenceis only a matter of degree.
One might think that talking about poems is inappropriate in connec-
tion with mathematics. Of course, there are many other types of repre-
sentational objects, and none is more important than the onesthat arisein
the world of computers: computer programs, piecesof software. It is in
the domain of computer software that we seethe characteristicsof repre-
Structuralism in Mathematics 63

sentational objects most clearly. The talk about the software becomes
sharply separatedfrom the representation, which is a particular imple-
mentation of the software. The existenceof the software as such becomes
more pronounced; we are now more ready to accept the existenceof
the software as something separate and independent from any imple-
mentation, more ready than in the caseof the poem (perhaps). Another
element in favor of the existenceof the software as such is its potency.
One may doubt whether poems do anything; but we do not doubt that
pieces of software do things. They have very specific and occasionally
large effects- of course, never in themselves, always in an implementa-
tion . Discussing, describing, planning around the effect of the software
takes place mostly in talk that is independent of any particular imple-
mentation. I am inclined to the view that mathematical objects are like
software, more specifically, like the datatypesof software.
But let us return to poetry once again. Maybe our brave talk about
poems as genuine representational objects can be shown up as hot air,
after all. It is very well to talk about the many representationsof " The
Chimney Sweeper" in books, in people's minds, and so on; but there is a
distinguished, an authentic, an original representation: the one that Blake
himself put to paper. When we talk about the poem " The Chimney
Sweeper," we talk about this original copy; we talk about a first-class
object pure and simple. When we use ostensionand point to the copy on
page 11 of my book, what we do is deferred ostension, to use Quine's
expression. We mean referring to the Original; we refer to the Original.
This is what Benaceraffmeant by the " homogeneoussemanticaltheory of
the rest of the language," that is, the language that refers to reality as
opposedto mathematics.
Before we try to answer this devastating argument, we have to ac-
knowledge its force. In this we can see why we want the Platonistic,
uniquely referring, standard semantical theory: we want authenticity, an
objective point of reference, to which we can return when in doubt. There
may always be a question whether a particular printing of the poem is
correct; ultimately, the only way to answer this question is to go to the
original and verify whether the copy in question and the original can be
mapped to each other perfectly.
The attempt to answerthe argumentis this. Assumeyou discoverthat in
fact, there is no original " The Chimney Sweeper." In fact, worsethan that,
you discover that there is somethingentitled " The Chimney Sweeper" in
a manuscript of Blake's, but it differs substantially from the " standard
 64 Makkai

version," the version that has been most frequently reprinted; moreover,
no trace of a version of the poem as we know it can be found in Blake's
manuscripts. You counter: This does not matter; the poem as we know it
was composedby someoneat some definite time and place, and when we
talk about the " The Chimney Sweeper" as we know it, we talk about that
original, even though it may differ from anything attributable to Blake
himself. To which I say that now the original of " The Chimney Sweeper"
as we know it is a purely fictive entity; it has lost its role as the grounding
of authenticity. However, in fact still there is no problem about what
" The Chimney Sweeper" as we know it is; it simply is what it is, and no
one will try to authenticate it , especially now (we are still within the
assumption that we have lost the good connection to Blake) that we
know that it cannot be authenticated. To which you say: There is no way
of stopping people from fixing the first, or most visible, or whatever, but
in some definite way unique source where the ' 'The Chimney Sweeper"
can be found, a source which becomesthe standard referenceto " The
Chimney Sweeper" as we know it. And then I reply: Oh, but now you
cannot say that " The Chimney Sweeper" as we know it is, by definition,
that given by the standard reference, since in fact there was something
entirely clearly defined before we had located that standard reference;
the location of the standard referencewas an a posteriori act of mock
authentication, an act entirely dependenton the clear and prior idea of
" The Chimney Sweeper" as we know it .!
Maintaining " the fiction of the homogeneoussemanticaltheory for the
language of the nonmathernatical world " (Benaceraff' s words, essen-
tially ) is hard work. I am suggestingthat it is too hard, what with all those
gadgetsthat comeflooding into our lives, eachhaving little or large pieces
of structure, defined in terms of someabstract functionality or what have
you. To make senseof all this, we resort to representational objects,
rather than trying to find unique authentic originals to which we would
have to run in caseof doubt. Certainly, theserepresentationalobjects are
abstractions (I do not mean to imply that the teffi1 I just used explains
them; rather, its use puts them in their place as inferior in some sense
to concrete things) and thus decidedly second-class objects that cannot
live without being rooted in medium-sized physical objects. But still, we
cannot do without them.
Is this a pragmatic argument? Yes, it is; but remember, the argument is
not about the existenceor nonexistenceof representational objects, but
about the fact or nonfact of the homogeneity of the semanticsfor the
Structuralism in Mathematics 65

language of the nonmathematical world . The conclusion may be that

the idea of the homogeneous , uniquely referential semantics vis-a.-vis
the world is in fact incoherent; if so, it is even worse for the fiction that
Benaceraff admits to be indulging in. It now seemsto me that we do
mathematics precisely to get away from all this compulsion of authenti-
cation, and deal with what we call mathematical objectsaswe know them.
At this point, I feel optimistic that a foundational stance in mathe-
matics that lacks the standard feature of unique referenceto objectsmay
be viable. However, the real test of it still remains a Fregeanfull articula-
tion . Before we have built the whole language, we cannot be surewhether
we do not, ultimately, smuggle in the " originally baptized," authentic
entities that the Platonists want.

Note
1. Ray Jackendoffhas suggestedto me a real exampleparalleling the hypothetical
case describedabove of being unable to find an authentic original. I describeit
essentiallyin his words.
There is a piece of music that has been in the repertoire at least since the early
nineteenthcentury, which is called " Sinfonia Concertantefor oboe, clarinet, horn,
bassoon, and orchestra by Mozart ." Only there is no known manuscript for this
piece. In a letter, Mozart himself referred to having written a sinfonia concertante
for flute, oboe, horn, and bassoon- which is totally unknown- and people gen-
erally assumethat the piecewe know is a later reorchestrationof that. The trouble
is that all attempts to " reconstruct" the original have been awkward at best;
basically, the clarinet part in the piece we know cannot easily be transferred to
idiomatic writing for any of the instruments mentioned in Mozart 's letter. More-
over, the last movement of the piece we know has someparts that Mozart never
would have written . So no one knows where this piececame from, prior to its first
publication someyears after Mozart 's death.
Jackendoff also pointed out to me that he has had to deal with representational
objects in his work on music, which involves an additional layer of performance,
making the semanticsof, say, " Beethoven's Fifth Symphony" even stranger.

References

Benaceraff, P. (1965). What numbers could not be. Philosophical Review, 74,
47- 73.
Benaceraff, P. (1973). Mathematical truth . The Journal of Philosophy, 70,
661- 675.
Davis, P. J., and Hersh, R . (1980). The mathematical experience. Birkhauser
Verlag.
Eilenberg, S., and Mac Lane, S. (1945). General theory of natural equivalences
.
Transactionsof the American Mathematical Society, 58, 231- 294.
66 Makkai

Kelly , G . M . ( 1964 ) . Complete functors in homology I . Proceedings of the Cam -

bridge Phzwsophical Society , 60 , 721 - 735 .

Kripke , S . ( 1972 ) . Naming and necessity . Oxford : Blackwell .

Lawvere , F . W . ( 1969 ) . Adjoints in foundations . Dialectica , 23 , 281 - 296 .

Lawvere , F . W . ( 1970 ) . Equality in hyperdoctrines and the comprehension schema

as an adjoint functor . In A . Heller ( Ed .) , Proceedings of the New York Symposium
on Applications of Categorial Logic ( pp . 1- 14 ) . Providence , RI : American Math -
ematical Society .

Lawvere , F . W . ( 1971 ) . Quantifiers and sheaves . In Actes du Congres International

des Mathematiciens Nice 1970 ( Vol . 1, pp . 329 - 334 ) . Paris : Gauthier - Villars .

Maddy , P . ( 1992 ) . Realism in mathematics . New York : Oxford University Press .

Makkai , M . ( 1996 ) . Avoiding the axiom of choice in general category theory .

Journal of Pure and Applied Algebra , 108 , 109 - 173 .

Makkai , M . ( 1997 ) . Generalized sketches as a framework for completeness theo -

rems . Journal of Pure and Applied Algebra , ll5 , 49 - 79 ( part 1) , 179 - 212 ( part II ) ,
and 241 - 274 ( part III ) .

Makkai , M . ( 1998 ) . Towards a categorical foundation of mathematics . In J. A .

Makowski and E . V . Ravve ( Eds .) , Logic colloquium '95 ( Lecture Notes in Logic
11 , pp . 153 - 190 ) . New York : Springer -Verlag .

Makkai , M . (to appear ) . First - order logic with dependent sorts ( Lecture Notes in
Logic ) . New York : Springer - Verlag .

Parsons
303,.C.(1990
-346 ).The
structuralist
view
ofmathematical
objects
.SyntheseJ
84
,
Chapter~ 4

The Natural Logic of Rights Ray Jackendoff

and Obligations

4 .1 Introduction

The focus of John Macnamara ' s work over three decades was an in - depth

inquiry into the fundamental structure of human knowledge . Though less

publicly spectacular than Chomsky ' s results on the nature of syntactic

knowledge , John ' s research went beyond linguistic expression to delve

into the character of meaning , thought , and reason themselves . Like

Chomsky , John was asking what it is possible for a child to learn on the

basis of the input in the environment - and what parts of the child 's

knowledge cannot be learned , but must serve as the basis for learning .
Some of his results were absolutely startling , at least in the context of the

sort of empiricist philosophy of meaning that prevailed at the beginning

of his career and that has enjoyed a strong resurgence in these con -

nectionist times . For those of us of a more rationalist cast , his way of

chewing over issues of epistemology has been a continual inspiration .

Anyone acquainted with John sensed that he was a man of deep moral

convictions . Occasionally the nature of these convictions became for him

a topic of study as well , for instance in his paper " The Development of

Moral Reasoning and the Foundations of Geometry ." Here he argued

that a sense of morality has to be as much a part of our inborn character

I am grateful to my late father Nathaniel J ackendoff , my brother Sparky

J ackendoff , Larry Solan , Carlos Otero , Storrs McCall , Al Bregman , Paul Bloom ,
Jay Conison , Gordon Bower , and members of the audience at the 1996 Piaget
Centennial in Geneva for discussion and comments on this material . I \\,Tas given

the honor of presenting this work as the first John Macnamara Memorial Lecture
at McGill in April 1997 , and I benefited a great deal from the discussion there as
well . This research was supported in part by National Science Foundation grant
92 - 13849 to Brandeis University .
 68 Jackendoff

as our understandingof physical space- that moral ideals come to us as

naturally as geometric ideals. The present chapter is an exercise in the
same spirit, if not so ambitious, investigating a pair of notions closely
related to moral reasoning : rights and obligations . I will be concerned
especially with what might be called " social/ legal/ contractual " rights and
obligations . Parts of the analysis will apply as well to " moral obliga -
tions ," " moral justifications ," and " human rights " ; other parts will not .
Some of the differences will be mentioned in section 4 .6 .
Rights and obligations are fundamental to the fabric of human social
organization . For example , ownership of an object confers on the owner
(or consists of ) rights to use the object and rights to prevent others '
using it (Miller and Johnson-Laird 1976, following Snare 1972). Giving
someone a promise places one under obligation to fulfill the promise .
Conferring on someone a social status (e.g., an official title , a professional
degree, or membership in an organization ) grants this person certain rights
and places him or her under certain obligations . Any sort of contract -
including not only financial / legal contracts but also marriage in many
societies- places the participants under obligation to perform certain
acts. Inasmuch as the main issues addressed by a society's legal system
(written or unwritten ) include the privileges of ownership , the making and
enforcing of contracts , and the rights and duties of officials and of citi -
zens, it is clear that rights and obligations playa central role in under -
standing concepts of law .
The notions of rights and obligations appear to be universal in human
societies. A great deal of anthropological description is devoted to how
societies differ in what rights and obligations pertain to their members ,
how such rights and obligations are obtained and lost , and how they are
taken to be grounded in religion or government . But such descriptions
invariably take the notions of right and obligation themselves for granted ,
not subject to discussion . 1 Yet , as I will show , these notions are remark -
ably complex and subtle. Thus , these concepts raise interesting questions
about learning and the evolution of cognition - just the sort of questions
with which John Macnamara was so deeply concerned . I will turn to these
questions briefly at the end.
I will be investigating rights and obligations in the context of a theory
of linguistic meaning - of the contextually integrated interpretations of
utterances- that in my view forms the bridge between linguistic semantics
and a larger psychological theory of how humans understand the world .
This theory investigates conceptual structure , a form of mental represen-
Logicof RightsandObligations 69

tation over which principles of inference (including invited inferenceand

heuristics) can be defined and in terms of which planning of actions takes
place. Conceptual structure is mapped into the strictly linguistic levels
of syntactic and phonological structure by sets of correspondencerules
(Jackendoff 1990).
This approach to meaning contrasts with many standard philosophical
approachesin the Frege-Tarski tradition , in that I do not take conceptual
structure to map directly into the real world; rather, it maps into the
world as human beingsunderstandit, quite a different notion (Jackendoff
1983). I am trying to study human concepts, not " ultimate reality."
However, conceptualstructure is still connectedto and constrainedby the
external world , but indirectly, via the complex mappings betweensensa-
tion and cognition that are establishedby the perceptual systemsof the
brain. Here I think I am basically in agreementwith John Macnamara's
approach, though we did disagreedeeply, in some way I still fail to un-
derstand, about how a theory of conceptsshould relate to the " real real
world ."
Like any theory of meaning, the theory of conceptual structure should
be supported by linguistic (including crosslinguistic) evidenceand by its
ability to formally support reasoning. However, becauseit is supposedto
be embeddedin a larger psychologicaltheory, it should also interact with
evidence from perception, child development, and potentially neuro-
science. And since conceptual structure is meant to be to a degreeinde-
pendent of the languagecapacity per se, we should in principle be able to
test the theory against evidencefrom the cognition of animals, especially
primates, both in the laboratory and in natural settings.
The domain of concepts investigated most intensively by myself and
many others (see, e.g., Talmy 1983; Herskovits 1986; Vandeloise 1986;
Bloom et at. 1996) is spatial cognition, the position and movement of
physical objects and substancesin space, the forces they exert on each
other, and the temporal structure of the statesand eventsthat result. This
domain is especiallyfruitful becausethere is a vast range of lexical items
expressingspatial concepts and becausethese correspond to a rich and
precise set of perceptually based intuitions. This line of researchis now
well establishedwithin linguistic semantics. In addition, it has long been
recognized that language expressing spatial concepts is mirrored to a
considerable extent by language expressingconcepts in other domains
(Gruber 1965; Jackendoff 1976; Lakoff and Johnson 1980; among many
70 Jackendoff

others). So understandingthe organization of spatial conceptshelps set a

foundation for investigating other domains.
The domain in which the study of rights and obligations is situated is
that of social cognition. Evidence from animal societies, especially pri -
mates (e.g., Cheney and Seyfarth 1990), suggeststhat a substantial com-
ponent of our conception of the world is devoted to the understandingof
the social context in which individuals find themselves(Jackendoff 1992,
chap. 4; Searle 1995). The fundamental unit of this domain is not the
physical object but the person, a seat of intention and volition . People
(and to some degreeanimals) have representationsin both this domain
and the physical; trees, rocks, streets, and buses are representedonly
physically.2 Within this framework, I will be asking how people concep-
tualize situations in which someonecan be said to have a right or an ob-
ligation. It makeslittle senseto ask what rights and obligations really are,
outside of people's understandingof their social context. In other words, I
am interestedin a theory of the " folk theory" of social relations. Like the
theory of the conceptualizationof objects, space, and force in the physical
domain, this forms part of the theory of human conceptualization- but
one far lessdirectly tied to perception.
In the course of my exposition, I will make use of the formal frame-
work of conceptual semantics developed in my previous work (e.g.,
Jackendoff 1983, 1990). Any alternative notation that makes similar dis-
tinctions is adequatefor my purposes. However, it is important that the
notation be construed as encoding regularities in the mind, not some
generalset-theoretic construct (involving, say, possibleworlds) as in most
varieties of formal semantics.

4.2 The Argument Structure of Rights and Obligations

An important aspectof my approach to conceptual structure is that the

grammatical structure of the languageused to expressconceptscan help
reveal the organization of the concepts themselves. So let us start by
looking at someof the ways that rights and obligations can be expressed
.
We seeimmediately that they form a closely related pair.
About the simplest way to expressa right in English is with the modal
verb may; an obligation can be expressedwith the modal verb must.
(1) a. One may use one's possessionsas one likes. (Right)
b. One must pay salestax in Pennsylvania. (Obligation)
Logicof RightsandObligations 71

One immedia te im pulse for formalizing these meanings might be to

take the modals to express operators (notated as RT and DB ) over a
proposition , as in (2). This is essentially the formalization found in von
Wright 's ( 1963) foundational work on deontic logic .
(2) a. Sue may (i .e., has a right to) leave when she wants to .
=: R T (Sue leaves when she wants to)
b . Sue must (i .e., has an obligation to ) leave before noon .
=: OB (Sue leaves before noon )
Such a treatment , however , misses the basic point that a right or an obli -
gation is a relation between a person and his or her action . The readings
of may and must that do express propositional operators for possibility
and necessity lend themselves to paraphrases like (3a), whose syntactic
structure reflects the semantic structure rather well . Such paraphrases are
impossible with rights and obligations (3b) .
(3) a. It is possible/necessary that Sue will leave.

b *It. "h/ bl" . {that

. IS a fIg t 0 Sue(will
)leave
IgatIon}3 .t:"
lor Sue to 1eave "

Rather , asrecognized by somemorerecentwritersondeonticlogicsuch

asForrester(1996 ), thepropertreatmentrecognizes two separateargu-
mentsof theseoperators : theholderof theright/obligationandthesitua-
tion with respect to whichthis personis entitledor obligated . Thefirst
argument of theseoperators mustbea person . Rocks , clouds
, andcom-
putersdonothaverightsandobligations . Animalsaresometimes asserted
to haverights, by construing themassemipersons ; theyneverhaveobli-
gations .4 In moderncapitalistlegalthought , corporations areconstrued
assusceptible to rightsandobligations andtherefore canenterinto con-
tracts;thelanguage usedto effectthisconstrualis thatcorporations count
as"legalpersons ."
In English , thesecond argument of theseoperators mustsyntactically
bea verbphrase(VP) whoseunderstood subjectis theholderof theobli-
gationor right. Thus, thearguments in (4a) areacceptable but thosein
(4b) arenot.

(4) a. 8uehas{ an obligation

a right } { to talk tothe
attend Harold
party} .

b. *8uehas{ a 0bl.19atlon
anright . } lor B.II to Ieave
1 sky
c { the to beblue} .
 72 Jackendoff

ThisVPis subject to semantic constraints . Bothrightandobligation

require thesituation tobenonpast withrespect tothetimeoftheobliga -
tion:theVPmaybepresent , future
, orgeneric time.
rightnow
(5) Suehas{ an . . } to1eave tomorrow .
aright
oblIgatIon whenever shegets annoyed
*yesterday5
TheVPofanobligation must express anaction thattheholder ofthe
obligation cancarry outvolitionally (6).
leave
scratchhernose
(6) Sue hasanobligation to *betall .
*bedescended fromroyalty
*bepaidforherwork
TheVPofarightcanexpress either anaction carriedoutbyitsunder -
stood subject (7a,b) orasituation inwhich itsunderstood subject receives
abenefit (7e).
a. leave
b. scratch hernose
(7) Sue hasarightto c. *bedescended fromroyalty.
d. *betall
e. bepaidforherwork
(8) illustratesthedifference between rightandobligation in thisrespect .
Theverbreceive does notdenote a voluntary actiononthepartofthe
recipient, buttheverb accept does
. Onlythelatter ispossible asanargu -
ment ofobligation.
(8) a. Sue hasarighttoaccept /receive payforherwork .
b. Sue hasanobligation toaccept /*receive payforherwork .
I willcallthekindofrightillustrated in (7e) a "passive " right , thatin
(7a,b) an"active " right ; obligationscanonlybeactive .
Forconvenience , I willcalltheperson having therightorobligation
theActor , andthesituation towhich therightorobligation pertains the
Action , withtheunderstanding thatthisincludes asaspecial case passive
rights, which donotinvolve anAction inthestandard sense. Given this
much ,wecanformalize rights andobligations asin(9), where RTand DB
areone -placeoperators applying totheAction . Asshown in(10 ), wethen
Logic of Rights and Obligations 73

say the Actor " has" the right or obligation (the use of HA VE will be
further justified shortly ).6

(9) a. RT (ACT (1'1)) == 'the right to do Action '

b . OB (ACT (1'1)) == 'the obligation to do Action '

( 10) a. HAVE (Xa , R T (ACT (a))) == 'X has a right to do Action '
b. HAVE (Xa , OB (ACT (a))) == 'X has an obligation to do
Action '

In these formulas , the Action is notated as a function of one variable , its

Actor (it may have further variables , irrelevant in the present context ) . In
( 10), this Actor position is bound to the holder of the right or obligation
by the bound variable Ct, where the superscript (X on X indicates that X
binds the variable in the argument position of ACT .
The fact that righ "ts and obligations have an Action rather than a prop -
osition as their argument places them in the general domain of deontic
logic , which deals with such notions as permissions and prohibitions and
with the logic of may , must, should, and ought, and which contains moral
reasoning as a particular subcase (von Wright 1963).
However , passive rights do not fall altogether comfortably into the
standard deontic domain , since their arguments are not volitional actions .
It is interesting therefore that the modal may cannot be used comfortably
to express passive rights : Sue may be paid for her work does not para -
phrase (7e). (Deserve to VP, which has a related meaning , has constraints
on its VP argument similar to those of right .)
A further constraint on the Action argument is deeply rooted in the
notions of right and obligation . Essentially , a right normally concerns
something one wants to do , whereas an obligation normally concerns
something one does not want to do .
( 11) a. Sue has a right / ??an obligation to eat her ice-cream sundae.
b. Sue has an obligation / ??a right to scrub the toilets .

The interpretations marked ?? are sensible just in case we assume Sue

doesn't like the ice-cream sundae and does like scrubbing toilets . (There
are exceptions , however , in which one has a right to do something odious
or an obligation to do something pleasurable .)
I will state this intuition in teffi1s of the value of the Action to the

Actor - positive for a right , negative for an obligation . I will notate it as a

function VAL VE that maps two arguments , a Stimulus and an Experi -
encer , into a Value .
 74 Jackendoff

( 12 ) VALUE ( Y , X ) = = + / -

' the value of Y ( Stimulus ) to X ( Experiencer ) is positive / negative '

The opposition between positive and negative value here is a primitive

affective distinction that may also be paraphrased by ' in vs . not in X ' s

interest ' . ( This seems to bear a relation to Freud ' s notion of Pleasure

vs . Unpleasure ( Lust vs . Un / ust ) . ) Using this notion , we can state the

constraint on rights and obligations as ( 13 ) . The principles are stipulated

to be defeasible ( ' defeatable ' ) to allow for cases in which other pragmatic

factors intervene to create exceptions .

( 13 ) a . HAVE ( XCi , RT ( ACT ( a ) ) ) defeasiblypresupposes

VALUE ( ACT ( X ) , X ) = = +

b . HAVE ( XCi , OB ( ACT ( a ) ) ) dejeasiblypresupposes

VALUE ( ACT ( X ) , X ) = = -

There are cases , such as the right / obligation to vote , that one may con -

strue with either value . In the right to vote , we take voting as a desirable

action ; in the obligation to vote , as somewhat burdensome . This confirms

the intuitions expressed by ( 13 ) . Similar effects can be discerned with the

choice between right and obligation in ( 4a ) and ( 5 ) .

In addition , rights and obligations have their own values : a right is

generally a good thing to have , an obligation a bad thing to have . We can

state this as ( 14 ) .

( 14 ) a . VALUE ( [ HAVE ( XIX , RT ( ACT ( a ) ) ) ] , X ) = = +

b . VALUE ( [ HAVE ( XIX , OB ( ACT ( a ) ) ) ] , X ) = = -

There is an indirect connection between ( 13 ) and ( 14 ) , to which I will

return in section 4 . 8 .

4 . 3 What One Can Do with Rights and Obligations

Next let us explore the range of things one can do with rights and obli -

gations beyond having them .

First , one can perform the action to which the right or obligation per -

tains . We speak of so doing as exercising the right or fulfilling the obliga -

tion . Notice that the collocations for right and obligation involve different

verbs for what ( at this level of description at least ) appear to be parallel

actions . We will see that such differences pervade the whole range of verbs

used with rights and obligations .

Logic of Rights and Obligations
75

Second, a right or obligation can be created. Sometimesthe creator of

an obligation is the Actor him- or herself. For example, promising is
(in part) creating and declaring an obligation upon oneselfto perform the
promised action. We speakin this caseof undertakingthe obligation. By
contrast, though one can declare or claim one's own rights, one cannot
thereby create them without the assentof other relevant parties. A per-
son's rights and obligations can also be created by an outside party,
whom I will call the Authority . We speak of the Authority 's giving,
granting, or conferring rights on the Actor , and of imposing obligations.
(I return to the status of the Authority in section4.9.)
For a slightly more complex case where rights and obligations are
created, considerX 's making an offer to Y to do such and such. This can
be construed as X 's conferring the right on Y to demand (i.e., impose an
obligation on) X to do such and such- an embedding of an obligation
within a right.
Third , a right or obligation can go out of existence. In certain cases,
performing the Action has this effect. For instance, handing the usher
one's ticket confers on one the right to attend a performance, after which
point the right ceasesto exist. Similarly, when a debt is paid, the obliga-
tion to pay it ceasesto exist. (Not all rights and obligations have this
property; seesection4.5.)
An Actor can also causea right to go out of existenceby renouncingit .
The counterpart for an obligation would be for the Actor to reject or
possibly renounceit . However, renunciation of an obligation does not
automatically make it go out of existence, even if the obligation is self-
imposed: we do not think well of someonewho revokespromises.
Under certain conditions, an Authority who has imposed an obligation
on an Actor can releasethe Actor from the obligation, or removethe ob-
ligation from the Actor , in which casethe Actor isfree of it . In the caseof
rights granted by an Authority , we speak of the Authority 's revoking or
taking away theserights- in which casethe Actor losesthem.
Fourth, one can transfer rights from one person to another, the first
party relinquishing them and the secondacquiring them. The parallel case
might be one person's taking on someoneelse's obligations.
Fifth , in a situation of conflict between the Actor and the Authority ,
the Actor may insist on a right, which the Authority is supposed to
acknowledgeor recognize. Alternatively , the Actor may try to get out of
an obligation, and the Authority may try to hold him or her to it .
Thesesituations are summarizedin table 4.1.
 76 Jackendoff

Table4.1
What one can do with rightsand obligations(# indicates'not necessarily
felicitous
')
Right Obligation
Perfofllling
Action exercise fulfill
Creating
by Actor # declare , claim undertake
by Authority give, grant impose
Voiding
by Actor renounce # reject
by Authority revoke , take away release, remove
(effect on Actor ) lose be free of
Transfer transfer take on
Conflict

Actor insist on get out of

Authority acknowledge hold to

The use of give, transfer , and take away alongside have suggests that
a right might be conceptualized along the lines of a possession, that is, as
an independent entity that one may have, give, or take away . On this
analysis, the verb have in have a right , notated as HA VE in ( 10), is essen-
tially the ordinary have of alienable possession.? This leads to analyses as
in (15) . (INCH is inchoative , or 'coming to pass' .)

( 15) a. CAUSE ( Y , [INCH HAVE (XCl, ( RT (ACT (cx))))])

== 'Y gives X the right to do Action '
b . CAUSE (Y , [INCH NOT HAVE (XCl, (RT (ACT (cx))))])
== 'Y takes away the right to do Action from X '

The language associated with obligations in table 4.1 has more incon -
sistent and opaque associations . Undertake , impose , remove , and take on
seem to image the obligation as a burden to be borne , as does the phrase
under ( the weight of ) obligation (s) . On the other hand , release, get out of,
and possibly hold to suggest the obligation is imaged as a constraining
force . In particular , the notion of an obligation as a constraint relates it to
force -dynamic expressions ( Talmy 1988): it is a social force that affects
one' s course of action . The expression fulfilling an obligation , through its
association withfill , might suggest an image of the obligation as a container .
The almost synonymous phrase meeting an obligation carries overtones of
Logic of Rights and Obligations 77

yet another image, one whosecharacteris difficult to pinpoint . In the face

of all these distinct associations, none of them coincident with those for
rights, I will adopt a foffi1alization for obligation exactly like (15), only
substituting 0 B for RT.

4.4 The Ontological Status of Rights and Obligations

Consider more closely the " images" associatedwith rights and obliga-
tions. A theorist in the vein of Lakoff and Johnson (1980) would claim
that rights and obligations are understood " metaphorically" and that
they derive their conceptual properties from another domain, called the
" source domain." According to Lakoff and Johnson's methodology, the
evidencefor identifying the sourcedomain comesprecisely from the col-
locations in which the words in question appear. In this particular case,
we would be inclined to claim that rights and obligations are understood
metaphorically in terms of different source domains- rights as posses-
sions, obligations as burdens or constraints. Yet, as we have already seen
to someextent, and as I will continue to document, rights and obligations
are near-twin concepts, with altogether parallel logic. So there is some-
thing a bit suspiciousabout the metaphor view.
An alternative view is that rights and obligations have their OM'n logic.
This logic is shared superficially with possessionsand burdens, but it is
close enough to draw an associativeconnection. In choosing verbs to
expresswhat one can do with rights, the languageis swayedtoward verbs
of possessionbecauserights, like possessions , generally are of positive
value; verbs relating to obligations are swayed toward verbs of physical
burden and constraint becauseobligations, like burdens and constraints,
generally are of negative value. On this view, one does not understand
rights and obligations metaphorically in terms of possessions and burdensf
constraints. Rather, becauseof what one understandsabout rights and
obligations, one choosesverbal collocations in a motivated fashion. While
acknowledging the insights that Lakoff and Johnson seekto express, this
view turns the notion of metaphorical understanding on its head: it is
becauserights and obligations are understood as they are that the meta-
phorical connection is possible- not the other way about. (For more
detailed discussionof Lakoff and Johnson's approach, seeMurphy 1996;
Jackendoff 1992, chap. 3; Jackendoff and Aaron 1991.)
Looking a little more deeply, what sort of conceptual entities might
rights and obligations be? In the formalization in (9), (10), and (15), the
 78 Jackendoff

right is an independententity, which someonemay have, may be given, or

may have taken away, and toward which someonemay expressattitudes
such as insistenceor acknowledgment. A more or less standard philo-
sophical approach to rights and obligations, observing that they involve
clausal complements, might suggestthat they are propositional attitudes,
like beliefs and desires. However, we have seenthat their arguments are
not propositions, but Actions (except for passive rights). More impor-
tantly, rights and obligations are emphatically not attitudes. Beliefs and
desiresare conceptualizedas being " in an individual 's mind" ; having a
belief or desire is being in a subjective mental state. By contrast, rights
and obligations are abstract entities, and having a right or obligation is
being in a certain objective social situation. To see this contrast more
clearly, notice that Bill 's belief/desireis in his mind is a sort of tautology,
but Bill 's right/obligation is in his mind attributes to Bill some kind of
delusion about his social relations. In fact, understandingsomeone's rights
and obligations requires no understanding of that person's mind. Here
rights and obligations differ sharply from the more standardly studied
speciesof deontic concept, moral/ethical understanding, which depends
heavily on empathy, that is, one's understanding of others' minds and
others' values (Goldman 1993; Hoffman 1987).
Beyond this observationI have little to offer about what sort of abstract
objects these are. They are perhaps a bit like facts- objectively deter-
minable persistent entities- a bit like responsibilities and opportunities,
a little less like credit and blame (both mass terms). More unexpected
parallels, pointed out by Searle(1995), are to the value of money and to
points in a game, which are again abstract but objective entities within a
social context. But sincethe ontology of such abstract objects in all their
variety has not been explored (to my knowledge), for the moment the
matter has to be left hanging.

4.5 " Existentially" versus" UniversallyQuantified

" Rightsand
Obligations

Section 4.3 spoke of rights and obligations that go out of existenceby

virtue of exercising or fulfilling them. To repeat, handing the usher a
ticket gives one the right to attend the perfoffi1ance- once; one has the
right no longer. Paying a debt erasesone's obligation to pay it ; returning
a borrowed item erasesone's obligation to return it . On the other hand,
not all rights and obligations are like this. Showing the usher one's year-
Logic of Rights and Obligations
79

long pass to the theater gives one the right to enter the theater , but one
retains the right for future occasions. Similarly , one's obligation to obey a
police officer does not go out of existence when one obeys an officer once:
this obligation persists.
We must distinguish , then , between rights and obligations that pertain
to exactly one action and those that pertain to all actions of a given type .
The former are exercised or fulfilled by an appropriate action 's taking
place, at which point they go out of existence. We could think of these
as " existential ," in the sense that if there comes to exist an Action that
satisfies the argument of the right or obligation , the right or obligation
ceases to exist . The latter , by contrast , are " universal " : they pertain to
every Action of the appropriate type . ( Von Wright calls these varieties
particular and general, respectively ; the legal tradition uses the terms in
personam and in rem, respectively .)
This distinction could be encoded by a subscript on the operators RT
and DB : for example , RTEx and OBEx versus RTun and OBun. Inference
rule ( 16) then pertains only to the " existential " variety .

(16) HAVE (XCi, RTEx/ OBEx(ACT (a))) at time t1, and

ACT (X ) at time t2, where t2 > t1,
entails

NOT HAVE (XCi, RTEx/ OBEx(ACT (a))) at time t3, where t3 > t2
It is important to our story that this inference rule involves a succession of
times- that is, the principles of rights and obligations require a dynamic
logic . (This temporal dependency is not present in the formalism of von
Wright and his successors, even if they acknowledge it informally .)
It is worth observing that various other concepts have this curious
property . For instance, an intention to perform an action is fulfilled - and
thereby goes out of existence- when the action is performed . Parallel
effects obtain with bodily sensations such as hunger , thirst , and some
itches. So this odd metaphysical property of rights and obligations is
actually more broadly attested among our concepts.
The situation is actually still more complex . One has the right to vote in
every election ; that is, exercising the right does not eliminate it . On the
other hand , one has the right to vote only once in each election ; having
voted eliminates the right till the next election . So this right has a mixed
flavor , partly " universal " and partly " existential ." This suggests that the
simple subscripts Ex and Un are not subtle enough to capture the range of
possibilities ; there is more internal structure to be teased out .
80 Jackendoff

4.6 Consequencesof Noncompliance

So far , I have not done much to distinguish social/ legal/ contractual rights
and obligations from other kinds of deontic operators expressed by
modals such as may , should, ought to, and must- in particular , from
" moral obligations " and " human rights ," which turn out to be a bit
different in character . Now I turn to what gives social/ legal/contrac -
tual rights and obligations their distinctive flavor : the consequences of
noncompliance .
What happens if one fails to fulfill an obligation ? Very simply , one runs
the risk of getting in trouble . Suppose I have undertaken an obligation -
say, by promising to wash the dishes. Making a promise involves another
individual to whom one has made the promise , who typically will benefit
from having the promise fulfilled . Let us call this individual the Benefi-
ciary of the obligation .
N ow suppose I do not wash the dishes within a reasonable amount of

time . (What counts as a reasonable amount of time is a delicate matter .)

Then the Beneficiary has the ( Existential ) right to impose some sort of
punishment on me. The Beneficiary does not necessarily exercise this
right , but we clearly understand that this right exists.
What does it mean to impose punishment ? Roughly , if Z punishes X , Z
performs some action with negative value to X , in return for some previ -
ous action on the part of X with negative value to Z . (Reward is the same,
except with positive values.)
This is enough information to enable us to state a preliminary version
of the inference rule for nonfulfillment of obligations . First , we must add
to the predicate DB an argument position for the Beneficiary , as in (17).
(This character is not recognized , even implicitly , in the work on deontic
logic with which I am familiar .)

( 17) OB (ACT (a), TO Z ) = 'the obligation to Z to do Action '

We may further add that the Beneficiary normally benefits from the
Acti on .

( 18) OB (ACT (lX), TO Z ) (defeasibly) entails V ALOE (ACT (a), Z ) == +

The inference rule for noncompliance now comes out as ( 19); again it is
stated dynamically .
 81
Logic of Rights and Obligations

(19) HAVE (XCi, OB (ACTI (C(), TO Z )) at tl and

NOT ACTI (X ) in period from tl to t2
entails

ACT2 ( P )

HAVE ZP , RTEx ; ~a ( V ALUE ( a , X ) = = - ) at t2

EXCH ( NOT ACTl ( X ) )

In the bottom line of (19), ACT 2 is the action that Z has the right to carry
out; the lambda- expression says that this action has a negative value
for X ; the EXCH operator says that this action is in exchangefor X 's
nonperformance.
Rule (19) saysnothing about the appropriate time interval to wait for
compliance. More important , (19) saysnothing about what kind of retal-
iatory act is appropriate- only that it should be somethingthat the Actor
won't like. Many such actions, especially for culturally loaded obliga-
tions, are prescribed in a culture's stock of customs and oral or written
law.8 In particular, there are large classesof obligations, including such
things as debts, for which the appropriate action in the face of non-
compliance is to call in the authorities to determine the appropriate pun-
ishment. In turn, social norms may dictate that the authorities have an
obligation to the Beneficiary to mete out the appropriate punishment-
one of the foundations of legal theory (Stone 1968). I return to the notion
of " the authorities" in section4.9.
In addition to (19), the nonfulfillment of an obligation has a broader
consequence . Roughly speaking, ever} 'one- not just the Beneficiary- is
justified in criticizing the Actor (or thinking less of the Actor ) for failing
to fulfill the obligation. The threat of such criticism seemsto me to con-
stitute the moral dimension of an obligation. Here I would be reluctant to
use the term has a right to describewhat everyonemay do to the Actor ;
something like is mol'ally justified seemsmore appropriate. I will not
attempt to formalize this inference, in part becausean appropriately gen-
eral form is beyond the scopeof this chapter.
Thesetwo inferences, the one social/ contractual, the other moral, seem
both-to be involved in social/ contractual obligations suchas promisesand
debts. The kinds of things we might call " moral obligations," such as the
obligation to preservethe environment, seemto me to invoke only the
latter inference rule.9 It is interesting how the two rules are somewhat
parallel in structure, but not entirely. In particular, only the Beneficiary
has a right of retribution , whereaseveryoneis morally justified in sanc-
tioning the offender.
 82 Jackendoff

Turning to rights , we find a related situation . Suppose I have given

the usher my ticket . This gives me a right to enter the theater to see the

performance . Now suppose someone tries to prevent me from going in ,

saying , " You have no right to go in ! " I am thereby entitled to take action

against this person : my right has been violated . Notice , by contrast , if I

have tried to go in without presenting a ticket , I 'm not entitled to express

disapproval , since it ' s the rights of the theater ( as corporate body ) , not

mine , that are being violated .

This scenario leads us to the counterpart of rule ( 19 ) for rights . If I

have a right , I may choose or not choose to exercise it . But if I attempt

to exercise it , and some other party tries to prevent me , I then have the

( Existential ) right to exact punishment on that person . ( 20 ) formalizes this

intuition .

( 20 ) HAVE ( Xa , RT ( ACTI ( a ) ) ) at tl and

CAUSE ( Z , NOT ACTI , ( . X -) )- at t1.

entails

ACT 2 (P )

HAVE XP , RTEx Aa ( V ALUE ( a , Z ) ) == -

EXCH ( CAUSE ( Z , NOT ACT1 ( X ) ) ) at t1

at t2 > t1

Again , there are many cultural customs and norms concerning what kind

of retaliatory Action2 is appropriate to what kind of Action ! , and under

what sort of relationship between X ( the Actor ) and Z ( the Right -

Violator ) . In particular , just as in the case of unfulfilled obligations ,

X 's right in many cases will consist of a right to go to the authorities

to demand retaliation . For instance , if the usher doesn 't let me into the

theater despite my having presented my ticket , the appropriate action ,

after due remonstration , is to go to the manager , and if that doesn ' t work ,

to the police .

As in the case of obligations ; there is also a moral dimension to violation

ofa right : everyone is morally justified in criticizing ( or thinking less of ) the

person who violates someone ' s rights . Again , only the Actor has the right

of retribution , but everyone is morally justified in criticizing the offender .

Rules ( 19 ) and ( 20 ) , though nearly symmetrical , have one important

difference . In the case of obligati 'ons , there is a specific Beneficiary of the

obligation . As seen in ( 19 ) , this is the individual who acquires the right of

retaliation if the obligation is not met . In the case of rights , there seems to

be no such specifically identified individual . Anybody who tries to prevent

one from exercising a right is a potential target for justified retaliation .

Logic of Rights and Obligations 83

This difference , however , conceals a deeper similarity : the individual

entitled to retaliation is always the one for whom the potential Action is of

positive value . In the case of a right , the Actor is the potential Beneficiary

and therefore receives the right of retaliation for interference with receiv -

ing the benefit . In the case of an obligation , the Action is of negative value
to the Actor ; the reason the Action is to be performed is to benefit some -

one else . It is that someone else , then , who receives the right of retaliation .

Notice too that this parallelism extends to the moral dimension . Every -

one is morally justified in criticizing the person who exacts the cost from

the intended Beneficiary : the Actor in the case of an obligation , the person

who interferes with the Actor in the case of a right .

My sense is that inference rules ( 19 ) and ( 20 ) are the central principles

that make rights and obligations what they are . By contrast , a moral /

ethical principle takes the form " One should / should not do such and

such " ; that is , it is similar relation of an individual to an action . But

it does not carry inferences concerning rights of retaliation . Rather , it

carries only what I have called the moral dimension , the fact that every -

one is morally justified in criticizing the Actor for noncompliance . It is

the need to regulate modes of retaliation in response to violations of

social / contractual obligations and rights that leads to the development of

legal and judicial systems in a society . So these inference rules lie at the

foundation of social / cultural cognition .! 0

4 .7 Reciprocal Rights and Obligations

To amplify the symmetries we have been observing : It is often felt that

there is a sort of reciprocity between particular rights and obligations . In

particular , the Beneficiary of an obligation is felt to have a " passive right "
to receive the benefit . This sense can be stated as inference rule ( 21 ) .

( 21 ) HAVE ( XCi , [ OB ( ACT ( a ) , TO Z )])

entails

HAVE ( Z , [ RT ( ACT ( X ))])

The bottom line of ( 21 ) is a passive right because Z is not the Actor of the

Action . However , by entailment ( 18 ) , X ' s action is beneficial to Z , so this

Event satisfies the conditions for the argument of a passive right , as illus -

trated in ( 7e ) and ( 8a ) .

Now let us apply inference rules ( 19 ) and ( 20 ) to both lines of ( 21 ) .

When we apply ( 19 ) to the first line , the inference is that Z is entitled to

 84 Jackendoff

retaliation if X does not perform the Action . When we apply ( 20 ) to the

bottom line , the inference is that Z is entitled to retaliation toward anyone

who prevents Z from receiving the benefit of X ' s acting - in particular ,

toward X if X does not perform the Action . So the two lines lead to

related inferences just in case X does not comply .

A right imposes a similar sort of reciprocal obligation : the obligation

falls on everyone not to infringe on someone ' s rights . 11 In particular , the

individual who grants the right is particularly bound to respect it . The

general rule can be stated as ( 22 ) . ( In the interest of uncluttered typogra -

phy , I use the notation EVERYONE instead of a standard universal

quantifier with bound variable .)

( 22 ) HAVE ( XIX , RT ( ACT ( a ) ) )

entails

HAVE ( EVERYONEP ,

[ OB ( NOT CAUSE ( P , NOT ACT ( X ) ) ) , TO X ] )

Notice that the obligation in ( 22 ) is not to perform an action - a sort of

counterpart of a passive right . Again , the outcome of someone ' s infring -

ing on a right , as entailed by rule ( 20 ) , is the same as the outcome of

someone ' s failing to meet the obligation not to prevent exercise of a right ,

entailed by rule ( 22 ) . So perhaps ( 22 ) is logically unnecessary . Neverthe -

less , something like it is often stated explicitly ( e . g . , in Stone 1968 ) , so I

include it for completeness .

4 . 8 Equality of Value Implied by EX CH

Next let us consider more closely the notion of exchange , symbolized by

EXCH in ( 19 ) and ( 20 ) . This notion not only appears in the entailments

of rights and obligations , but is also expressed explicitly by means of a

certain use of the preposition for in English .

( 23 ) a . Susan praised her son Sam for behaving nicely .

b . Fred cooked Lois dinner for fixing his computer .

( 24 ) a . Susan insulted Sam for behaving badly .

b . Lois slashed Fred ' s tires for insulting her sister .

These sentences describe situations in which someone does something

" in return " for someone else ' s action . Those in ( 23 ) describe actions with

positive values ; those in ( 24 ) describe actions with negative values . Ex -

changes can felicitously take place only with another person , an entity
Logic of Rights and Obligations 85

that can be regarded as having values and responsibility . One cannot

sanely punish one' s car for getting a flat tire .
If we switch around the actions in (23) and (24), we get sentences that
sound odd or perhaps ironic .
(25) a. Susan insulted Sam for behaving nicely .
b. Lois slashed Fred 's tires for fixing her computer .

This shows that we expect a positively valued action in return for a posi-
tively valued action , and a negatively valued action in return for a nega-
tively valued one.
So far I have spoken of values as only positive or negative . But ex-
change situations show that values are conceived of as roughly quantita -
tive . We find it odd if the two actions related by for do not match in
quantity . The sentences in (26) convey some of this oddness.

(26) a. Fred cooked Lois dinner for saying hello to him .

b . Fred cooked Lois dinner for rescuing all his relatives from certain
death .

c. Fred slashed Lois 's tires for eating too little at dinner .
d . Fred slashed Lois 's tires for murdering his entire family .

In (26a) and (26c), we sense Fred as overreacting , as doing something

unwarranted in return for Lois 's action ; in (26b) and (26d), we sense him
as underreacting , as doing something that is not nearly enough to recog-
nize the importance of Lois 's action .
The intuition , then , is that an exchange situation presupposes rough
equivalence of value between the two actions . For a first approximation ,
then , we could state a principle for the value of exchanges along the lines
of (27). ( The presupposition is defeasible in order to allow for exceptions
- for example , merely thanking someone publicly for performing some
heroic act .)

(27) [ACT2
EXCH(Z ) )] ]
[ACT1(X
defeasibly presupposes
VALUE (ACT2 (Z ), X ) == VALUE (ACT } (X ), Z )

In practical situations , the equation is not this simple : each participant 's
judgment of the value of his or her own and the other 's actions , both to
himself or herself and to the other , may differ . In particular , there seems
to be a general cognitive bias toward overestimating costs to oneself and
benefits to others , and toward underestimating benefits to oneself and
 86 Jackendoff

costs to others . Thus , it often requires negotiation to achieve a " fair " ex-
change, where both participants judge the exchanged acts to be of equiv -
alent value to themselves and to the other . Here is an important point
where " folk theory of mind " and Cosmides ' ( 1989) " cheater detection "
enter : one is more inclined to compromise if one believes the other 's
assertions of value are made in good faith .
The logic of exchange laid out here is a cognitive elaboration of a be-
havioral strategy well documented in the ethological literature , recipro -
cal altruism : " You scratch my back and I ' ll scratch yours ." I leave open
how much of its detail can be attributed to nonhuman primates (not to
mention elephants and bats). What strikes me as particularly human ,
though , is the broad generality of the actions available to be entered into
exchange equations .
Turning back to rights and obligations more specifically , we see in (27)
a more or less formal statement of " The punishment fits the crime " : this
helps guide what actions are appropriate in retaliation for breaking obli -
gations and violating rights in ( 19) and (20). (27) implies that the offend -
ing party should receive retribution : the cost of the retaliatory action to
offenders should equal the cost of their offense to the injured party - that
is, the two parties should end up equally badly off . An alternative to (27)
is that the injured party should receive restitution : the benefit of the re-
taliatory action to the injured should equal the cost of the injury , what -
ever its cost to the offender - that is , the offended party should end up as
well off as before .

(28) Retribution : VALUE (ACT2 (Z ), X ) == VALUE (ACTI (X ), Z )

Restitution : VALUE (ACT 2 (Z ), Z ) == - VALUE (ACT 1 (X ), Z )
We know that the two are not equivalent . For one extreme case, " N oth -
ing you can do to the murderer will bring my son back " ; that is, restitu -
tion is impossible , though retribution might be. (The distinction between
these two in conceptual development is noted by Piaget ( 1932); he claims
the notion of restitution is established later than that of retribution .)
Combining (27) with ( 19) and (20) gives us an inferential link between
( 13) and ( 14), that is, between the value of rights and obligations and the
value of the actions to which they pertain . Consider first obligation . Since
the action to which one is obligated has a cost, and not performing the
action risks a cost (just in case the Beneficiary exercises the right of
retaliation ), then having an obligation is basically a lose-lose situation -
that is, obligation itself is of negative value . Conversely , since the act to
Logic of Rights and Obligations 87

which one has a right is of positive value , and being prevented from
exercising the right grants one a right of restitution , then having a right
is a win -win situation , that is, a benefit . So some of the pieces of the logic
of rights and obligations begin to hang together .

4.9 Authority

Consider who can impose an obligation on yoU.12 As pointed out in sec-

tion 4.3, the simplest case is a self-imposed obligation such as a promise .
Other things being equal , you are free to make whatever promises you
wish . But no one else can impose an obligation on you unless particular
conditions obtain . For example , if a random person says to you , " I here-
by oblige you to wash my feet," you are justifiably offended and baffled .
The felicity conditions for such a speech act are not met , just as if a ran -
dom person were to declare to you , " I hereby name you Fuzzy -Wuzzy ."
One felicitous condition under which an obligation can be imposed by
someone else is if you have granted the other person the right to impose it ,
either by making an offer or by making an agreement or contract . You
are perfectly free to grant such a right .
However , there is another situation in which someone can impose
obligations on you : when that person has authority over you through
position in the social hierarchy .13 A society presents many authority
relationships - for instance , parent to child , boss to worker , sergeant to
private . The authority relationship entails the Authority 's right to impose
obligations on the subordinate . If , for a first approximation , we encode
the authority relationship as (29), then (30) expressesthe Authority 's right
to created obligations for subordinates .

(29) HAVE (Z , AUTHORITY (X ))

== 'Z has authority over X '

(30) HAVE (Z , AUTHORITY (X ))

entails

HAVE (ZIX, RT (CAUSE (a, INCH HAVE (Xp ,

(OB (ACT (P), a))))))
The language used to express what one can do with authority is virtu -
ally identical to that for rights listed in table 4.1. We speak of the exercise
of rights and of authority ; a higher authority can give or grant rights or
authority - or revoke them or take them away . One can renounce rights or
authority that one currently holds; or one can through malfeasance lose
88 Jackendoff

rights or authority . One can insist on one's own rights or authority ; one
can acknowledge someone else's. On the other hand , there is more com -
plexity in authority , since one can resist another 's authority , but the
phrase resisting anotherJs right makes little sense.
Like all obligations , those imposed by an Authority must have a
Beneficiary . (30) encodes the typical case in which the Beneficiary is the
Authority him - or herself (this is notated by the (Xin the second argument
posiion of OB , bound to Z ) . Thus , in case of noncompliance the Author -
ity also has the right of punishment . Other Beneficiaries are possible- for
instance, when a judge obliges a divorced parent to pay child support to
the ex-spouse. In such a case, the ex-spouse's right of retaliation in case of
noncompliance is typically deteffilined by the judge as well . That is, if
an Authority imposes an obligation on an Actor , with a third individual
as Beneficiary , the Authority retains the right to punish the Actor for
noncompliance , with or without appeal from the Beneficiary . So the logic
becomes still more complex ; I will not attempt to foffilalize it here.
Even with these added complexities , the account in (30) is still missing
an important caveat, in that it is necessary to recognize limitations of
authority . For instance , in our society , we believe that your boss does not
have the right to oblige you to take your clothes off . A more adequate
formalization relativizes authority to a particular class of actions , as in
(31) .

(31) HAVE (Z , AUTHORITY ( Xp , ACTa (P)))

= 'Z has authority over X with respect to actions of type a '
The appropriately revised form of (30) is (32) (where I also add the
Authority 's right to confer rights ) .

(32) HAVE (Z , AUTHORITY ( Xp , ACTa (P)))

entails

HAVE (ZCX , RT (CAUSE (a, INCH HAVE (Xp ,

( RT / OB (ACTa (P), a))))))

(32) leaves about the right loopholes for social negotiation (and conflict ) :
Over exactly what actions can a given Authority impose obligations ? And
how are those decided? These are issues with which every society must
grapple .
How does one obtain authority ? One way is to be granted it by a
higher Authority , who is then said to be delegating authority . But this
leaves open who grants authority at the top of the pyramid . This problem
Logic of Rights and Obligations 89

of the " apex norm" (Stone 1968, following Hans Kelsen) lies at the root
of a society's conception of itself. Three possible solutions,: despotism,
where the ultimate authority simply assertsauthority without recourse
and maintains it through the exerciseof power; supernatural authority
such as the " divine right of kings," in which the top-ranked personis said
to be granted authority by a deity whose rights in turn require no justifi -
cation; and representative government, in which authority is taken to
arise from the " consentof the governed." In addition, all societiesrecog-
nize the " natural" authority of parentsover children, which seemsto need
no justification . Perhapsthere are other possibilities.

4.10 WhereDoesIt ComeFrom?

I have surely left many subtletiesstill untouched- and somemajor points

as well, such as how to reason about conflicting rights and obligations,
and how to characterizerights as legitimate or illegitimate.14 Neverthe-
less, I will now step back a bit .
I have shown that notions of right and obligation are richly inter-
connectedby inferencesto eachother and to the notion of the value of an
action to an individual . The central inference rules (19) and (20) depend
on the notion of linking two actions as exchangesof value, and through
that on notions of fairness or justice. The notion of authority depends
heavily on the notion of rights. In short, many of the conceptual founda-
tions of social organization either depend on an understanding of rights
and obligations, or else are developedin justification of the assertion of
particular rights and obligations.
Like the sorts of conceptswith which John Macnamara was so deeply
concerned, the conceptsof right and obligation are, it seems,quite abstract,
not linked to perception of the physical world except very indirectly. In
fact, the analysis here suggeststhat their content may lie entirely in the
inferencesthat can be drawn from them. They are, as it were, part of an
elaborate social accounting systemfor keeping track of the implications
of an individual 's actions with respect to others, a system rooted ulti -
mately in the notion of value.
In order for an individual to function in a society, then, it is essential
for him or her to intuitively grasp the conceptsof right and obligation.
Indeed, most of the discussionhere has consisted of pointing out intui -
tions that all of us share. So the question arises of how people acquire
theseconcepts. As Macnamara puts it (1991), how does one gain entry to
90 Jackendoff

a systemof interrelated terms and ideas, if they cannot be definedin terms

of some other system?
There is no question that people must learn the particular network of
rights and obligations inculcated (or presupposed ) by their society: who
has an obligation to whom, who has a right to impose obligations and
grant rights over what actions, what retaliation is appropriate for failure
to meet what obligation, and so forth . This must by all meansbe a major
part of cultural learning. But, returning to a point made at the outset, it is
less clear that people must learn that there are such things as rights and
obligations. As far as I know, every culture sharestheseconcepts. They
seemto be building blocks as fundamental to understanding the social
world as force is to understandingthe physical world . (This point is also
made by Forrester (1996).)
Moreover, the inferential patterns of rights and obligations have
no analogue in the physical (or sensorimotor) domain, such that there
could be a progressionin learning along the lines of Piaget, or a learning
through metaphor along the lines of Lakoff and his colleagues. The latter
possibility was rejected in section 4.4, even before approaching the com-
plexities of retaliation and exchange, for which a physicalistic metaphor
is still more far-fetched. (In fact, if anything, the tendency often goes
the other way: people attempt to understand the physical world by
anthropomorphizing it into a metaphorically social world full of wills and
desires.)
It seemsto me, therefore, that an important question for researchinto
social cognition is how the child learns the conceptsof right and obliga-
tion- or whether they are learned at all. The latter possibility, not to be
discounted, is that theseconceptsare largely if not entirely innate, a spe-
cialized " way of thinking" wired into the brain by the human genome.
Such a hypothesiswould certainly account for the cultural universality of
theseconcepts: they would form a preestablishedspecies -wide skeleton of
social understnding over which each particular culture builds its own
flesh. Under this hypothesis, the child learning a culture would then come
to the task predisposedto interpret the social world in terms of rights and
obligations, among other things. If there is an identifiable developmental
stage where such concepts become available, relatively uniform across
cultures, this might well be interpreted as evidenceof biological matura-
tion of the brain. If so, the argument would be parallel to the arguments
for the biologically basedlanguagecapacity that makes language acqui-
sition possible(Chomsky 1965; Lenneberg 1967).
Logicof RightsandObligations 91

I am not aware of any researchthat bearsdirectly on the acquisition of

these precise concepts. However, suggestive evidence appears in the
experimentalwork of Piaget (1932), who discussesthe developmentfrom
ages6 to 11 of the child's understandingof related deontic conceptssuch
as the rules of gamesand of moral conceptssuch as prohibition and fair-
ness, as well as their relation to authority . In particular, rules of games
have some of the same " objective" ontological status as obligations, so
should provide suggestiveevidence.
Looking earlier in child development, work by Cummins (1996, to
appear) and Harris and Nunez (to appear) investigateschildren's under-
standing of the deontic concepts of prohibition , permission, and reci-
procal exchangeat the ages of 3 and 4. These authors find a degreeof
understanding of theseconceptsmore sophisticatedthan one might have
expectedfrom Piaget's research- in fact, an understandingmore reliable
than with equally complex propositional statements. For instance, Cum-
mins compares3- and 4-year-olds' understandingof a prohibition such as
" All squeaky mice have to stay in the house" with that of a declarative
statementsuch as " All the squeakymice are in the house." The children's
understanding is tested by asking them which (toy) mice they have to
check in order to determinewhether the order has beencarried out or the
statement is true. The question is phrased in such a way that the very
samemice must be checkedunder both conditions. Cummins finds that
the children are much more reliable with the prohibition than with the
statement. (Notice also that this is not exclusively ego-centered deontic
understanding, as might be acquired through a child's experiencewith
parents' orders: the task involves checkingwhether other individuals have
obeyed someone else's order.) Cummins concludes that this aspect of
deontic understanding is in place early on in development. This is not a
test of the entire logic of rights and obligations, but at least an important
part.
If the logic of rights and obligations were part of the human endow-
ment, it would have likely emergedfrom some evolutionary antecedent.
So someprecursorsmight be expectedin the social behavior of primates.
In section4.8 I have already alluded to reciprocal altruism and aggression
as a precursor of exchange. More pointedly, Hauser et al. (1995) describe
a behavior in rhesusmacaquesthat appearsto presenta precursor of obli-
gation. It seemsthat when an individual finds a food sourceon his own,
there is a special call he utters to signal to the troop that food has been
found. If the individual does not utter the call, of course, he will benefit
92 Jackendoff

from a greater quantity of food for himself, so there is a payoff in not

uttering the call. However, if other membersof the troop discover the in-
dividual eating and the call has not beenuttered, then they beat him up-
even if he is the most dominant individual in the troop. It is striking,
however, that nonmembersof the troop will not be beatenup under par-
allel conditions. One is highly tempted to interpret this scenario as an
understanding on the part of the monkeys that one is " obligated" to the
membersof one's own troop to signal the finding of food. In particular,
what makesthis suite of behaviorslook like obligation is that the right of
punishment, expressedhere as inferencerule (19), seemsto be an essential
component. Although it is important not to overinterpret ethological ob-
servations, the detail of current work like that of Hauser and his col-
leaguesbids us not to underinterpret either. I would not go so far as to
call this behavior illustrative of an understandingof obligation in the full
human sense, but it certainly appears to be on the way. If so, it again
provides indirect evidencethat the notion of obligation has a biological
basis.
Three directions suggestthemselvesfor investigating the issuefurther.
First, crosslinguistic and crosscultural work on the language and under-
standing of rights and obligations would add a great deal to the analysis.
Second, I have barely scratchedthe surface of the formal detail of rights
and obligations, their dynamic functioning in social reasoning, and their
relationships with other social concepts(including dominance hierarchies
and moral and ethical thought). To figure out exactly what the child has
to learn- and what the child can learn- it is crucial to pull theseconcepts
apart further into their components and to see what external evidence
could lead to the acquisi~ion of such components. A third direction is to
use formal analysis of this sort to help guide further researchin anthro-
pology, primatology, and especiallychild developmenton this topic that
is so vital to our social existence. It is my hope that researchersbetter
versedthan I in thesedisciplineswill be stimulated by the presentchapter
to undertake the challenge.
Notes

1. To be politically correct, one might justifiably ask whether taking thesenotions

for granted is a cultural bias on the part of anthropologists, and whether other
cultures might indeed have quite different notions underlying their social organi-
zation. I don't think so: the apparent successof anthropological description- the
fact that one can make senseof cultures while taking thesenotions for granted-
suggeststhat there is little danger of conceptual chauvinism on this particular
point of analysis. I might be wrong, of course; but I suggestthat proving me
Logic of Rights and Obligations 93

wrong requires more than a blanket invocation of cultural relativism . See Brown
1991 for discussion .

2. Although , of course, the conceptualized functions of streets and buses must

involve the actions and goals of persons.
3. I follow the notational convention of linguistics in marking unacceptable sen-
tences with a *.

4. Though I gather that during the medieval period there were such things as
trials of pigs for killing children , pushing pigs toward a more responsible status .
5. Sue had a ,~ight /obligation to leave yesterday is of course acceptable . In this case,
the time for which the right / obligation is asserted is yesterday or earlier - not five
minutes ago , for instance .

6. For readers unacquainted with this style of formalization , the point is to have a
canonical and precise way of expressing claims about meaning , which permits us
to state rigorous principles by which meanings may be composed and by which
they may be used to draw inferences . The notation used here can be thought of as
an enriched form of predicate logic . For those uncomfortable with the formalism ,
I will take care to explain the formulas in ordinary English as well .
In ( 10) and succeeding examples I use the predicate HA VE to express posses-
sion , following Pinker 1989. An alternative in line with the practice of Jackendoff
1983 would be (i); I leave it to the interested reader to work out the differences in
the rules throughout the chapter .
(i) BEposs([RT / OB (ACT (ri))], ATposs XIX)
7. There is however a degree of circularity in this suggestion , in that , as men -
tioned at the outset , possession in large part consists of rights . I will not attempt
here to grapple with problems this may raise.
8. We can include among " customs " the sorts of things discussed in " Miss Man -
ners " columns in the newspaper : what do I do if someone doesn ' t fulfill the cus -
tomary obligation of writing a thank -you note for a wedding present?
9. What I am calling here the moral dimension is the main one considered by
Conison ( 1997) , who draws the inference that everyone is entitled to sanction
someone who breaks a promise . Yon Wright ( 1963, 12), however , says, " 'By defi -
nition ' , one could say, promises ought to be kept . But this is only one aspect,
beside others , of the obligation in question . . . . To try to explain the obligation to
keep promises , for example , in terms of the 'normative pressure' of customs seems
utterly out of place ." This is the intuition I am trying to capture here.
10. In certain religious traditions , moral / ethical strictures are taken to be obliga -
tions to a deity , and the deity acquires the right of retaliation . I take this to be a
cultural construct whose purpose is to sharpen ethical norms , giving them the
same " objective " status as laws .
11. " Infringe " : yet another thing one can do with a right . This verb has associa-
tions of treading on another ' s territory , a variation on possession.
12. This section is a particular area where I realize I run a risk of overly pre -
supposing something like the Western democratic tradition , with its stress on the
 94 Jackendoff

rights of the individual . If the logic works out differently for different cultures , so
be it .

13 . Authority is a form of social dominance , but not the only one . For example ,
siblings often fall into a dominance hierarchy by age , but that does not necessarily
make it possible for an older sibling to impose obligations on a younger .

14. In particular , I have not dealt with what many take to be the most important
issue concerning rights : the notion of universal human rights . I see some difference
between these and the more mundane type of rights discussed here . Unlike con -
tractual rights , issues of human rights typically arise in the context of claiming
rights that are not acknowledged by governmental authority . In the time of the
American Revolution , such rights were asserted to be granted by higher (divine or
natural ) authority (as in " . . . endowed by their Creator with certain unalienable
rights . . ." ) ; in this century , they have been to some extent acknowledged through
international agreements such as the United Nations Universal Declaration of
Human Rights . In either case , there is to some extent a failure of mutual belief
in the existence of the rights in question , as well as questionable effectiveness in
enforcement on the part of the higher authority .

References

Bloom , P ., Peterson , M . A ., Nadel , L ., and Garrett , M . F . ( Eds .) . ( 1996 ) . Lan -

guage and space . Cambridge , MA : MIT Press .
Brown , D . ( 1991 ) . Human universals . New Yark : McGraw -Hill .

Cheney, D ., and Seyfarth, R. (1990). How monkeyssee the world. Chicago: Uni -
versity of Chicago Press.
Chomsky, N . (1965). Aspectsof the theory of syntax. Cambridge, MA : MIT Press.
Conison, J. (1997). The pragmatics of promise. Canadian Journal of Law and
Jurisprudence, 10, 273- 322.
Cosmides , L . ( 1989). The logic of social exchange: Has natural selection shaped
how humans reason? Studies with the Wason selection task . Cognition , 31, 187-
276.

Cummins , D . D . ( 1996) . Evidence for the innateness of deontic reasoning . Mind

and Language , 11, 160- 190.

Cummins , D . D . (to appear ). Evidence of deontic reasoning in 3- and 4-year-old

children . Memory and Cognition .
Forrester , J. W . ( 1996). Being good and being logical : Philosophical gro undwork for
a new deontic logic . London : M . E . Sharpe .
Goldman , A . ( 1993). Ethics and cognitive science. Ethics , 103, 337- 360.
Gruber , J. S. ( 1965) . Studies in lexical relations . Reprinted 1976 as part of Lexical
stl'uctures in syntax and semantics. Amsterdam : North -Holland .
Harris , P., and Nunez , M . (to appear). Children 's understanding of permission
and obligation . In L . Smith , J. Dockrell , and P. Tomlinson ( Eds .), Piaget ,
Vygotsky and beyond. London : Routledge .
Logicof RightsandObligations 95

Hauser , M ., Clark -Schmidt , K ., Davis , R ., and Leeman , B . ( 1995) . Punishment

and convention in rhesus monkeys . Unpublished manuscript , Departments of
Anthropology and Psychology , Harvard University , Cambridge , MA .
Herskovits , A . ( 1986) . Language and spatial cognition . Cambridge : Cambridge
University Press.
Hoffman , M . L . ( 1987) . The contribution of empathy to justice and moral judg -
ment . In N . Eisenberg and J. Strayer ( Eds.), Empathy and its development. Cam -
bridge : Cambridge University Press.
Jackendoff , R . ( 1976) . Toward an explanatory semantic representation . Linguistic
Inquiry , 7, 89- 150.
Jackendoff , R . ( 1983). Semantics and cognition . Cambridge , MA : MIT Press.
Jackendoff , R . ( 1990) . Semantic structures . Cambridge , MA : MIT Press.
Jackendoff , R . ( 1992) . Languages of the mind . Cambridge , MA : MIT Press.
Jackendoff , R ., and Aaron , D . ( 1991). Review of George Lakoff and Mark
Turner , More than cool reason. Language , 67, 320- 338.
Lakoff , G ., and Johnson , M . ( 1980) . Metaphors we live by . Chicago : University of
Chicago Press.
Lenneber ~, E . ( 1967) . Biologicalfoundations of language. New York : Wiley .
Macnamara , J. ( 1991) . The development of moral reasoning and the foundations
of geometry . Journalfor the Theory of Social Behaviour , 21, 125- 150.
Miller , G ., and Johnson -Laird , P. ( 1976) . Language and perception . Cambridge ,
MA : Harvard University Press.
Murphy , G . ( 1996). On metaphoric representation . Cognition , 60, 173- 204.
Piaget , J. ( 1932) . The moraljudgment of the child (trans . M . Gabain ) . New York :
Free Press .

Pinker , S. ( 1989). Learnability and cognition : The acquisition of argument struc -

ture . Cambridge , MA : MIT Press.
Searle, J. ( 1995) . The construction of social reality . New York : Free Press.
Snare, F . ( 1972) . The concept of property . American Philosophical Quarterly , 9,
200 - 206 .

Stone, J. ( 1968) . Human law and human justice . Stanford , CA : Stanford University
Press .

Talmy , L . (1983) . How language structures space. In H . Pick and L . Acredolo

( Eds .), Spatial orientation : Theory , research, and application . New York : Plenum .
Talmy , L . ( 1988) . Force dynamics in language and thought . Cognitive Science, 12,
49 - 100 .

Vandeloise , C . ( 1986). L 'espace en jran9ais . Paris : Editions du Seuil . Trans . as

Spatial prepositions . Chicago : University of Chicago Press ( 1991) .
Wright , G . H . von ( 1963). Norm and action : A logical enquiry . London : Routledge
& Kegan Paul .
Chapter
- 5
Deliberation Reasons and Storrs McCall
Explanation Reasons

Practical deliberation - bouleusis- is discussed by Aristotle in books 3

and 6 of the Nicomachean Ethics . Although Aristotle doesn't mention
this , it is necessary to distinguish practical deliberation from what may be
called " cognitive deliberation ." Cognitive deliberation is deliberation
over whether something is true or false, whereas practical deliberation is
over what to do . In jury trials , for example , the jury is asked to decide the
cognitive question , is the accused guilty as charged? Its deliberations
concern this matter exclusively . If the judgment is " guilty ," the judge
must deliberate about something quite different , namely , what sentence to
impose . The jury ' s deliberations are cognitive , the judge 's practical .! In
this chapter , I shall be concerned with practical deliberation .
Aristotle makes two remarks about deliberation that are familiar to

every philosopher : first , that we deliberate about means not ends; second,
that we deliberate about things that are in our power and can be done
( Nichomachean Ethics 1112a30, 1112bll - 16, 1139b6, 1140a32). I need a
table ; to make a table requires a hammer and saw; I have a hammer but
no saw; so let me go out and buy a saw. Again a doctor , qua doctor , does
not deliberate about whether a patient should be healed, but how to heal
him . No one will disagree with this . To be sure, one can imagine circum -
stances in which a doctor might deliberate over whether a patient should
be healed- say, if there were 1,000 patients and drugs for only 500. But in
this case the doctor isn't really deliberating about ends, but about the
practical problem of how to treat as many people as possible with the

John Macnamara and I talked about many things , over many years, in the area of
practical deliberation and intention . As I write this chapter I imagine his voice ,
objecting , sometimes chiding , but always encouraging , and I dedicate these pages
to the memory of a true friend .
98 McCall

means available . If a doctor were truly to deliberate about ends and ask,
for example , " Is healing worthwhile ?" , the question would fall under
cognitive rather than practical deliberation .
What exactly is deliberation ? In what follows , I try to give a philo -
sophically adequate account of it and to answer some difficult questions .
I finish up by speculating about what sorts of structures and neuro -
physiological functioning in the brain would make possible in real life the
philosophical description of the deliberative process that I have given .
Let 's start off with a concrete example . Marsha has to decide whether
to accept an offer of graduate study in philosophy at UBC , Western
Ontario , or McGill . Call these alternatives A , B , and C . Each one has its

advantages and its disadvantages , and it is important to make the right

decision . Marsha deliberates .

The first step in the deliberative process is to be clear about the alter -
natives , to represent them accurately and keep them in focus . Is the list
exhaustive ? Should she make a last-minute application to McMaster ,
where her sister is studying ? Should she simply do nothing , and not accept
any of the offers? Call this last option D , the do-nothing alternative . Let 's
suppose that A , B , C, and D exhaust the alternatives facing her and that
each one is a " live " alternative in the sense that (a) she can choose it , and
(b) if she chooses it , her choice determines what happens subsequently .
Thus , if she chooses UBC , she goes to UBC . We may imagine that
Marsha has lined up three envelopes on her desk, addressed to UBC ,
Western , and McGill . All she has to do is to put a stamp on one of them
and mail it , or alternatively forget about graduate study . The first re-
quirement for deliberation , then , is the existence of a " choice set," a set of
two or more alternative coursesof action {A , B, C, D , . . .} each of which
it is physically possible for the deliberator to perform or implement , and
whichtogether
exhaust
theavailable
options
.
Once the choice set has been established , the process of deliberation
begins in earnest. Each option has its advantages, which constitute the
reasons for choosing it , and its disadvantages , which constitute the reasons
for not choosing it . I shall call these deliberation reasons. For example ,
UBC has made a generous scholarship offer , but on the other hand has no
staff member who works directly in Marsha 's area of interest . These facts
constitute positive and negative deliberation reasons for A , and there will
be other positive and negative deliberation reasons for B, C, and D . In
deliberation , we weigh deliberation reasons. We compare their relative
strength with the aim of arriving at an overall comparative evaluation .
Deliberation Reasonsand Explanation Reasons 99

The processof evaluation is normally but not invariably the most time-
consuming part of deliberation and ideally should result in a list of the
options ordered by preference.
Once evaluation is completed, it might seemthat the deliberative pro-
cessis at an end. But this is not so. There is one more step, frequently
ignored in studies of rational choice and decision but still essential: the
elementof choice or decisionitself. It may seemdifficult to imagine, once
an ordered evaluation of the options has been made, what more a delib-
erator could need. But supposethe first two alternatives are very close
in the ordering? Supposethey come out equal? What if the deliberator is
faced with a difficult decision? Even in the casewhere an evaluation is
unambiguous, with a clear-cut winner, something elsebesidesevaluation
is neededfor action. The missing element is what Aristotle calls prohai-
resis, deliberative choice.
More needsto be said about prohairesis, but I should first sum up what
has been established so far. The philosophical account of deliberation
given above has distinguishedthree separatecomponentsof the delibera-
tive process, ordered in strict temporal sequence
:
1. Representationof the alternatives
2. Evaluation
3. Choice
With decision, which is choice of one of the alternatives, the deliberative
processends. I turn now to decisionsand the reasonsfor them, the latter
being distinct from the deliberation reasonsfor the different options.
In deliberation we weigh and assessdeliberation reasons, each being a
reasonfor or against one of the alternatives. When eventually we decide,
and chooseone of the options, there is normally also a reason why that
option was selectedover the others. I shall call the latter an explanation
reason, or, occasionally, a decision reason. In the deliberative process
there are deliberation reasons, which are the reasonsfor or against the
different options, but there are also explanation reasons, which are the
reasons why, after examination of the deliberation reasons, one of
the options is chosen.
A principal objective of this chapter is to be clear about the exact dif-
ferencebetweendeliberation reasonsand explanation reasons. Without a
good understanding of the difference, I don't think we can know what
deliberation is. Therefore, I shall spend some time discussingthe rela-
tionship betweenthem.
 100 McCall

First, is it certain that a sharp line can be drawn betweendeliberation

reasonsand explanation reasons? On the face of it , yes. Deliberation rea-
sons, in the deliberative process, have a relatively long life. Somemay be
presentfrom the start, as when a diner at a restaurant perceivesan excel-
lent reason for choosing a chocolate moussedessertat the very moment
he becomesaware that it is on the menu. Others may emergelater, when
the calorie count of the mousseis compared with that of the fruit salad.
The explanation reason, on the other hand, comesinto existenceonly at
the end of the deliberation, when the choice has been made. As long as
there is not yet a choice or decision, there cannot as a matter of logic be
any explanation of it . Deliberation reasonsare reasonsfor deciding this
way or that; explanation reasonsare reasonsthat or why a certain deci-
sion was made. A typical explanation reason might be of the form " X
choseA rather than B becausein the end X attached more weight to the
deliberation reasonsfor A than to the deliberation reasonsfor B." If it
exists at all, the explanation reason comes into existenceonly when the
decision comes into existence. This isn't the case with deliberation rea-
sons. Hence, the two are quite different.
That being said, we still want to know more about explanation reasons.
Could they perhapsbe associated, or identified, with comparisonsamong
deliberation reasons, or more strictly perhaps with comparativerankings
of alternativesbasedon deliberation reasons? If Marsha for examplewere
to conclude that the program and the secondlanguage opportunities at
McGill were more important than the funding and the friends she has
in Vancouver, would this not be a perfectly good explanation reason for
the decision to go to McGill ? We must proceed carefully here. A reason
for going to McGill , even though it results from a comparative weighing
of all the deliberation reasons, is still a deliberation reason. One can, if
Aristotle is right in maintaining there is such a thing as akrasia, conclude
that overall the reasonsfor doing X outweigh the reasonsfor doing Y,
and still do Y. In Davidson's memorableexample(1980, 4), someonemay
conclude that there is nothing to gain and everything to lose from drink -
ing a can of paint, and still one day drink it . No matter how strong it is,
and how much it dominates all other deliberation reasons, a deliberation
reasonis not an explanation reason. Not until the decisionhas beentaken,
and the choice made, does the explanation reasoncome into existence.
Second, must every decision have an explanation reason? Or are there
sometotally irrational decisions, for which there is no explanation at all?
Is the difference between a decision that is made lightly , casually, or
Deliberation Reasons and Explanation Reasons 101

thoughtlessly , and a decision that is made carefully and deliberately , the

difference between a decision that lacks , and a decision that has , an ex -

planation reason? These questions are not easy to answer. Let us start
with Buridan -type situations , in which a choice must or should be made ,
but in which there is absolutely no reason for choosing one thing rather
than another .
Buridan 's ass starved to death halfway between two piles of hay because
there was nothing to incline his choice toward A rather than B .2 Here the
deliberation reasons for A and B are equally balanced , and as a result
no decision is taken . A similar problem , based on fear rather than desire,
is the " railroad dilemma ." 3 What fascinates us in these examples is not
just the specter of decisional paralysis , but the feeling that the outcome -
death by indecision - is in a genuine sense an affront to reason. If the ass
had been rational , or more rational than he was , he would not have
starved . He would have drawn straws and said, " Long left , short right ."
Failing this , if he had been clever enough , he could simply have made
an arbitrary or criterionless choice (Ofstad 1961; McAdam 1965). If you
believe this is impossible , reflect on how you manage to choose one of
a hundred identical tins of tomato soup in a supermarket . The example of
the soup tins was a favorite of Macnamara ' s, and he used it with great
effect in discussion . If you made the mistake of saying that an arbitrary
choice in these circumstances was difficult or impossible , that would in -
dicate that you , like Buridan 's ass, were not very intelligent .
Moving from Buridan -type examples to cases where there is a signifi -
cant difference between the alternatives , but where the deliberation rea -

sons are still equally balanced , the same question arises: can a choice be
made for which there is no explanation or decision reason? If Marsha
finds it difficult or impossible to decide between UBC and McGill because
between them there is " nothing to choose," can she use some tie -breaking
mechanism like a coin and in so doing make a choice that lacks an
explanation reason? Not really . Although in this case the explanation is
different from what it would have been if the evaluation process had

produced a winner , an explanation reason still exists. If Marsha is asked

why she chose UBC rather than McGill , she may answer that she flipped
a coin . The reason why she chose UBC is that the coin fell heads. There
need be nothing irrational about this . Compared to the paralysis of
the Buridan example , and bearing in mind the approaching deadline for
mailing the letter of acceptance, it is a supremely rational procedure .
Sometimes we just have to stop deliberating and decide. It 's a sign of
rationality that we're able to do this .
 102 McCall

Although the conclusion toward which we're moving seemsto be that

decisionsalways have explanation reasons, even if the reasonin question
may be something like " Becausethe coin fell heads," there still remain
other casesto be examined. All our exampleshave been decisions that
were made carefully, thoughtfully , deliberately. Even if a coin had to be
used to break ties, its use was deliberate and rational. But what of deci-
sions that are taken carelessly, without adequatethought? What if we are
inebriated, and our deliberative powersimpaired? Suchcasesmay provide
examplesof decisionswithout explanation reasons.
The making of a decision, like the activity of evaluation, is a rational
process. In fact, deliberation as a whole is a rational process, and each of
the components that make it up is subject to norms by which it can be
judged.
First, in the formation of the choice set, the alternatives should be
realizable, and each option should be such that the decision of the delib-
erator alone is sufficient to implement it , or to initiate its implementation.
The alternatives should also be exhaustive: no important or significant
alternative should be excludedfrom the choice set.
Second, for the process of evaluation, different norms apply. Have
all the deliberation reasonsfor all the alternatives been recognized and
collected? Has each deliberation reason been appropriately weighted? If
there are several deliberation reasonsfor the same alternative, has the
total weighting of all of them, including possible negative reasons, been
reasonably assessed ? Finally , given appropriate total weighting for each
alternative, has the comparison betweenthem been carried out honestly
and without bias? Or if there is bias, is the deliberator consciousof it ,
rather than having it work behind the scenesand affect the evaluation
without the deliberator's being aware of it? These are the questions to
be considered if an evaluation is to be judged rational. It is true that
frequently we lack sufficient insight and self-knowledgeto be surewhether
in a particular case we are giving due and appropriate weight to a set
of deliberation reasons, particularly if the decision is one that touches
us emotionally. In many cases, the degreeof objectivity and detachment
required may be beyond us. Nevertheless, if the deliberative processis
to be a rational one, a certain level of objectivity and self-knowledge is
needed.
Finally , in addition to the above requirements, there are norms for
rational choice. The first requirement is that to be consideredrational,
a choice or decision must have an explanation reason. If , for example, a
Deliberation Reasonsand Explanation Reasons 103

choiceis made for no reasonat all, not evenwith the aid of a randomizing
devicelike a coin as part of the decision-making process, then the decision
cannot be considered rational. Furthermore, even in cases where an
explanation reasonis provided by the deliberator, norms of rationality still
apply to it , and the rational assessmentof a decisionmust take them into
account. For example, if Marsha were to explain her decision to go to
UBC by saying, " BecauseI 'm a Libra ," the relevanceand consequently
the rationality of this supposedexplanation could be questioned. Deci-
sion, therefore, like choice set foffilation and evaluation, is a procedure
that it is appropriate to subject to rational norms, and to judge rational
only if it satisfiesthesenOffilS. The sameapplies, mutatis mutandis, to the
whole deliberative process.
An important question that should be addressedis whether an expla-
nation reason for a decision is a cause of that decision. The immediate
answerwould seemto be no. An explanation reasonfor a decisioncomes
into existenceonly once the decision is made, whereas a cause should
antedatethe thing it causes. The whole question of causality and its role
in the deliberative processis neverthelessof great interest and must be
looked at carefully.
First, our discussionfrom the start has taken us outside the framework
of causaldeterminism, sincethe very existenceof a choice set with two or
more physically realizable alternativesis not consistentwith determinism.
One might attempt to remove the inconsistencyby requiring that in any
choice set one and only one alternative was causally possible, the delib-
erator necessarilybeing ignorant of which one that was. As long as the
deliberator believed (falsely) that each alternative in the choice set was
realizable, deliberation could proceedasusual, the eventualdecisionbeing
taken in favor of the one option that was in fact open. This is deliberation
according to the script that would be written for it by a determinist. The
rewriting does indeed remove the inconsistency, but at the price of (a)
abandoning choice setswith two or more real alternatives and (b) draw-
ing a veil of ignorance over the eyes of deliberators about which is the
sole choosableoption. Behind the veil, " deliberation" can proceed. But
rewriting the script in this way is unnecessary, if it can be shown that it is
simpler and more elegantto abandon determinism.
Since the 1920s, quantum mechanicshas provided the example of a
sciencethat is probabilistic rather than deterministic. As a result, replac-
ing deterministic models of deliberation by indeterministic ones is not
such a daunting or unthinkable project today as it would have been for
 104 McCall

Hume, or Mill , or Brentano, or any philosopher working within the para-

digm of Newtonian science. This is not the place to speculateexactly
how or in what respectthe overall neuronal functioning of the brain could
be regarded as probabilistic rather than strictly deterministic, but at least
the hypothesis that the brain is a complex indeterministic mechanism
makes sense, is testable, and may one day be confirmed or falsified.
What is important for present purposesis the fact that if the central
nervous system functions probabilistically rather than deterministically,
then this functioning permits the formation of choice setsin deliberation.
At the neurophysiological level, a choice set requires different physically
possible neural statesn(A ), n(B), n(C), . . . corresponding to the different
actions A , B, C, . . . , the states being such that if n(A ) obtains then A is
performed, if n(B) obtains B is performed, and so on. Just as each of
A , B, C, . . . excludesthe others, so each of n(A ), n(B), n(C), . . . excludes
the others. These states are " missing" states of the brain that neuro-
scientistsshould one day be able to specify and describe, but that during
deliberation are not actual existents. What is essential, however, for the
existenceof a choice set is that throughout the deliberative processeach
one of the neural statesn(A), n(B), n(C), . . . should be physically possible,
that is, capable of becoming actual at the end of the process. Since each
state is incompatible with the others, the only way this could be the case
would be if the brain functioned indeterministically. Indeterministic neural
functioning consequentlyunderpins choice set formation in the sensethat
it makesit possible, that it is a necessarycondition for its existence. If we
believe in choice sets, with each option separatelyrealizable, I don't see
how we can avoid regarding the operations of the brain as in someyet-to-
be-discoveredway indeterministic.
The next stepis to considerprobabilities. If each of the outcome neural
statesn(A), n(B), . . . is physically possible, it will have a probability value,
and the probability values of all the members of a given choice set will
sum to one. Needlessto say, in keeping with the idea of the brain as a
complex indeterministic mechanism, we are talking of objective proba-
bility values, not epistemic ones. With the introduction of probability
comes the possibility of saying something about causality and perhaps
ultimately being able to conclude whether deliberation can be considered
a causalprocess.
At the beginning of the deliberative process, once the choice set is
formed, the existenceof the underpinning neurological mechanism, with
the different target neural states, gives each alternative of the choice set a
Deliberation Reasonsand Explanation Reasons 105

precise probability value . An interesting and important question is

whether these probabilities change during deliberation .
If the probabilities change during the deliberative process, consid -
erations of probabilistic causality may apply. For example, if J is some
factor that enters consciousness during deliberation - a thought , a desire,
a fact , a memory , a goal - and if A is one of the options of the deliber -
ator ' s choice set , then if

p (AIJ ) > p (Alnot -J ),

we may say that J is a ( prima facie) probabilistic cause of A (Suppes 1970,
12; Fetzer 1988, 118; Cartwright 1989, 55; Eells 1991, 56; DeVito 1997).4
This is not equivalent to saying that J causes the alternative A to be
chosen, since together with the above inequality we may also have

p (BIJ) > p (Blnot-J ),

and J could not be the cause of A 's being chosen and also B 's being
chosen. Hence , probabilistic causality is of limited interest in connection
with deliberation .

Again , the factor J may increase the probability of option A 's being
realized , but can it increase it to unity ? Can any deliberation reason, or
deliberative factor , be of such strength that it makes it physically or
causally impossible for any alternative other than A to be selected? Might
it , before the choice is made, reduce the choice set to a single choosable
option ? This takes us back once more to determinism . One cannot rule
out the possibility that during deliberation something should occur (e.g.,
paralysis or a violent fit ) that restricts the alternatives to one option only .
But such cases are so far removed from the norm as not to qualify as
instances of deliberation at all . At the end of a deliberation , we are left
either with the choice set with which we began, or with a reduced choice
set, or an enlarged one, but in any case a set consisting of at least two
members . The probabilities of these options can change, but if any of
them reaches the value one , deliberation ceases .

A final question concerning probabilities , which reintroduces the sub-

ject of explanation or decision reasons, is this . At the end of the deliber -
ative process , before a decision is made , the deliberator is faced with a
choice set of different alternatives A , B , C , . . . , each with a corresponding
neural state n(A), n(B), n(C), . . . . The objective probabilities of realizing
these different options , and hence of the different target neural states'
becoming actualized , will not in normal circumstances be known to the
 106 McCall

deliberator. Each probability will in fact be nonzero and some may be

exceedinglysmall. Is it possible for a deliberator to choose, deliberately
and intentionally, an alternative of very low objective probability ?
From the point of view of probabilistic science, the occurrenceof an
event of low probability is not impossible, but would not be expected. If it
happened, the only explanation that could be given would be something
like " By the laws of chance, improbable eventsdo occasionallyoccur. By
pure chance, a bridge player may be dealt a hand of 13 spades." There
exists no better causal/ probabilistic explanation than this of why an
improbable event occurs. But this isn't true of a choice made by a deliber-
ator. No matter how objectively improbable the chosenoption may be, a
perfectly good explanation reason for the choice may be forthcoming.
The deliberator who has for years drunk nothing but whiskey may say
that this time he chose a Tia Maria , no matter how objectively unlikely
his choice may have been, becausehe felt like it .
The lessonhere, I think , is that in the realm of deliberation and choice
there may occur eventsfor which there is no causal/ probabilistic expla-
nation, but for which there is an intentional explanation. Someintentional
explanationsmay be of the form " Ivan chosethe Tia Maria becausehe felt
like it " ; others may be of the form " Unbelievably, Sally isn't here today
becauseshedeliberately and intentionally choseto go to jail ." Perhapsthe
clearest examplesof intentional explanations are those given in circum-
stanceswhere nonintentional explanations(i.e., causal/ probabilistic ones)
don't exist. But detailed examination of the differencebetweenintentional
and causal/ probabilistic explanation is a lengthy matter and must await
another occasion.
Summing up, I began by listing the three essentialcomponents of a
deliberation. Theseare, in strict temporal order, (1) choice set formation,
(2) evaluation, (3) decision. In evaluation the reasonsfor and against the
different options (deliberation reasons) are both appropriately weighted
and weighed by the deliberator. Eventually, one option is chosen. Each
elementof the deliberative processis subjectto rational norms by which it
may bejudged, a common but nonmandatory norm of element(3), choice
or decision, being the existenceof an explanation reason for the choice
once it is made. If choice set formation is underpinned by an indeter-
ministic neural mechanism, with a specificneural staten (Ai ) and a specific
probability corresponding to each alternative Ai of the set, then the
explanation of why the deliberator choosesone of the alternativeswill not
in general be a causal/ probabilistic explanation based on those proba-
Deliberation Reasons and Explanati on. Reasons 107

bilities. Instead, it will be an intentional explanation. The precise rela-

tionship between intentional and causal/ probabilistic explanations in
deliberation remains to be investigated.

Notes

1 . The distinction between practical and cognitive deliberation is discussed in

McCall 1987 and McCall 1994 , 254 .

2 . The dilemma antedates Buridan . See Rescher 1959 for why the problem of
" Buridan ' s ass " should more justly be known as " al - Farabi ' s dates ." Buridan 's

problem is discussed in Bratman 1987 .

3 . You are hiking with your partner through mountainous wooded country and
come to a horseshoe - shaped valley with a railway track at the bottom . While you
are crossing the track , a heavy branch falls on your partner ' s leg , pinning him or
her to the rails ; and while trying vainly to move the branch , you hear the whistle
of an approaching train . Unfortunately , the echoes in the valley make in impos -
sible to tell which direction the train is coming from . Game theory tells us you
have a 50 % chance of saving your partner ' s life if you pick a direction arbitrarily
and run as fast as possible down the track to stop the train , but how many of us
would have the strength of mind to do this as the whistle grows louder and the
uncertainty becomes more agonizing ?

4 . This definition of probabilistic cause is normally supplemented by requiring

that p ( AIJ ) > p ( Alnot - J ) in all partitions of A ' s causal factor - space , meaning in
all circumstances in which combinations of other causal factors of A are held fixed

and only J is allowed to vary . In the case of deliberation , however , we are con -
cerned with a single unique process , not in general repeatable , and so this supple -
mental requirement cannot be met . Nor would it be appropriate to require , in
addition to p ( AIJ ) > p ( Alnot -J ) , that the anticipated probabilistic effect A actu -
ally occurs , since if A occurs then in the probability calculus p ( A ) == 1 , and hence
p ( AIC ) == 1 for any C , provided p ( C ) # O.

References

Bratman , M . ( 1987 ) . Intention , plans , and practical reason . Cambridge , MA :

Harvard University Press .

Cartwright , N . ( 1989 ) . Nature 's capacities and their measurement . Oxford : Clar -
endon Press .

Davidson , D . ( 1980 ) . Essays on actions and events . Oxford : Clarendon Press .

DeVito , S. ( 1997 ) . Probabilistic causality without propensities . Unpublished

manuscript , University of Pittsburgh .

Eells , E . ( 1991 ) . Probabilistic causality . Cambridge : Cambridge University Press .

Fetzer , J . H . ( 1988 ) . Probabilistic metaphysics . In J . Fetzer ( Ed .) , Probability and

causality . Dordrecht : Reidel .
108 McCall

McAdam , J. ( 1965). Choosing flippantly or non -rational choice . Analysis , 25,

132 - 136 .

McCall , S. (1987). Decision. CanadianJournal of Philosophy, 17, 261- 287.

McCall , S. ( 1994). A model of the universe. Oxford : Clarendon Press.
Ofstad , H . ( 1961) . An inquiry into the freedom of decision. Oslo : Norwegian Uni -
versities Press .

Rescher, N . ( 1959). Choice without preference . Kant -Studien, 51, 142- 175.
Suppes, P. ( 1970). A probabilistic theory of causality . Amsterdam : North -Holland .
Chapter
- 6
Truth and Its Negation : David R . Olson
Macnamara ' s Analysis of the
Place of Logic in a Cognitive
Psychology

Logicians, I gather, toil away at pecifying the relations betweensymbols

and the things the symbols represent. Thus, they concern themselveswith
how names can refer to things, how propositions can be true, how the
logical particles and, or, not, and if-thenwork, and generallywith issuesof
reference, truth , and validity . They formulate the results of their inquiry
in terms of logical schematasuch as
p ~ q , ""' q
,

""' p

which reads something like : ifp implies q, and q is not the case, then p is
not the case either . The question is, what is the status of such a logical
schema? Is it , for example , a law of thought ? Is thinking achieved by
applying rules of logic , or is logic merely a description of how thinking
proceeds, or , yet again , is logic a set of " regulative " or noffilative princi -
ples quite independent of the way people ordinarily think ?
The question has a venerable history , as John Macnamara infoffils us
in his daunting book A Border Dispute ( 1986) . The idea that logic con-
stitutes the laws of thinking is known as psychologism and it is well
expressed in J. S. Mill 's System of Logic , in which the author claims that
the necessary truths of logic are simply those things that we find it impos -
sible to doubt . That is, Mill identified necessity with certainty . Frege in his
Begriffsschrift distinguished the two , arguing that certainty is based on the
facts of experience whereas necessity is connected " with the inner nature
of the proposition considered " ( 1879/ 1967, 5) . As Macnamara points out ,

Parts of this chapter were first published as a review of Macnamara 1986 (see
Olson 1987).
 110 Olson

Frege claimed that logic is a " normative " law , a law that states how
thinkers ought to judge , not a description of how they do judge .
Macnamara wants to add a third stage to this history by showing that
logic is not something different and apart from thought but , rather , that
" logic supplies a competence theory for the psychology of human rea-
soning " ( p . 185) . In so doing , he echoes the views of George Boole , his
more famous Irish predecessor, who claimed that the laws of logic " pos-
sess an authority inherent and just , but not always commanding obedi -
ence" ( 1958, 409) .
Macnamara 's line of attack is borrowed from the grammarians rather
than the mathematicians . The work of the logician is somewhat similar to
the work of the grammarian who attempts to spell out the internal rela-
tions within linguistic structures . Since Saussure and especially since
Chomsky , it has been common to think of linguistic structures as con-
stituting the competence of native speakers. Grammars do not just
describe products or sentences; they specify the implicit knowledge of
speakers, knowledge that figures both in the generation of sentences and
in intuitions of grammaticality .
Those developments in linguistics have given us a new respect for the
rich structure of the linguistic system and , consequently , a new respect for
the complexity of structure of the human mind . Rules , representations ,
constituent structure , meaning , propositional structures are all part of
current discourse about mind . There is no longer any talk of tabula rasa.
Why not approach logic this way ? Why not indeed , says Macnamara .
Rather than leave logic to the logicians , he argues that logical competence
is part of the endowment of rational creatures just as linguistic compe-
tence is part of the endowment of verbal ones . And then he sets out to
explicate that competence .
Macnamara spent several of the last few years of his life with logicians
learning about logic and then trying to specify what is logically implied in
children 's learning to do such apparently simple things as refer to objects
by proper names and to make assertions. In the course of this study he
examined the logic of proper names, common names or sortals , kinds ,
existential predicates , truth predicates, and sentential connectives includ -
ing negations . Needless to say, children 's acquisition of these structures is
no small achievement !

Consider Macnamara 's analysis of the child 's learning of a proper

name . To call a particular dog " Freddie " requires , Macnamara points
out , that the child have the notion that this particular dog, Freddie , is a
Truth and Its Negation 111

member of a category , dog , that the child recognize that dog is a kind of
thing , that the particular individual in front of the child is what is being
named , and that there is some equative relation between the thing pointed
at and the proper name . He convincingly shows how much more a child
brings to the learning of a proper name than Saint Augustine or anyone
else had noticed . And he provides interesting evidence that very young
children distinguish names, expressed by proper nouns , from sortals ,
expressed by common nouns , by the presence or absence of the articles a
and the . Remarkable !

But Macnamara goes much further to insist , in addition , that all of

these things must be represented in a language of thought in the form of
a full proposition that the child judges to be true . For those of us who
believe in genuine development - that knowledge is in some sense con-
structed - this goes much too far . Is there some way to escape Macna -
mara 's claim that such elaborate representation is presupposed in the
learning of a proper name? As truth and its negation , falsity , are the fun -
damentallogical notions and as they play the central role in Macnamara ' s
analysis and because truth is one of those axes that it does no harm to
grind , I shall consider it in some detail .
Utterances express propositions that are, in fact , true or false. " To say
of what is that it is , is true ," said Aristotle . Moreover , truth is the crite -
rion for agreement that parents use in talking to their children . If a child
says " Freddie " while pointing at Muffin , they say " No , Muffin ." So
truth and falsity , expressed by the negation , are implicated in talk . But
Macnamara is digging deeper. He is not examining truth and falsity in
discourse. He is examining how truth and its negation function in the
internal language of thought . Specifically , Macnamara claims that the
l -year-old , in saying " Freddie ," is asserting a proposition , " That is
Freddie ," which , fully expressed in the language of thought , has some
form equivalent to " That [which is a member of the kind dog , which
refers to a kind ] is Freddie is true " ( p . 62) . In saying " Freddie ," the child
is not only asserting a proposition that we adults may characterize as
expressing a truth judgment . The child , Macnamara claims , understands
that the expression asserts that particular proposition and in so doing has
judged that proposition to be true . Lest you think I am putting words in
Macnamara ' s mouth (as I am implying he has put them into the child 's),
let us allow Macnamara to speak for himself :

Assertion presupposes a grasp of the notion of truth ; it presupposes a judgment

of truth . . . . The learner does not need the concept of judgment because he can
112 Olson

- --
perform the act of judging without knowing that he is judging. But the learner
must
have
106
-107
) the
concept
of
truth
;there
is
noaction
that
can
stand
initsplace
.(pp
.
So Macnamara claims that the predicate is true (in the language of
thought) must be used in judging the sentencetrue any time the child
names an object. Its use implies the presenceof the concept true and its
opposite, not true, that is, false. " Children are endowedby nature with an
understandingof both" (p. 108). Finally , " the conclusion that forces itself
on us is that neither the conceptsof truth and falsity nor the fundamental
principles of logic are learned" (p. 109).
Thus, Macnamara has concluded that children have concepts of
truth and falsity representedin an innate languageof thought that permits
them to learn, among other things, proper names. But are we compelled
by logical argument to grant that young children not only produce utter-
ancesthat are, in fact, true or false, but also that l -year-old children have
concepts of truth and falsity and know the principle of contradiction?
Macnamara claims that we are, the primary reason being that it seems
impossible to him that " logic-free activities can ever yield logical re-
sources" (p. 30). This, of course, is a denial of the Piagetian hypothesis
that representational structures can develop from sensorimotor ones, a
central plank of Piaget' s epigenetictheory.
Even after severalcareful readings of Macnamara's arguments, I can
see no reason why the predicates true and false could not be acquired
(rather than being presupposed ) . Children do encounter utterances that
are objectively true and others that are objectively false, and they do en-
counter adults' judgments of utterancesas being true or false. Why could
not theseobjective facts be sufficient grounds for children's acquisition of
thesepredicates? I seeno reason that these predicatesare preconditions
for naming any more than that the predicates say and mean are pre-
conditions for saying and meaning things. Admittedly , all predicatespre-
supposesome mechanismfor representingpredicates, but that is not the
issuehere. The issuehere is whether it is a necessarytruth that children
must have predicates for truth and falsity representedin a language of
thought in order for them to learn proper names.
Concepts of truth and falsity (false being not true) do presupposea
conceptof negation, and it is worth asking whether, as Macnamara insists,
the concept of negation is also innate. Negation is fascinating in that it is
clear that negation does not exist in nature. Russell (1948, 520) pointed
out that " the world can be describedwithout the use of the word not."
TruthandIts Negation 113

Geach (1957, 23), too, wrote that " nowhere in the sensibleworld could
you find anything, nor could you draw any picture, that would suitably be
labeled or or not." Yet a mind that could representonly what is the case
would be a mind without logic, hypotheses, or language(Harrison 1972).
Although negations are absent from nature, they are central to lan-
guage. Comparativists have argued that negation is conspicuouslyabsent
from the communication systemsof other animals (Sebeok 1962), and
although nonhuman animals can issuewarnings and register disappoint-
ment, they cannot mark in their communication that someexpectedevent
has not occurred. Visalberghi and Fragaszy (1996, 294) write, " We can
think of no impediment to the use of these forms of negation in non-
humans; but they simply do not use them." Indeed, it is possiblethat an
animal cannot mentally representnegation and that the ability to repre-
sent a cat as " not a dog" is peculiar to language-using humans. Could,
contrary to Macnamara, the concept of negation (and hence falsity) be
acquired as a child acquiresa natural language?
There is a rich literature on children's acquisition of negation in lan-
guage (for a summary, seeHorn 1989) that indicates a clear progression
in understanding and use of negatives beginning with expressionsof
rejection, disappearance , and finally denial, so-called truth-functional
negation. Although rejections, refusals, and prohibitions are not neces-
sarily linguistic- even dogs can reject food or refuse to come- denials
constitute the " linguistic category of negation par excellence" (Tottie
1982, 96). As we shall see, only when children are approximately school-
ageddo negation and falsity come apart in such a way that they canjudge
a sentencecontaining a negation to be true. They come to agree with
Aristotle in the secondpart of the definition mentioned earlier: " To say of
what is not that it is not, is true." Thus, to judge a proposition as true or
false is a late and sophisticatedachievemententirely dependentupon the
acquisition of negation.
Macnamara could say of such a counterargumentthat it confuseslogi-
cal performance with logical competence. It is easy to design tasks that
make it difficult for children to display their underlying logical ability .
The tasks I discussexamine complex judgments about English sentences ,
whereas he discussesthe logical competence that must be present to
account for children's acquisition of names, assertions, and inferences.
So, he may claim, I misunderstand. That, and worse, is what he said
about Piaget. In a comment that is as close as one gets to a call to arms,
 114 Olson

Macnamara says, " It is in fact doubtful that Piaget had any sharable
insights into the fundamental processes of cognitive growth " (p . 114).
Piaget 's subjects, he claims , are " faced with a complicated induction
problem and simply become muddled " (p . 116). Piaget 's tasks, he sug-
gests, test only subjects' ability with an algebra for possible combinations
of propositions (he'd say the same of Wason and 10hnson-Laird 's " four -
card " task), whereas he is interested in only " the most basic logical con-
nectives themselves" ( p . 115). Of those basic notions he says, " There is
nothing in [Piaget 's] evidence to even suggest that [the child ] was so
totally lacking in logical intuitions as to knowingly assent to both a
proposition and its negation " (p . 116) .
However , consider the evidence provided by Alison Dewsbury and later
explored in some detail by Janet Astington and myself (Astington and
Olson 1984; Olson 1997) . Four - and five -year-old children are shown a
picture of a man with a hat , accompanied by a series of sentences. They
are asked to put a check mark by the ones that are true , right , correct , or
OK and to put an X by the ones that are false, wrong , or in error . Here is
what children reliably do :

Themanhasa hat. J
Themanhasno hat. J
As if that was not enough , they go on thus . Shown a picture of a cat, and
asked to mark the following sentences, they do thus :
This is a dog . X
This is not a dog . X

Such judgments would seem to contradict Macnamara 's claim directly ;

children do seem to knowingly assent to both a proposition and its nega-
tion . In fact , the explanation is not that simple . I have argued that rather
than judging truth or falsity as requested, children are either agreeing or
disagreeing with the assertion of a speaker and not judgments of the truth
or falsity of the proposition . Thus , in response to " The man has no hat ,"
children are using a check mark to express " Yes, he does [have a hat ],"
and in response to " This is not a dog," they are using an X to express
" No , it 's not [a dog] ." Thus , both supposedly wrong responses are in fact
appropriate . Only after children are of school age (and perhaps even then
only in a literate society) do they master the more specialized predicates ,
true and false . Judging a proposition as true or false, far from being a
presupposition for naming and saying things , is a consequence of treating
TruthandItsNegation 115

utterances in a certain way - namely , as propositional objects that exist

independently of the sentences that express them and the situations they
represent and, hence, as objects that can be judged as true or false of those
situations . Admittedly , children are sensitive to the fact that something
has gone wrong if they disagree, but the nature of this " going wrong "
may be pragmatic rather than truth -functional (Volterra and Antinucci
1979).
Rather than treating logic as a competence theory for thought , I would
prefer to see logic as a metalanguage for expressing certain properties of
statements in situations . As Macnamara acknowledges , logic is a device
for characterizing arguments , not a device for making them . However ,
there is something deeply counterintuitive in arguing that the predicates
represented in the metalanguage are presupposed in learning the language
in the first place. As mentioned , Macnamara borrows this assumption
from grammatical theory , where the abstract properties of a grammar are
attributed to a learner as an explanation of what makes the learning of
language possible .
Extending this approach to logic leads Macnamara to claim that calling
one' s dog " Freddie " is an exemplar of the class of assigning proper
names. Proper names have a number of logical properties including that
they assert the truth of a proposition " This particular individual dog is
Freddie ." Macnamara next argues that if this abstract characterization of
the logical structure of proper names is correct , then this logical structure
must be known by learners if they are to learn a proper name . Indeed ,
Macnamara claims , it is presupposed in the acquisition of a proper name .
What is required , he argues, is that children possess an unlearned lan -
guage of thought that has all the logical resources that the natural lan -
guage will come to have when it is acquired . It is seen as the unlearned
competence that makes the learning of logical particles in natural lan -
guage possible . Is such attribution inescapable?
In my judgment , the argument is analogous to the following . When
children learn the word fish , they learn a word , a one-syllable word , a
four -letter word , and a number of other things . These are all properties of
the stimulus that can be characterized by means of a metalanguage . Why
couldn 't the same be true of logic ? Children learn to make utterances in
contexts . Utterances constitute new objects in the world , objets trouves. A
metalanguage is evolved to talk about the properties of those newly cre-
ated objects, propositions and truth being the most important of these.
Truth is, in this view , an emergent property of a created object . Concepts
 116 Olson

of proposition and of truth are not presupposedin learning a language;

they do not characterize a language of thought that makes learning of
languagepossible. Rather, they are the properties of symbolic objectsthat
the metalanguagebrings into reflectiveconsciousness(to usea term that is
not to be found in Macnamara's index).
Now there is a senseof competencethat this metalanguagereflects. The
metalanguage, like the language, is a human invention, and human
inventions do reflect, however indirectly, the structures of the human
mind. Thus, as the late Robin Gandy, a friend of Alan Turing , claimed,
exploration of mathematical structuresis essentiallyan exploration of the
properties of the human mind. That notion of competence is more
appealingto me than Macnamara's notion of competence, which seesany
acquired logical or linguistic structure as an explicit representationof the
structures already present in a language of thought. Needlessto say, it
was not more appealing to Macnamara.
In developing his own views, Macnamara does justice to a complex
body of knowledge and at the sametime engagesthe reader in a complex
but important debate. Just as languageand thought came to be seenin a
new light when psychologistsfinally began to pay some attention to lin-
guistic structure, so thinking and reasoningmay come to be seenin a new
light when psychologistsbegin to pay attention to the issuesMacnamara
set out. There can be no doubt that Macnamara is a Moses leading a
reluctant peopleout of bondageto an oversimplifiedview of mind. I would
be much happier if I could be surethat this wildernesswas, in fact, a route
to the promised land.

References

Astington , J . W . , and Olson , D . R . ( 1984 , May ) . Children ' s comprehension of true

negation . Paper presented at the University of Waterloo Conference on Child

Development . Waterloo , Canada .

Boole , G . ( 1958 ) . An investigation of the laws of thought . New York : Dover .

Frege , G . ( 1879 / 1967 ) . Begriffsschrift . In J . v . Heijenoort ( Ed . ) , Frege and Giidel :

Two fundamental texts in mathematical logic . Cambridge , MA : Harvard Univer -

sity Press .

Geach, P. T. (1957
). Mental acts , their content and their objects . London : Rout -

ledge& KeganPaul.
Harrison, B. (1972). Meaning and structure: An essay in the philosophy of lan -

guage. New York : Harper & Row.

Horn
.,L.R.(1989
Press ).Anatural
history
ofnegation
.Chicago
:University
ofChicago
Truth and Its Negation 117

Macnamara , J. ( 1986) . A border dispute: The place of logic in psychologiv. Cam -

bridge , MA : MIT Press.
Olson , D . R . ( 1987). Thinking about logic ( Review of J. Macnamara , A border
dispute). Canadian Journal of Psychology , 41, 392- 398.
Olson , D . R . ( 1997). The written representation of negation . Pragmatics & Cog-
nition , 5(2), 239- 256.
Russell, B. (1948). Human knowledge, its scopeand limits. New York : Simon &
Schuster.
Sebeok, T . A . (1962). Coding in the evolution of signaling behavior. Behavioral
Science
, 7, 430- 432.
Tottie, G. (1982). Where do negativescome from? Studia Linguistica, 34, 101-
123.
Visalberghi, E., and Fragaszy, D . (1996). Pedagogyand imitation in monkeys. In
D . R. Olson and N . Torrance (Eds.), Handbook of educationand humandevelop-
ment. Oxford: Blackwell.
Volterra, V ., and Antinucci, F. (1979). Negation in child language: A pragmatic
study. In E. Ochs and B. Schieffelin(Eds.), Developmentalpragmatics . New York :
Academic Press.
Chapter - 7

Names for Things and Stuff: SandeepPrasada

An Aristotelian Perspective

Freddie is a dog . As an undergraduate in one of John Macnamara 's

courses , I spent the better part of a semester studying what knowledge

must be attributed to someone who understands that sentence . I intend to

borrow a page from John and will spend much of this chapter discussing

what knowledge must be attributed to the person who understands the

sentences ( 1 ) - ( 4 ) . In doing so , I will be addressing an aspect of the ques -

tion posed by Macnamara ( 1978 ) : How do we talk about what we see ?

Given that we can use any of the sentences ( 1 ) - ( 4 ) to talk about the same

entity in the world , we need some kind of account of how it is that these

sentences differ in how they are related to our visual representations of the

entity . As we will see , the simple fact that we can use any of these sen -

tences to talk about the same entity in the world , if taken seriously , leads

to interesting conclusions concerning the nature of our perceptual and

conceptual systems . Recent work by Landau and J ackendoff ( 1992 ) has

addressed the question of how we talk about objects and locations . In this

chapter , I will be focusing on what is involved in talking about objects

and materials .

It is impossible for me to adequately express the intellectual debt I owe John

Macnamara . In addition to the vast amount of knowledge he imparted , he has ,

through example , shown me the virtues of taking problems in cognitive psychol -

ogy seriously and trying to make progress on the problems by using a wide array
of methods and insights from different disciplines . I am grateful to have had the
opportunity to have him as a teacher , a colleague , and a friend . I thank John Kim
for helping me clarify my thinking on all of the issues discussed in this chapter ; he
is , however , in no way responsible for any confusion or mistakes that remain .
Finally , I ' d like to thank Paul Bloom and Ray Jackendoff for helpful comments
on an earlier version of this chapter . Preparation of this chapter was supported
in part by National Science Foundation grant SBR - 9618712 .
 120 Prasada

(1) That is a table.

(2) That is wood.
(3) That is a piece of furniture .

(4) I have no idea what that is, but it is brown, hard, and weighs 10
pounds.
In each of the sentences(1)- (4), the demonstrative that directs our atten-
tion to a particular part of our environment. Typically , this is a spatio-
temporally bounded solid entity, though this doesnot have to be the case
(e.g., That is a shadow, That is coffee, That is a forest, That is a somer-
sault). Clearly, there are principles of perception that operate to organize
the information available in ways such that particular spatiotemporally
bounded entities are segregatedfrom others. These entities serve as the
natural objects of our attention, action, and understanding. Exactly what
types of information are used and how they are used to define theseenti-
ties, which are the natural objects of our attention, is what a theory of
perception should give us.
Crucially, that which is being attended to can be distinguished from
that which is not being attendedto in purely spatiotemporal terms. This is
what allows for the possibility of a given entity (describedin purely spa-
tiotemporal terms) being understood (in nonspatiotemporal terms) in
more than one way by a given person (e.g., as a table or as somewood),
or in different ways by different people. The fact that we can disagree
about what somethingis, or think about it in more than one way, requires
that we be able to attend to the thing in a manner that is independentof
our understandingof that thing. That is, the perception of an entity has to
be autonomous of its conception (Macnamara and Reyes 1994a). Note
that this claim doesnot rule out the possibility of top-down influencesin
perception; rather, it claims only that it is possibleto perceiveentities in a
manner that is independent of any particular way of understanding the
nature of the thing- how often this happens, or how easythis is to do, is
another matter. In fact, sentencessuch as (4), which are noncommittal
with respectto the nature of an entity, are rare except perhapsin archae-
ological contexts. The claim of autonomy also does not imply that the
perceptualproperties of an entity are not related to how it is conceived; it
only means that there are principles of conception that are distinct from ,
and are not solely determinedby, principles of perception and! or percep-
tual properties.
Names for Things and Stuff 121

Given that we can perceive an entity without having committed our-

selvesto what it is, it is necessaryto provide answersto the following
questions: (1) what is the nature of the conceptual representationsthat
underlie different construalsof the entity (e.g., table/wood)?, and (2) what
are the factors that determinewhich construal is made? The answerto the
first question should have implications for the answer to the secondone.
In approaching theseissues, I have taken to heart Macnamara's 0bserva-
tion that whereasmodem-day physicistsand biologists may safely ignore
Aristotle 's writings, modem-day psychologistshave much to gain from
Aristotle . In the next section, I will briefly describeAristotle 's account of
the nature of things and stuff because, although it is lacking as a scientific
account, I think it accurately capturesour natural ways of understanding
and talking about the world and is implicit in our conceptual structures.
Furthermore, such an account will help us formulate answers to the
questionsraised above. The account of Aristotle 's thought on the nature
of objects, essence , and the four causesis based largely, though not
exclusively, on the interpretations of Lear (1988), Scaltsas(1994a,b), Witt
(1989), Bolton (1976), Cohen (1996), and Moravcsik (1975).

7.1 An Aristotelian Account of the Nature of Things

I and Stuff

According to Aristotle , an appropriate answer to the question " What is

it?" requires citing the essenceof the entity in question or the causeof its
being what it is. Furthermore, he proposes that there are four ways in
which we cite cause. The four types of causeswe cite are the material, the
formal , the final , and the efficient. It is important to note, however, that
even though theseare often presentedas four distinct causes, there really
are only two basic kinds of causes, the material and the foffi1al, the final
and efficient causesbeing aspectsof the formal cause. Aristotle felt that
we can adequatelycharacterizethe essenceor nature of both things and
stuff in terms of thesecauses.
Let us unpack theseclaims. Aristotle called things like dogs and tables
(what we would call basic-level objects) substancesand things like wood
and sand matter. Becausewe now use substanceand material synony-
mously, much confusion arises. To try to minimize confusion, I will use
basic-level object, object, and thing to refer to what Aristotle would have
called substancesand matter, material, and stuff to refer to what Aristotle
would have called matter. It is important to remember, however, that in
calling entities objects or things I do not mean simply spatiotemporally
122 Prasada

bounded solids, as thesecould be understood either as being an object of

some kind or as being a piece of matter of some kind . Aristotle claims
that basic-level objects are a composite of matter and form . What he
meansby this is that certain aspectsof an object's being are causedby its
matter, whereas others are caused by its form. For example, certain
aspectsof a table's being, suchas its being hard or burnable, are causedby
its matter. This is why it is appropriate to cite the material of which the
table is made when we are seekingto explain why it is hard and why it
bums. The properties of being hard and burnable are not causedby the
structure of the table, its function, the processby which it was created, its
color, its size, or any property other than the material of which it is made.
Thus, there is a subsetof properties of objects, which includes properties
suchas hardnessand burnability , that are causedby, and thus explainable
by referenceto, the kind of matter of which the object is constituted.
Other aspectsof an object's being are causedby its form . For example,
a ball's ability to roll , a dog's ability to walk, a bird's ability to fly, or a
car's ability to transport us are all causedby, and are explainablein terms
of, the structure of these objects. Aristotle noted, however, that the
structure of an object does not arise spontaneously, nor could it , because
having a particular structure is not a property of matter. In fact, matter is
indeterminate with respectto structure. Thus, the structure itself must be
causedby and explainable in terms of a processof growth or construc-
tion, what Aristotle called the principle of growth and maintenance(or
sometimesthe principle of changeand rest). Furthermore, this process,
which causeschangesin structure, doesnot simply changethe structure of
the matter in random ways but is directed at producing a particular
structure (or a particular range of structures). Thus, the principle of
growth and maintenance cannot be defined independently of the struc-
tures that it is aimed at producing, and the presenceof a particular
structure in matter cannot be explained without referenceto a processof
growth or creation.
Finally , we are led to ask why the process of growth or creation is
directed at producing the structuresit does, and not others. Once again,
the explanation for this cannot be found in the material constitution of
the object, becausematter, in and of itself, lacks both determinate form
and any principle of growth or creation. However, we might expect that
some aspect of the form may provide the relevant explanation. This is
what Aristotle proposes. For example, he would argue that if we are to
explain why a carpenterengagesin the activity of building that culminates
Names for Things and Stuff 123

in the creation of this particular structure and not some other structure ,
we must cite certain functional properties of the created structure as being
the reason why the carpenter created this structure and not some other
structure .! In the case of natural kinds such as plants or animals , the fact
that they have the particular structures they have can be explained by
citing the fact that these structures allow the organisms to function in var -
ious ways that allow them to grow , maintain , and reproduce themselves.
To summarize , the principle of growth and maintenance (often called
the efficient cause) is responsible for changes in the structure of matter
and is the cause of a determinate structure . This structure (often called the
formal cause), in turn , is the cause of certain functional properties ' being
present. These functional properties (often called the final cause), in turn ,
are the cause of the existence or continued existence of the process of
growth and maintenance . Thus , there is a single formal cause that has
three aspects or components , rather than three independent causes. This
cause is called the substantial form or sometimes just form orformal cause;
but it should be understood that this term does not refer to shape but to
the complex cause discussed above .
The introduction of the formal cause into matter causes it to cease
being matter and become the matter of the object that results. This is not
simply a terminological distinction , but a metaphysical distinction . It
involves a change from something that is essentially matter to something
that is essentially an object . Recall that matter lacks any determinate
structure , and any structure that is present in an amount of matter is
accidental and is not due to its material nature . It follows that any unity
that is found in an amount of matter (e.g., a piece of wood ) is also acci-
dental and not due to its material nature . The unity is merely spatio -
temporal and is not due to its material essence. This is the reason why the
matter would survive as the same matter even if it lost its spatiotemporal
unity (e.g., by being broken into separate pieces). The introduction of the
formal cause into matter changes all this . The structure of the matter
found in an object is not accidental but is due to the substantial form .
Furthermore , the unity found in the matter of an object is not accidental
or merely spatiotemporal but is due to its substantial nature . It is essen-
tially a unitary object . The unity is due to the fact that it is the product of
a process directed at creating a structure of a particular kind that is
essential for its functioning in ways that are responsible for the existence
and/ or continued existence of the process and the object itself .
124 Prasada

We are now in a position to provide an account of what is involved in

offering (I ) and (2) as responsesto the question " What is it?" In
responding with (2), we are saying that the entity in question is (essen-
tially ) wood. In giving this answer, we are citing only the material cause.
That is, we are claiming that various aspectsof its being, such as its being
hard, brown, and burnable, are causedby, and thus explainable in terms
of, the kind of material it is. The nature of the material doesnot, however,
account for the structure of the entity. Therefore, we are also claiming
that the particular structure the entity has is not the result of any process
that was directed at creating that structure (becausein this answer we
do not cite the presenceof the formal cause); instead, the structure is the
result of a processthat was not directed at creating that structure (e.g.,
erosion, breaking, tearing). It follows also that any functional properties
of the entity that depend on its structure must be consideredaccidental
(i.e., they are not caused by the entity's material nature). "Finally , in
stating that the entity is simply wood, we are claiming that the unity that
is present in the entity is accidental and not due to what it is essentially
(in this case, wood).2
Conversely, in respondingto the question " What is it?" with (I ), we are
saying that the entity is (essentially) a table. This means that it is essen-
tially a composite of matter and form, which in turn means that the
structure of the entity is the result of a processthat is directed at creating
that structure. Furthermore, we are claiming that at least some of the
functional properties that are dependent on the structure are not acci-
dental, but are the reason why the structure was created in the first place
and will govern modifications and repairs that may be made (seeBloom
1996a).3 Finally , the unity that is found is not merely accidental or spa-
tiotemporal but is the result of the directed processthat createsthe struc-
ture that is essentialto the entity's being what it is.
I submit that theseAristotelian responsesto the question " What is it?"
are implicit in our conception of entities as kinds of objects and stuff. In
the next section, I presentsomeevidencefor this claim.

7.2 Some for the Aristotelian Natureof Our Conception

of
Things

A spatiotemporallycontiguousamountof matter is distinguishedfrom

an objectby the presenceof the formal causein the object. Given this
understandingof the natureof thingsand stuff, we predictthat subjects
Names
forThings
andStuff 125

should be more likely to construe a solid entity as an object if they are

presented with evidence that the entity has a definite foffi1 that is the
product of a processdirected at creating that structure and that it pos-
sessesfunctional properties that dependon that structure than if they are
led to believe that the structure is the product of an undirected process
and are not presentedwith any evidenceof structure-dependentfunctions
of the entity. Furthermore, entities construedas objectsshould be referred
to through the use of count nouns (as in (1)) becausethe unity of the
entity is essentialto its construal as an object, whereasentities that are
construedas amounts of material should be referred to through the use of
massnouns (as in (2)) becausethe unity presentin an amount of matter is
not essentialto its construal as some stuff. Data from a recent seriesof
experimentsby Prasada, Ferenz, and Haskell (in preparation) bear out
thesepredictions.
In a first experiment, subjectswere presentedwith either " regularly"
shaped entities or " irregularly" shaped entities and asked to choose
a count noun description (There is a blicket in the tray) or a mass noun
description (There is a piece of blicket in the tray or There is blicket in
the tray) of the entities. Not surprisingly, subjectschose the count noun
description significantly more often when describing regularly shaped
entities than when describing irregularly shapedentities.4 Crucially, the
regularly shapedentities had beenjudged, in a pilot experiment, to have
structuresthat were likely to be the product of a processthat was directed
at creating that particular structure, whereasthe irregularly shapeditems
were rated as having structuresthat were unlikely to be the product of a
processthat was directed at creating that particular structure. Although
thesedata are consistentwith the prediction that entities that are seenas
having structures that are the result of a directed processshould be con-
strued as objects and spoken about through the use of count nouns more
often than entities that are seenas having structuresthat are the result of
an undirectedprocess, it is possiblethat someother differencebetweenthe
regularly and irregularly shapedentities was responsiblefor the results. A
third condition in the experimentruled out this possibility.
In the third condition, subjects were initially shown three identical
instancesof an irregularly shapedentity. Then one of the instanceswas
moved into a tray of its own and, as in the previous conditions, the sub-
jects were askedto chooseeither a count or massnoun description of just
a single instance. The irregularly shaped entities used in this condition
were the same as in the previous single-instance condition, and the con-
 126 Prasada

ditions under which subjectschose the description were identical to the

conditions in the single-instancecondition (i.e., with a single entity in the
tray). The only difference between this condition and the previous one
was that before subjectswere asked to pick a description for the target
entity, they saw it in another tray with two other identical entities. Sub-
jects in this condition chosethe count noun description significantly more
often than the subjects who had not initially seen multiple identical
instances of the entity. Furthermore, subjects in the multiple-identical-
instancescondition rated the structure of a singleirregularly shapedentity
as more likely to be the product of a processthat was directed at creating
that particular structure than subjectswho saw only a single instance of
the irregularly shapedentity. Theseresults show that if subjectsare given
evidencethat a particular structure is unlikely to have arisen through an
undirected process(e.g., becausethere exist multiple identical instances),
they are more likely to construe the entity as an object and refer to it
through the use of a count noun than if they are not provided with such
evidence.
Finally , there was a fourth condition in which subjectswere initially
presentedwith three nonidentical irregularly shapedentities before being
asked to choose a description for the single irregularly shapeditem. As
predicted, simply seeingmultiple instancesthat were not similar in struc-
ture to the target entity did not lead subjects to choose a count noun
description any more often than they did when they saw only the single
instance. Also, as predicted, seeingmultiple nonidentical instancesdid not
affect subjects' intuitions concerning the likelihood that the structure of
the target item was the result of a directed process.
A secondexperiment investigatedwhether providing subjectswith evi-
dence of a structure-dependent function would lead them to be more
likely to construe an irregularly shapedentity as an object and refer to it
through the use of a count noun than if they were not provided with evi-
dence of a structure-dependentfunction and were instead only provided
with evidence of a structure-independent function. Half of the subjects
were shown a seriesof videotaped trials in which an irregularly shaped
entity was usedto perform a structure-dependentfunction (e.g., as when a
key is used to open a lock); the other half were shown a presentation
in which the irregularly shapedentity was used to perform a structure-
independentfunction (e.g., as when a key is used to ring a bell). At the
end of each trial , the entity appearedon the screenand, as in the previous
experiment, subjectshad to indicate whether they preferred a count noun
Namesfor ThingsandStuff 127

or a mass noun description of the entity. As expected, subjectschosethe

count noun description more often in the structure-dependentcondition
than in the structure-independentcondition.
Finally , a large body of researchshows that even very young children
construe names for regularly shaped solid entities as names for objects,
whereas they interpret names for regularly shaped nonsolid entities
as names for the kind of material (Dickinson 1988; Hall 1996; Hall ,
Waxman, and Hurwitz 1993; lmai and Gentner, in press; Landau, Smith,
and Jones 1988; Markman and Wachtel 1988; Soja, Carey, and Spelke
1991; Soja 1992). These results are predicted if in our conception of
matter and objects, matter lacks definite form, but objects have a definite
form. Nonsolid entities clearly do not have a definite form (as is evident
from their lack of cohesiveness ); thus, according to the present proposal,
labels for them should be interpreted as names for the kind of material.
Conversely, regularly shaped solid entities have a definite form (as is
evident from their cohesiveness ), and furthermore, the forms are regular
and thus unlikely to have arisen by accident, but instead are likely to be
the result of directed processes ; thus, according to the present proposal,
labe~s for them should be interpreted as names for the kind of object
rather than as namesfor the kind of solid material.
The experimentsdiscussedabove suggestthat we expectobjectsto have
a definite form that is the result of a directed processand that they possess
functional properties that depend on that structure, whereaswe do not
have the sameexpectationsof amounts of matter.5 Thesedata show that
our conceptionsof what constitutesan object as compared to an amount
of matter is in important ways Aristotelian . The hypothesisthat our con-
ceptual representationsof things are Aristotelian in nature is developed
further in the next few sections.

7.3 AristotelianEssences

Given that, according to Aristotle, it is essencesthat are the cause of

things' being what they are, it is important to get a better understanding
of the nature of Aristotelian essencesand how they may be represented.
Aristotelian essentialism differs from many contemporary essentialist
theories in important ways. For example, most contemporary theories
would not consider having two feet an essential property of humans;
however, Aristotle claims that humans are essentially two-footed. How
are we to reconcile these positions? Does Aristotle think that when a
128 Prasada

person loses a foot , he or she stops being human ? Clearly not . Aristotle
would agree with our intuitions that a person who loses a foot continues

to be a person . However , he would not conclude that having two feet

is merely a stereotypical but nonessential property of humans . Instead ,
he would argue that humans are essentially two -footed , though certain
individual humans may be accidentally one-footed . Thus , the person who
has lost a foot may be said to be essentially two -footed but accidentally
one - footed .

The difference between these two conclusions is best understood by

considering how essential and nonessential properties are determined by
the two types of theories . Many contemporary approaches determine
essential and nonessential properties on the basis of whether something
continues to exist as the kind of thing it is even when the property in
question is no longer true of it , or by considering whether things that lack
the property in question may nevertheless belong to the kind in question .
This is the view of essentialism expressed in the following quotations from
Kripke and Lakoff :6

Another example one might give relates to the problem of essentialism . Here is
a lectern . A question which has often been raised in philosophy is: What are its
essential properties ? What properties , aside from trivial ones like self-identity , are
such that this object has to have them if it exists at all , are such that if an object
did not have it , it would not be this object ? ( Kripke 1971, 178- 179)

Essentialism : Among the properties that things have, some are essential: that is,
they are those properties that make the thing what it is, and without which it
would not be that kind of thing . ( Lakoff 1987, 161; original emphasis)

To see why Aristotle does not pursue this strategy in determining essence,
we must recognize first that the essence, for Aristotle , is not a set of nec-
essary and sufficient properties or a probabilistic cluster of properties .
Instead , an Aristotelian essence is the cause that defines what something
is. This is why it provides an adequate response to the question " What is
it ?" In this respect, the Aristotelian view of essencesis similar to " psycho -
logical essentialism" (e.g., Medin and Ortony 1989; Keil 1989; Gelman ,
Coley , and Gottfried 1994) . Although the exact relation between the
present proposal and psychological essentialism is unclear , one way to
think about the foffi1er is as an explicit proposal concerning what con-
stitutes an essencefor us and the basis upon which we distinguish essential
and nonessential properties .
In the case of an object such as a human being , the essenceis the sub-
stantial foffi1 that , by infoffi1ing the appropriate kind of matter , gives rise
to a human being . Recall that the substantial form is the process of
Namesfor ThingsandStuff 129

growth and maintenance that is directed at creating a particular kind of

structure that allows the entity to function in ways that are responsible for
its process of growth and maintenance . In the case of human beings, the
process of growth and maintenance is directed at creating an entity that
has two feet . In some cases , however , because of causes other than the
essence, such as the environment in which the growth occurs, or an acci-
dent , the resulting body will have only one foot . Thus , the contrast
between essential and nonessential properties , for Aristotle , lies in the kind
of cause that is responsible for the property in question , rather than in
whether the property is required for membership in the relevant kind .
If a property is part of the " blueprint " that the process of growth or
creation is trying to implement , and thus in normal circumstances caused
by it , then the property is considered to be essential. If , on the other hand ,
the property is the result of other processes (causes), or due to errors in
the " implementation " (growth / construction ) process, then the property is
considered accidental . Therefore , particular objects (e.g., dogs) are what
they are by virtue of their (dog) essence, but they are the way they are by
virtue of their essence and other causes. For example , some dogs are wet
because they are outside and it is raining ; other dogs may have three legs
because of an accident , or have fleas because of where they live ; others
may be skinny and weak because they don 't get enough , or nutritious ,
food ; still others may be overweight because they eat too much fatty food ;
and so on .
It should now be clear that Aristotle does not require that all members
of a given kind actually possess a specific set of properties , even essential
properties , because actual members of the kind are subject to other causes
that might introduce other properties , fix parameters left open by the
essence, and even remove essential properties . In fact , the Aristotelian
conception of essencesand objects predicts that , in general, not all mem -
bers of a kind will have all their essential properties because different
members will be subject to different causes and some of those causes will
destroy various essential properties . In general, it will also be the case that
those properties that are shared by most members of the kind are more
likely to be essential than those that are shared by only a few members .
Finally , essential properties should , in general, be found to be present in
most members because, in most cases, there should not be any systematic
cause that interferes with the expression of essential properties . It is
important to note , however , that these statistical generalizations are only
heuristic and are neither necessary , nor constitutive of , essences .
 130 Prasada

7.4 Representationof Essences

: Evidencefor the Aristotelian Conception

The manner in which we interpret sentencessuch as (5) and (6) pro-

vides evidenceconcerning the manner in which we think and talk about
essencesand how they are related to essential and nonessentialprop-
erties (Chomsky 1975a; Moravcsik 1994).7
(5) a. Chickenslive to be x years of age.
b. Chickenslive to be x years of age, by virtue of what they are.
c. All chickenslive to be x years of age.
(6) a. Chickenslive in chicken coops.
-.
b. Chickens live in chicken coops, by virtue of what they are.
c. All chickenslive in chicken coops.
Notice first that although we feel that (Sa) and (5b) are almost identical in
meaning and may be consideredto be paraphrasesof eachother, the same
is not true of (6a) and (6b). This shows that although we feel that the
property of living for x years is causedby the essenceor nature of chick-
ens, the property of living in a chicken coop is not causedby their essence .
As has often been noted, the bare plural is not equivalent to universal
quantification, and thus neither (Sa) nor (6a) is interpreted as a statement
about all chickens, and both (5c) and (6c) are judged to be false. Instead,
both (Sa) and (6a) can be interpreted as making statementsabout what is
generally, though not always, true of chickens. Thus, either statement
could be made as long as the predicatesare true of a suitably large per-
centage of chickens. Crucially, though this statistical interpretation is
possible for both (Sa) and (6a), it is necessaryfor (6a), but not (5a).
Therefore, it would be false to say (6a) if a suitably large number of
chickensdid not live in chicken coops- for example, if tomorrow we let
all chickens out of chicken coops. However, we continue to judge (Sa) to
be a valid statementeven though it is the casethat the vast majority of
chickensare raised on poultry farms and are killed long before they reach
the age of x . Finally , notice that even though no particular percentageof
chickensneed to live to be x years of age in order for (Sa) to be true, we
tend to assumethat the predicateis true of most chickensand only become
aware of the contingency of this fact when a systematicreason why the
essentialproperty is not expressedis pointed out to us. Unless such a
cause is made explicit, we assumethat there is no systematiccausefor
chickens' lacking an essentialproperty and thus assumethat the property
is true of the majority of chickens.
Names for Things and Stuff 131

These data show that (1) we distinguish essential and nonessential

properties in terms of their causes, not on the basis of whether all
instancesof the kind have the property, or need the property to belong
to the kind , (2) essentialproperties need not be true of the majority of
instances of the kind , and (3) it is assumedthat most members of the
kind will have essentialproperties, unlessthere is an explicit reason why
a cause other than the substantial cause is systematically present and
responsiblefor destroying an essentialproperty. Thus, the interpretation
of generic sentencessuch as (Sa) provides evidencefor the Aristotelian
notion of what an essenceis and how essentialand nonessentialproperties
of things are to be distinguished.
It should also be noted that thesedata are not explained by essentialist
theories that assumethat essentialand nonessentialproperties are to be
distinguishedon the basis of whether a property is necessaryfor existence
as a particular kind of thing. The qualitative differencein the interpreta-
tion of sentencessuch as (5a) and (6a) also poses serious problems for
theories that do not posit a distinction betweenessentialand nonessential
properties in our conceptual representations and instead assume that
properties are to be distinguished simply in terms of strength of belief,
frequency, or centrality. Note that this does not mean that we cannot
revise our beliefs about which properties are essential- we can, and
sometimesdo. The point is that the properties we deem essentialplaya
qualitatively different role in our thought and speechthan the properties
we deemnonessential.

7.5 AcquiringNamesfor ThingsandStuff

The foregoing discussionsuggeststhat we have at least two ways in which

we naturally understand the entities that populate our world . These two
ways of understanding explain different aspectsof the spatiotemporally
defined entities. To construe an entity as an object (an object of some
kind) is to understand the structure of the entity as being the product of
a process directed at creating that structure (out of a suitable kind of
matter), such that the functional properties that the structured matter
affords are responsible for the growth and maintenance of the object.
Furthermore, this entails that the unity present in the entity is causedby
the substantial form and thus is essentialto its being that object. Finally ,
this also leadsus to considercertain structure-related properties including
parts and structure-dependentfunctions to be essentialproperties of the
 132 Prasada

entity becausethey are causedby its essence . Alternatively , to construe

the entity as an amount of matter is to say that the structure of the entity
is not the product of a directed process; thus, we do not consider any of
the structure-dependent properties to be essential. Instead, we consider
some of the structure-independentproperties (e.g., burnability , hardness,
texture) to be essential(i .e., due to the material cause). Furthermore, the
unity presentin the entity is not consideredto be the result of the material
cause(and thus the unity is not essential).
This account allows us to fill in an essentialcomponent of what is in-
volved in the acquisition of namesfor things and stuff. Putnam (1975) and
Kripke (1982) have made a strong argument that there is an indexical
component to the representationof natural kind terms such as gold and
tiger as well as artifact kind terms such as table.8 The key idea is that the
referenceof a term such as tiger is not determinedby a description, a set
of descriptions, or beliefs we may have about tigers, but is determinedby
the real-world essenceof tigers. Within this type of theory, there is an
" introduction " or " baptism" event in which the word tiger is first usedto
talk about a particular tiger. After this, the word is taken to refer to all
things that have the same essence - that is, all the things that are tigers
(i.e., things that are of the samekind as the thing that was baptized tiger).
We may, of course, mistakenly call things tigers that are not tigers;
however, the mere fact that we recognize this possibility shows that we
distinguish the information on the basisof which we chooseto call some-
thing a tiger from the information that determineswhich things actually
are tigers.
Although the part of the theory that is concerned with the roles of
indexicality and descriptions in a theory of referencehas been discussed
in considerabledetail, a crucial aspect of the introduction event has not
received much attention. Putnam (1975) describeswhat is involved in
introducing a natural kind teffil as follows:
Therearetwo obviouswaysof tellingsomeone whatonemeansby a natural-kind
tenn suchas'water' or 'tiger'. Onecangivehim a so-calledostensive definition-
'this (liquid) is water'; 'this (animal) is a tiger'; 'this (fruit) is a lemon'; wherethe
parentheses aremeantto indicatethat the 'markers ' liquid, animal,fruit , may be
eitherexplicitor implicit. . . . (p. 229)
Clearly, the markers have a central role to play in the introduction of a
novel word. Without a theory of what markers are, the type of theory of
referenceproposed by Putnam and Kripke cannot be exploited. It is the
(implicit ) markers that help distinguish the referenceof the words table
Names for Things and Stuff 133

and wood in (1) and (2). I propose that the account of objects and matter
developedin this chapter constitutes a plausible theory of the semantic
markers 'object' and 'material' and that an adequateaccount of the ref-
erenceof terms such as table and wood requires thesemarkers.
Given this account, we can characterizewhat is involved in the acqui-
sition of the words table and wood in an ostensivecontext as follows. A
demonstrativedirects the learner's attention to a spatiotemporally defined
entity. Next, the learner must decide(unconsciously) whether the entity is
an object of somesort or simply an amount of matter. Actually , what the
learner really has to decide is whether the speaker, in using the word to
refer to the entity in question, does so with the intention of referring to it
as an object or as some matter. Insofar as the learner makes the wrong
guess, the learner will have interpreted the word " incorrectly." This, of
course, is the reasonwhy as pragmatically competent usersof a language
we would never chooseto teach the word wood through an ostensivedefi-
nition such as (2) in which the entity in question is a table if we knew that
the learner did not know that the name for that kind of object is table. If
we were to teach wood in this context, we would probably do so by first
teaching the object name and then making the relation of the material
name to the object explicit (e.g., " This is a table. The table is made of
wood" ). In general, we would chooseto introduce a material name such
as wood through a sentencesuch as (2) only if the entity in question was
unlikely to be construed as an object, as when the entity is irregularly
shaped and has no obvious structure-dependent properties. Pragmatic
competencemay minimize the probability of miscommunication but does
not, in and of itself, provide us with an analysis of what is involved in
learning a word like table, and how (or if ) this differs from what is
involved in learning a word like wood. In fact, the type of pragmatic com-
petencedescribedhere probably makes crucial use of this infoffi1ation.
In deciding that the entity is essentiallyan object, the learner interprets
the novel word as referring to all objects of that kind . In doing so, the
learner has posited that the entity is essentiallya unit and that the struc-
ture that is present in the entity is (for the most part) the product of a
process directed at creating that structure, which, in turn, functions
in ways that are the reason for the (continued) existenceof the process
itself.9 The decision about whether to think of the entity as an object or
an amount of matter is made upon the evidenceavailable to the learner.
In addition to the perceptual cuesdiscussedin section 7.2, adult learners
can usethe foffi1 classof the entity to deteffi1inewhether the entity should
 134 Prasada

be thought of as an object or as matter ; however , as Macnamara ( 1982)

has pointed out , this evidence cannot be used in the earliest stages of word
learning because the surface cues that signal count and mass nouns vary
from language to language . Soja, Carey , and Spelke ( 1991) have provided
empirical evidence that children can in fact exploit perceptual cues about
how to construe novel entities when learning names for the entities before
they have mastered the count noun / mass noun distinction .
Importantly , to construe the entity as an object , and to fix the reference
of the novel term as referring to objects of that kind , the learner need not
know (and usually does not know ) what the directed process is, or even
the nature of the directed process (biological / intentional ), or which func -
tions that are afforded by the structure are the essential ones, only that
such processes and functions exist and are responsible for the structure of
the object . In fact , it is even logically possible, though unlikely , that the
learner could find out that tables are not the result of intentional processes
but are the result of a biological process and thus organisms of some
sort rather than artifacts (see Putnam 1975 for a more detailed discussion
of this kind of possibility ) . In this case, it is perfectly natural to say
- - .,

something like (7). However , it does not make sense to say something like
(8a,b) .

(7) It turns out that tables are organisms .

(8) a. It turns out that tables are ( just ) wood ./ It turns out tables are
(simply ) matter .
b . It turns out that hunks of wood are organisms .

This is because the form of the understanding associated with an object

term such as table is incompatible with the form of the understanding
associated with a material term such as wood. In the baptism , table was
used to refer to something that is essentially a unit , whereas wood was used
to refer to something that is not essentially a unit .! 0 Thus , the logic of the
terms , which derives from the nature of the things referred to , is incom -
patible . That a term that is stipulated to refer to things that are under -
stood to be units essentially (objects) cannot refer to things that are
understood not to be units essentially (amounts of matter ) is a logical fact
that is unrevisable . We can , of course , revise what we believe the entities

in question to be. The change in belief can be expressed through the use of
sentences such as (9a,b) in which the entities in question are referred to
through the use of thing , which , like entity , is noncommittal with respect
to whether the entity is essentially a unit or not (Macnamara 1994).
Names
for ThingsandStuff 135

(9) a. Those things we thought were tables are simply hunks of wood.
b. Those things we thought were simply hunks of wood are tables.
Note that we may appeal to experts to decidewhen we should make this
type of revision (Putnam 1975); however, nothing experts could tell us
would allow us to meaningfully make statementssuch as those expressed
in (8a,b). The distinct logic of the terms table and wood derivesfrom their
status as namesfor objects and matter that are essentiallyunits and non-
units. Thus, learning words like table and wood requires the ability to
construe entities as essentiallybeing units or nonunits. The Aristotelian
conception of objects and matter allows us to formulate what this ability
consists in, and provides a characterization of what the form of the
essences in the two casesis, such that whether the entity is thought of as a
unit or not follows as a consequence .! 1

7.6 AbstractDefinitionsandtheRepresentation
of Superordinate
Concepts

Finally , I want to briefly consider how the Aristotelian conception of an

object also allows us to developMacnamara's (1982) suggestion, basedon
the ideas of Cassirer (1953), that we regard the representationof super-
ordinate categoriessuch as ANIMALand VEHICLE as being related to the
representationof more specificcategoriessuch as DOGand CARin a way
that is analogousto the way that mathematiciansregard the relation of an
expressionsuch as y :=: mx + b to the expressiony :=: 2x + b.
There are two standard ways in which psychologiststhink of the rela-
tion of a category to a superordinate category. The first is to think in
extensional terms, in which case the superordinate category is simply a
superset of the extension of the lower-level category. Alternatively,
thinking intensionally, superordinate categoriesare seenas being repre-
sented by a subset of the features used to represent the more specific,
lower-level categories. This is a perfectly reasonableway of proceedingif
one's representation of concepts consists of sets of features. If , on the
other hand, our representationof what somethingis, is not representedas
a set of features, but is instead a representationof the essencethat causes
somethingto be what it is, then an alternative way of representingsuper-
ordinate categoriesmust exist. Furthermore, we should expectthe form of
the superordinaterepresentationto be constrained by the representation
of the basic-level category.
 136 Prasada

It is in this connection that it is helpful to considerthe way in which the

formula that definesa particular kind of line (e.g., all lines of slope 2) is
represented(in a Cartesian coordinate system), and how this representa-
tion is related to the representationof the general class of lines of which
the first class is a subset. A line is a single thing that is generatedby a
formula that is defined in terms of two components: a slope and a y-
intercept. The slope and the y-intercept are not independently existing
featuresthat are combined to produce a line, but they are componentsof
a line that representand explain different aspectsof the line (slape: the
rate of change of y with respectto x; and y-intercept: its position within
the coordinate system, specifiedby where the line crossesthe y-axis). Both
the slope and the y-intercept are aspectsaflines . The formula for the line
is representedin terms of thesecomponents. The parallelism betweenthe
definition of a line and its componentsand the definition of an Aristote-
lian object and its componentsis, I hope, evident. An Aristotelian object
is a single thing that is generatedby an essencewhose formula is repre-
sentedin terms of four components. The four causesare not independently
existing featuresthat are combined to produce the object, but components
of an object that explain different aspectsof the object. The matter, form,
function, and principle of rest and changeare aspectsafobjects. Thus, we
can expect that the formula that representsan object will be stated in
terms of thesecomponents.
Given this parallelism, we can turn to the representation of lines of
a specific kind (e.g., lines of slope 2) and the relation of this representa-
tion to a representationof a more generalclassof lines (e.g., all lines) for
insight into how one might representthe relation between.a specifickind
of object (e.g., dogs) and a more general class of objects (e.g., animals).
Lines of slope2 are representedby the formula y == 2x + b, the classof all
lines by y = mx + b. The more general formula is arrived at by replacing
the constant that definesthe slope with a variable. Note, specifically, that
the more general formula is not arrived at by looking at lines of slope 2,
3.5, 7.1, 4, . . . and finding all the featuresthat are common to them. There
are no independent features that are common to an lines or classesof
lines. The only thing that the different classesof lines have in common is
thefarm of their definition, which capturesthe relevant relations between
the distinct components of lines. Generalization is achieved by replac-
ing the constantsassociatedwith the componentsthat define a given class
of lines (slope and y-intercept) with variables that can take on a range of
values.
Names for Things and Stuff 137

This notion of generalizationby replacing constantswith variables can

be applied to derive the representationof substantial conceptsat varying
levels of generality. Recall that objects are defined as essentiallybeing a
compositeof matter and form. Furthermore, becausethe substantialform
is itself defined in terms of the structure, processof growth and mainte-
nance, and structure-dependentfunctions, the essence , or definition, of an
object may be representedas in (10), where m == matter, s == structure,
p =: process, and f == function.
(10) Object = G (m, s,p , f )
Although we do not know the exact form of substantial form (and there-
fore G), we do know the constraints that must hold betweenthe different
componentsthat define an object. The constraints are as follows: (1) the
processmust lead to the structure in the matter, (2) the structure must
afford essentialfunctions, (3) the functions are responsiblefor the process,
and (4) the matter must be of the appropriate sort (suchthat the processis
able to form the required structure out of the matter and that it allows the
structure to have the function(s) required by the processof growth and
maintenance). It is important to note that the representationin (10) is
meant to be simply a generalway of stating the formula that generatesan
object and should not be taken to be a mathematical function.12 Given
this formulation , categoriessuch as DOGcan be representedby constants
for each of the componentsthat definethe object, whereascategoriessuch
as TABLEcan be representedas having constants for each of the compo-
nents except for the material component, which will have, in place of a
constant, a variable that can be filled in by any constant that satisfiesthe
relevant constraints on the nature of the matter of the object. Thus, the
representationprovides a straightforward way for us to representintu-
itions concerning the possibility of different ways of realizing an essence
through the use of a variable, and specificrealizations through the use of
specificconstants. The constantsin the caseof objectsare not numbers as
is the case with lines, but instead are representationsof the essential
shape, the essentialactivities, the nature of the principle of change and
rest (biological/intentional), and the type of matter.
It should now be clear how we can derive the representationfor the
category ANIMALfrom the representation of the category DOG. This is
achievedby simply substituting the variables for constants for structure
and matter. Dogs, by virtue of having a certain structure and being com-
posed of a certain kind of matter, are capable of moving in order to get
 138 Prasada

food, to avoid danger, and generally to engagein activities that allow

them to maintain themselves , grow, and reproduce. These functions,
which are responsiblefor (continued) existence, can be instantiated by a
variety of objects, each possessingits distinctive form and matter (and
thus satisfying these functions in its own characteristic manner) while
being the product of the sametype of processof growth and maintenance.
Thus, when we think of a dog as an animal, we are thinking of the entity
from a perspectivein which we consider the particular form and matter it
has as one of severalcompositesof form and matter that are capable of
functioning in ways that allow them, just as they allow the dog, to grow,
maintain, and reproduceitself. An advantageof this approach to forming
representationsof general categoriesis that it is possible to do so on the
basis of very limited information . In principle, this can be done on the
basis of the representationof a single basic-level category. It is important
to note that the present account is concernedwith the representationof
types of things and the relation betweentypes of varying generality (e.g.,
DOG /ANIMAL ; CAR/"EHICLE) rather than the representationthat is required
for conceptssuch as those representedby laundry, livestock, or groceries
in which only particular tokens of various types are grouped together to
form categories.
The type of relation that is found between the representationof cate-
gories such as DOGand ANIMALor CARand VEHICLE is one in which the
representationsof the former, more specific categoriesare a special case
of the representation of the latter, more general categories. A different
type of relation is found betweenthe representationsof categoriessuch as
TABLEand FURNITLTRE or SillRTand CLOTffiNGin English. In thesecases,
the more specificrepresentationsare not a specialcaseof the more general
representation. Instead, the more specific representationsprovide only a
partial representationof the more generalrepresentations,which are only
fully representedthrough the combination of multiple specific represen-
tations. Thus, we may think of the representationsfor FURNITURE or
CLOTHING as piecewise-defined representationswhose piecesare given by
the representationsthat are derived by substituting variables for specific
values for structure and matter. Furthermore, becausethe piecewise-
defined representationsdo not directly refer to anything that is essentially
a unit, they are lexicalized as massnouns rather than as count nouns.
This account of the representationof superordinatecategoriessuggests
that there are two ways in which we can think about functions: as being
discreteor nondiscrete. Discretefunctions are thosethat are either satisfied
Names for Things and Stuff 139

or not, and cannot be satisfied only partially or to varying extents. Con-

sider, for example, the ability to transport: either somethingis capable of
transporting (e.g., cars, buses) or it is not (e.g., chairs, necklaces ) , and
thus multiple items of different kinds do not help satisfy this function any
more than a single item of the relevant kind does. On the other hand, a
function such as furnishing or clothing is one that can be satisfied to
varying degrees.We even have expressionssuch aspartially furnished and
partially clothed. Single items within the category are unable to fully sat-
isfy the category function- a single chair does not furnish a room, nor
does a shirt clothe a person. Furthermore, simply putting on multiple
shirts or putting multiple chairs in a room doesnot allow one to be fully
clothed or a room to be fully furnished. What is required is that multiple
items of different kinds, each of which is capable of partially clothing or
furnishing, be used.13
A straightforward prediction of the proposed account of superordinate
representation is that instances of superordinate categories that are
referred to by mass nouns should be more likely to be used together to
perform somefunction than instancesof superordinatecategoriesthat are
referred to by count nouns. This is, in fact, what Wisniewski, Imai, and
Casey (1996) and Prasada, Kim , and Eschemeyer(1997) have found.
These data provide some initial support for the representationof super-
ordinate categoriesproposed above, which is based on the nature of the
representationof basic-level categoriesthat follows from the Aristotelian
conception of the essenceof things. Importantly , the current proposal
demonstrateshow the relation betweenbasic-level and superordinatecat-
egoriesmay be represented, if it is the casethat basic-level categoriesare
to be representedin terms of an essencerather than as a set of properties
as many theories of conceptual structure now maintain. There are, of
course, many questions that this approach to superordinate representa-
tion raisesthat cannot be addressedhere and must be the subjectof future
research.

7 .7 Summary and Conclusion

In this chapter, I have argued that the Aristotelian distinction between

object and matter plays a crucial role in the way that we think and talk
about things and stuff. The distinction is required for explaining the ref-
erential properties of words like table and wood. These words enter our
lexicons as namesfor a particular kind of object or as a name for a par-
 140 Prasada

ticular kind of matter. Although we generally do not know the nature

(essence) of the object or the matter in question, we do know the distinct
types of essencethat enter into the referenceof the terms. Furthermore,
becauseobjects are essentiallyunits whereasmatter is not, the two types
of words must be representedby logically different types. The modes of
understanding and talking about what we see that are representedby
thesetwo types of words are simply incompatible.
Given the Aristotelian characterization of the essencesof objects and
matter, and thus the mode of understanding that is required to grasp
something as an object or as matter, it was possible to formulate, in a
principled manner, the typesof perceptualevidencewe might useto trigger
a particular mode of understanding. Evidenceconcerning the definiteness
of structure comes from cohesiveness ; evidencethat the structure is the
product of a directed processcomes from the regularity of the shapeas
well as the repeatability of the structure; finally, evidencefor the presence
of a final causecan be indicated by the presenceof a structure-dependent
function. Each of thesetypes of evidencewas shown to influencewhether
experimental subjectsthought and spoke about entities as objects or as
matter.
Although the focus of this chapter has been on names for things and
stuff and trying to explicatewhat is involved in understandingan entity as
some one thing rather than an amount of stuff, the factors that account
for our thinking of objectsas units are likely to be at work in determining
what is involved in understanding multiple objects as some one thing
(e.g., as one forest) rather than as many objects (e.g., as many trees). The
work of Paul Bloom and colleaguessuggeststhat this is in fact the case
(Bloom 1996a,b; Bloom and Kelemen 1995; Bloom et al. 1995).
Finally , the representationsof objects in Aristotelian terms provide us
with a natural way of constructing a theory of how it is that we are able
to generaterepresentationsof categoriesof varying generality, as well as
an explanation for the levels of generalization that are most natural.
Pustejovsky (1995) provides extensive evidence that various linguistic
phenomena can be accounted for if we assumethat the nature of the
understanding provided by various lexical items is representedin Aris-
totelian terms. The findings in this chapter complement Pustejovsky's
findings and suggestthat much progress may be made concerning the
manner in which we naturally understand and talk about the world , if
we approach the issuesfrom the perspective provided by Aristotelian
metaphysics.
Names
for ThingsandStuff 141

Notes

1. Here , as elsewhere, a range of putative counterexamples can be formulated . To

keep the exposition clear and concise, I will address only the most obvious ones.
For more detailed discussion , see Prasada 1998. One might argue that the car -
penter built the table simply because of an urge to build . Note , however , that in
making this argument , one is conceding that the fact that a table was built is a
complete accident (the carpenter could just as well have built a chair or a house or
a random conglomeration of boards ) and thus there is no explanation for why the
structure is what it is and not some other structure . But this is not the kind of case
Aristotle is interested in . He is interested in what explains the regularity with
which certain processes culminate in the creation of a particular structure and not
others . That is, in the cases where there is an explanation , what is the nature of the
explanation ? His answer is that the explanation lies in functional properties of the
created structure . It is important to note that within the present approach , inten -
tional and functional explanations are not independent or opposed to one another .
A functional property of the intended structure provides the reason for the inten -
tion to create the structure in question . A different class of putative counter -
examples can be created by pointing out that the table may have been built simply
because the carpenter was paid to do so, was told to do so, was threatened into
doing so, and so on . Note , however , that these kinds of cases raise questions of the
sort , " Why would someone pay the carpenter to build this structure and not some
other structure ?" ; and these questions are ultimately to be answered by citing
some functional property of the structure . This is where explanation ends. Once
we have a reason for why the process is directed at creating the structure it does,
we feel we have a complete explanation . If , on the other hand , we are asked just to
accept as a brute fact that there exist processes that are directed at creating a
structure and there is no explanation for why they create the structures they do , we
feel that an important regularity has gone unexplained . Finally , note that teleo -
logical explanation is not necessarily intentional and thus is also applicable to
nonartifacts (on this point see Asma 1996; Broadie 1987).
2. What I have described is the case in which the entity is believed to be simply a
hunk of wood and not a table . If the entity in question is believed to be a table ,
then a statement such as (2) does not imply that no directed process gave rise to
the structure . Instead , it means that in thinking of the table as wood , the fact that
it has the structure it does as the result of a directed process is not considered to be
relevant .

3. The present account of our conception of artifacts accords a central role to

functional properties . This seems to conflict with arguments put forth by Malt and
Johnson ( 1992) and Bloom ( 1996a) that seem to suggest that functional properties
are not a necessary component of our conception of artifacts . It is unclear , how -
ever, the extent to which Malt and Johnson 's and Bloom 's criticisms apply to the
current way of thinking about essences(see text below ) and how various aspects of
the essence, including the final cause, are related to each other . Space limitations
prevent a more detailed discussion of these issues ; however , I take them up in
Prasada , in preparation .
142 Prasada

4. In this and all other conditions in this and the next experiment, there was no
differencein subjects' preferencefor either of the two massnoun locutions (There
is a piece ofblicket in the tray/ There is blicket in the tray).
5. It is important to quality this claim. There are entities such as those things that
we call clouds, rocks, pebbles, and boulders that we tend to think of as objects
rather than pieces of stuff, and yet we do not believe that these entities are the
result of processesthat are directed at creating the structure that the entities have.
Why is it , then, that we treat theseentities as objectsrather than as piecesof stuff?
There are three possibilities. First, it is possiblethat the above account of what it
meansto conceiveof an entity as an object of somekind rather than as an amount
of solid material is sufficient but not necessaryand thus there are other ways in
which we can conceive of entities as objects of some kind rather than as an
amount of solid material of some kind. Second, it is possible that thesecounter-
examplesshow that the account is wrong and there is a single alternative account
of what is involved in conceiving of entities as objects of somekind rather than as
amounts of material of somekind. Third , it is possiblethat the account accurately
representswhat is involved in our conceiving of entities as objects of some kind
and as amounts of matter; however, the world presentsus with casesthat seemto
fall betweenthesetwo conceptual schemesthat are given by our ontology, and we
assimilate them as best we can. The intermediate casesbetray their problems in
being variable from languageto language, in the existenceof ambidextrouslexical
items within a single language (e.g., a stone/made of stone), and in the largely
arbitrary baseson which thesethings in the world are thought of as objectsand as
stuff (e.g., pebblesf*gravels). Although it remains an open question which of these
possibilities is right, the third has the advantageof providing a natural explana-
tion of various " problematic" aspectsassociatedwith the conceptualization of
theseentities, whereasthe first two would render rocks and stonesequally good
objects as dogs and tables and thus would require some other way of accounting
for their " problematic" aspects.
6. Lakoff does not endorsethis or any other view of essentialism
; however, the
quotation provides a conciseexpressionof the common view of essentialism.
7. The interpretation of genericstatementsinvolves many complexities. Here, I am
concernedonly with using it as a tool to illustrate the manner in which our con-
ceptual systemsdistinguish essentialand nonessentialproperties. For a discussion
of some of the complexitiesinvolved in the interpretation of genericsentences
, see
Chomsky 1975aand the referencescited therein.
8. For an application of this insight to artifact kind terms from a psychological
perspective, seeBloom 1996a.
9. The qualification on the characterization of the structure is necessarybecause
particular substanceshave the structure they do by virtue of their essenceas well
as other causes. Therefore, one of the tasks that facesthe learner is to determine
which aspectsof the structure are causedessentiallyand which are due to other
causes.
Namesfor ThingsandStuff 143

10 . Sentences such as This ring is jewelry do not constitute counterexamples to

this claim because , as will be seen in section 7 .6 , mass superordinates of basic - level
count nouns do not require thinking of the entities in question as not being units
essentially . In fact , mass superordinates such as jewelry require that the entity be
thought of as essentially a unit ( e .g ., a ring ) rather than as essentially not a unit
( e .g ., some gold ) .

11 . It might be suggested that what is crucial here is the " standard " distinction
between concepts of objects and substance (which presumably are conceived of as
units and nonunits , respectively ) rather than the Aristotelian conception of objects
and substances . The question , of course , is just how this " standard " distinction
between our concept of an object of some kind and a material of some kind is
represented and how the representation entails , rather than stipulates , viewing the
entity as a unit or nonunit . The Aristotelian conception of things and stuff does
just this .

12 . It is assumed that an appropriate formalism could be developed for the rep -

resentation of objects ; however , the details of such a formalism are not important
to the nature of the solution concerning the relation of the representation of spe -

cific categories and more general categories . Category theory , concerned as it is

with models of becoming and the realization of plans , may be a promising direc -
tion in which to search for the right type of formalism ( see Lawvere 1994 and
other papers in Macnamara and Reyes 1994b ) .

13 . The general idea that the difference between the representation of count and
mass nouns for superordinate categories is related to whether multiple instances of
a category are used for some common function is also found in McCawley 1975 /
1979 , Wierzbicka 1985 , and Wisniewski , Imai , and Casey 1996 ; however , the
present account differs from these proposals in degree of explicitness as well as
substance . Space limitations prevent an adequate discussion of the similarities and
differences .

References

Asma , S . T . ( 1996 ) . Following form andfunction : A philosophical archaeology of

life science . Evanston , IL : Northwestern University Press .

Bloom , P . ( 1996a ) . Intention , history , and artifact concepts . Cognition , 60 , 1- 29 .

Bloom , P. ( 1996b ) . Possible individuals in language and cognition . Current

Directions in Psychological Science , 5 , 90 - 94 .

Bloom , P ., and Kelemen , D . ( 1995 ) . Syntactic cues in the acquisition of collective

nouns . Cognition , 56 , 1- 30 .

Bloom , P . , Kelemen , D ., Fountain , A ., and Courtney , E . ( 1995 ) . The acquisition

of collective nouns . In D . MacLaughlin and S . McEwen ( Eds .) , Proceedings of the
19th Boston University Conference on Language Development . Somerville , MA :
Cascadilla Press .

Bolton , R . ( 1976 ) . Essentialism and semantic theory in Aristotle . Philosophical

Review , 85 , 514 - 544 .
 144 Prasada

Broadie , S . ( 1987 ) . Nature , craft and phronesis in Aristotle . Philosophical Topics ,

15 , 35 - 50 .

Cassirer , E . ( 1953 ) . Substance and function . New York : Dover .

Chomsky , N . ( 1975a ) . Questions of form and interpretation . Linguistic Analysis ,

1 , 75 - 109 .

Chomsky , N . ( 1975b ) . Reflections on language . New York : Pantheon .

Cohen , S . M . ( 1996 ) . Aristotle on nature and incomplete substance . New Yark :

Cambridge University Press .

Dickinson , D . G . ( 1988 ) . Learning names for materials : Factors constraining and

limiting hypotheses about word meaning . Cognitive Development , 3 , 15 - 35 .

Gelman , S . A . , Coley , J . D . , and Gottfried , G . M . ( 1994 ) . Essentialist beliefs in

children : The acquisition of concepts and theories . In L . A . Hirshfeld and S. A .
Gelman ( Eds .) , Mapping the mind : Domain specificity in cognition and culture .
New York : Cambridge University Press .

Hall , D . G . ( 1996 ) . Naming solids and nonsolids : Children ' s default construals .
Cognitive Development , 11 , 229 - 264 .

Hall , D . G ., Waxman , S . R ., and Hurwitz , W . M . ( 1993 ) . How 2 - and 4 - year - olds

interpret adjectives and count nouns . Child Development , 64 , 1651 - 1664 .

Imai , M . , and Gentner , D . (in press ) . Linguistic relativity vs . universal ontology :

Cross -linguistic studies of the object / substance distinction . Cognition .

Jackendoff , R ., and Landau , B . ( 1992 ) . Spatiallanguage and spatial cognition . In

R . Jackendoff , Languages of the mind . Cambridge , MA : MIT Press .

Keil , F . C . ( 1989 ) . Concepts , kinds , and cognitive development . Cambridge , MA :

MIT Press .

Kripke , S . ( 1971 ) . Identity and necessity . In M . K . Munitz ( Ed .) , Identity and

individuation . New York : New York University Press . Reprinted in Moore , A . W .
( Ed .) , ( 1993 ) . Meaning and reference . Oxford : Oxford University Press .

Kripke , S . ( 1982 ) . Naming and necessity . Oxford : Blackwell .

Lakoff , G . ( 1987 ) . Women , fire , and dangerous things : What categories reveal
about the mind . Chicago : University of Chicago Press .

Landau , B . , Smith , L . B ., and Jones , S . ( 1988 ) . The importance of shape in early

lexical learning . Cognitive Development , 3 , 299 - 321 .

Lawvere , F . W . ( 1994 ) . Tools for the advancement of objective logic : Closed

categories and toposes . In J . Macnamara and G . E . Reyes ( Eds .) , The logical
foundations of cognition . New York : Oxford University Press .

Lear , J . ( 1988 ) . Aristotle : The desire to understand . New York : Cambridge

University Press .

Macnamara , J . ( 1978 ) . How do we talk about what we see ? Unpublished manu -

script , McGill University .

Macnamara , J . ( 1982 ) . Names / or things : A study o/ human learning . Cambridge ,

MA : MIT Press .
Names for Things and Stuff 145

Macnamara , J. ( 1994) . Logic and cognition . In J. Macnamara and G . E . Reyes

( Eds .), The logical foundations of cognition . New York : Oxford University Press.
Macnamara , J., and Reyes, G . E . ( 1994a). Foundational issues in the learning
of proper names , count nouns , and mass nouns . In J . Macnamara and G . E.
Reyes ( Eds .), The logicalfoundations of cognition . New York : Oxford University
Press .

Macnamara, J., and Reyes, G. E. (Eds.). (1994b). The logical foundations of

cognition . New York : Oxford University Press.
Malt , B . C ., and Johnson , E . C . ( 1992). Do artifact concepts have cores? Journal
of Memory and Language , 31, 195- 217.
Markman , E . M ., and Wachtel , G . F . ( 1988). Children 's use of mutual exclusivity
to constrain the meanings of words . Cognitive Psychology , 20, 121- 157.
McCawley , J. D . ( 1975/ 1979). Lexicography and the count -mass distinction . In
J. D . McCawley ( Ed .), Adverbs, vowels and other objects of wonder. Chicago :
University of Chicago Press.
Medin, D ., and Ortony, A . (1989). Psychological essentialism. In S. Vosniadou
and A . Ortony ( Eds .), Similarity and analogical reasoning. New York : Cambridge
University Press.
Moravcsik , J. M . E . ( 1975) . Aitia as generative factor in Aristotle ' s philosophy .
Dialogue , 14, 622- 638.
Moravcsik , J. M . E . ( 1994) . Essences, powers , and generic propositions . In T .
Scaltsas, D . Charles , and M . C . Gill ( Eds.), Unity , identity , and explanation in
Aristotle 's Metaphysics . Oxford : Clarendon Press.
Prasada , S. ( 1998) . Conceptual representation of artifact kinds . Manuscript sub-
mitted for publication , Dartmouth College .
Prasada , S., Ferenz , K ., and Haskell , T . (in preparation ) . Speaking of things and
stuff : The role of perceived arbitrariness of structure . Unpublished manuscript ,
Dartmouth College and University of Southern California .
Prasada , S., Kim , J., and Eschemeyer, L . ( 1997) . [Ratings of frequency with which
exemplars of superordinate categories are used together ] . Unpublished raw data ,
Dartmouth College and University of Southern California .
Pustejovsky , J. ( 1995). The generative lexicon . Cambridge , MA : MIT Press.
Putnam , H . (1975). The meaning of " meaning ." In Philosophical papers : Vol. 2.
Mind , language, and reality . New York : Cambridge University Press.
Scaltsas, T . ( 1994a) . Substances and universals in Aristotle 's Metaphysics . Ithaca ,
NY : Cornell University Press.
Scaltsas, T . (1994b) . Substantial holism . In T . Scaltsas, D . Charles , and M . C . Gill
( Eds .), Unity , identity , and explanation in Aristotle 's Metaphysics . Oxford : Clar -
endon Press .

Soja, N . N . ( 1992). Inferences about the meanings of nouns : The relationship

between perception and syntax . Cognitive Development , 7, 29- 45 .
146 Prasada

Soja, N . N ., Carey , S., and Spelke, E . ( 1991). Ontological categories guide young
children 's inductions of word meaning : Object tenus and substance terms . Cogni -
tion , 38 , 179 - 211 .

Wierzbicka , A . ( 1985) . Lexicography and conceptual analysis . Ann Arbor , MI :

Karoma .

Wisniewski , E ., Imai , M ., and Casey, L . ( 1996) . On the equivalence of super-

ordinate concepts . Cognition , 60, 269- 298.
Witt , C . ( 1989) . Substance and essence in Aristotle : An interpretation of Meta -
physics VII - IX . Ithaca , NY : Cornell University Press.
Chapter 8
The Unity of Science and the Steven Davis
Distinction among Syntax,
Semantics, and Pragmatics

One of the outstandingissuesin the philosophyof linguisticsis how to

draw the distinctionamongsyntax, semantics , and pragmatics . A related
issueis what falls within thesedifferentdomains. On the onehand, there
are those who see theoriesof syntax, semantics , and pragmaticsas
theoriesof humanlinguisticcapacitiesthat are supposedto accountfor
thepropertiesof humanlanguages and'ofcommunicationinvolvingnatural
languages . Thisviewhasit that humanlanguages andcommunicationare
sui generis . Becauseof this, theoriesof humanlanguagesand linguistic
communicationarenot to be assimilatedto a generaltheoryof signs,but
aretheoriesaboutparticularhumancapacitiesthat differ in fundamental
nonreduciblewaysfrom the phenomenainvolving systemsof signsthat
aresupposedly foundin myths, art, clothing, films, andsoon. Thisviewis
found in the syntacticwork of N oam Chomsky , the psychologicalwork
of John Macnamara , and the pragmatictheory of Dan Sperberand
DeirdreWilson. Let uscall this viewthenarrowviewof humanlanguages
andnaturallanguagecommunication .
On the otherhand, thereare thosewho seetheoriesof syntax, seman -
tics, and pragmaticsvery broadly to include all phenomenarelatedto
systemsof signs, includingcultural, social, and cognitivefeatures , where
the systemof signscan includesuchthings as literary works, scientific
theories, paintings, dance, cinema, clothes, gestures , facial expressions,
and humanlanguages . Therearetwo approaches to what I shallcall the
broadview of syntax, semantics , andpragmatics . First, therearetheorists
who take humanlanguagesto be systemsof signsno differentin kind
from any other systemof signs. The theoriesthat accountfor the prop-
ertiesof systems of signscanbe appliedto any domainin whichsignsare
used, includinghumanlanguage . Suchtheoristsconsiderall systemsof
signsto have an underlyingunity and to fall within the samesortsof
 148 Davis

theories. This approach finds its expressionin various forms of what is

called semiotic- that is, in the work of such people as Roland Barthes,
Claude Levi-Strauss, Ferdinand de Saussure , and Charles Morris .1 Sec-
ond, there are theoristswho take the broad view of syntax, semantics, and
pragmatics, but leave it open whether all systems of signs can be
accountedfor within the sametheoretical framework. This is the view of
pragmatics adopted by the International Pragmatics Association, for
which " pragmatics is interpreted very broadly as a cognitive, social and
cultural perspectiveon language" (5th IPrA ConferenceBrochure), a view
that is called perspectivalism. Pragmaticsis viewed, then, as a perspective
on language, rather than as a component of a grammar, a discipline, or
even a field of study. Perspectivalism, as I hope to show, is the direct
descendantof Charles Morris 's views about pragmatics.
In this chapter I shall concentrateon the work of Charles Morris , one
of the founders of the current movementin semiotic, and try to show how
the semiotic program has its roots in logical positivism and in particular
the project of logical positivism to provide a unified science. I shall try to
show why Morris took the broad view of syntax, semantics, and prag-
matics and the relationship that this view has to the project of a unified
science. In closing, I shall make somecritical commentsabout this project
and try to show what consequenceit has for semiotic and also for the
study of natural language.
Logical positivism can trace its beginnings in 1923 in Vienna to a
seminar led by Moritz Schlick. The members of the seminar came from
diverse academicand intellectual backgrounds. Many of them, including
Schlick, who did his doctorate in physics with Max Planck, had training
in scienceor mathematics. It is thesetwo domains that gave the central
characteristicsto the logical positivists' movement. One way of viewing
this fusion is to consider it , as does Morris (1955, 64), as the union of
empiricism and rationalism. The rationalists, like Descartes, thought that
it was possibleto construct a pure theory of the physical universewithout
any appeal to empirical observations to justify the theory. All that was
necessarywas the supposeda priori truth that spacewas Euclidean and
that it and any eventsin it could be describedby Euclidean mathematical
models. The empiricists, on the other hand, looked to observation for the
justification of theories. Locke thought that there could not be any a
priori theoriesof nature, but that any theory of nature would have to arise
from observation. What was lacking from his view about theories of the
world, however, was any appreciation of the role that mathematicsplays
Unity of Science 149

in science. Modern science, especiallythe physical sciences , borrows from

both the empiricist and rationalist traditions. The empiricist appeal to
experience appears in the scientist's controlled experiments, and the
rationalist appeal to mathematics appears in the use to which scientists
put mathematical models in formulating their theories of the universe.
Morris draws our attention to another dimension of sciencethat he
adds to the mix of empiricism and rationalism, namely, pragmatism. A
scientific theory, according to Morris , can be thought of as a set of sen-
tences, constituting for Morris a language, the formal structure of which
can be modeled by the apparatus of logical theories and which can be
justified by its relationship to the world of objects. But formal struc-
ture and relationship to the world do not exhaust the nature of scientific
theories and thus of science. As Morris points out, scientific theories also
are related to the usersof the theory. On his view, the particular " rules,"
" procedures," and " conventions" that are operative in science" involve
referenceto users of signs. . ." (1955, 67). Appealing to the underlying
behaviorism of pragmatism, Morris regards such rules, procedures, and
conventions as " habits actually found in operation or set up by delibera-
tive convention" (1955, 67). Moreover, pragmatism emphasizesthe fact
that scientific theoriesare createdby peoplewho are actors within various
sorts of social institutions, and thus, in order to have a fuller understand-
ing of the connection betweenscientific theories and their users, we must
have an understandingof the social context of science.
On Morris 's view, then, scienceincludes three elements: first, its formal
structure embodied in the mathematical models that scienceemploys in
constructing its theories and in the logical structure that the resulting
theories have; second, its empirical character in its relationship to the
world , namely, in the confirmation relationship between observational
consequences of scientific theoriesand the experiencesthat either validate
or invalidate theseobservationalconsequences , thereby confirming or dis-
confirming the theories; and third , the relationship that scientific theories
have to the social institutions in which they are created and used and to
the people who think up the theories, confirm or disconfirm them, and
use them in understanding and controlling the world in which they find
themselves. Thus, science, Morris claims, can be characterized as an
admixture of traditional empiricist, rationalist, and pragmatist elements.
The logical positivists, however, did not seethemselvesas merely con-
tinuing the past empiricist, rationalist, and pragmatist traditions. They
wished to break with past philosophy and viewed their activity as a rejec-
 150 Davis

tion of what they regarded as traditional philosophical practice issuing

in the construction of various metaphysicaltheoriesthat were supposedto
give a total account of the world . The logical positivists took as their goal
the laying bare of the foundations and meaning of human knowledge,
which they regardedas obtainable only through scientificpractice. In fact,
they held that traditional metaphysical theories are not only of no value
for human knowledge, but, given the verificationist theory of meaning
that many of them subscribedto, literally meaningless. One consequence
of this total rejection of metaphysicsand with it traditional philosophy is
that the logical positivists did not think of themselvesas doing philoso-
phy, since they regarded most of the statementsof philosophy, not only
those of metaphysics, as not being open to verification and therefore as
meaningless.
What then was left for the logical positivists to do? One of the central
tasks they took for themselveswas the understanding of the nature,
meaning, and logical structure of science, since it was sciencethat was
supposedto provide whatever understandingcould be obtained about the
world . The initial problem that the logical positivists faced was that there
are a number of sciences , including not only the physical sciencesbut also
the human sciences , such as psychology, sociology, and economics. To
solve this problem, the logical positivists took as one of their central goals
to form an Einheitswissenschaft , i.e., a unifiedscience
comprisingall knowledgeof
reality accessible to man without dividing it into separate
, unconnected special
disciplines
, . . . (Joergensen1951,4)
The logical positivists thought that this goal could be achievedby " clar-
ifying the notions and statementsof science" (Carnap 1934, 33). Philoso-
phy is thus transformed into a discipline about the language of science,
including its logical structure, its empirical connection with the world ,
and the use to which scientistsput sciencein explaining and controlling
the world . Unification would be achievedby showing, despite the differ-
encesin the subjectmatter of the diverse sciences , that they sharea com-
mon logical structure, the samemethods of verification, and similar uses.
Gone then from philosophy are the overarching metaphysical theories
that are supposedto tell us what there is and the nature of what there is.
The answersto these questions are left to the special sciences , and phi-
losophy becomesthe study of the language of science. Morris calls this
approach to reaching the goals of unified science" the scienceof science"
or " metascience " (1955, 69). Hence the explosion of interest in the nature
Unity of Science 151

of language among the logical positivists and their descendants in the

analytical philosophical tradition .
As we have seen , logical positivism had its beginning in Vienna between
the wars , but its influence quickly spread beyond Vienna . Soon it counted
important groups in Berlin , Warsaw , and the United States ; groups in
Paris , Uppsala , Munster , Finland , and Great Britain ; and individual
proponents in Denmark , Argentina , Peking , and Norway . It was , in other
words , a worldwide movement , with a journal , Erkenntnis , which began
publishing in 1930 ; a project for an International Encyclopedia of Unified
Science , which grew to two volumes , with 19 monographs ; and congresses
that were held from 1929 to 1939 . Morris was an important member of

the American group . He was a professor at the University of Chicago ,

where Carnap moved in 1936 , and an associate editor with Carnap of the
International Encyclopedia of Unified Science .
Morris ' s Foundations of the Theory of Signs was published in 1938 as
the second monograph in the International Encyclopedia of Unified Sci -
ence . His work in semiotic has to be viewed , then , as a part of the logical

positivist movement and in particular as part of the goal of achieving a

unified science . Morris is quite explicit that semiotic theory should be seen
as part of the project of a unified science . He claims that

[s]emiotic holds a unique place among the sciences . It may be possible to say that
every empirical science is engaged in finding data which can serve as reliable signs ;
it is certainly true that every science must embody its results in linguistic signs . . . .
The sciences must look to semiotic for the concepts and general principles relevant
to their own problems of sign analysis . Semiotic is not merely a science among
sciences but an organon or instrument of all sciences . ( 1938 , 56 )

Semiotic , then , can be viewed as metascience , that is , the science of

.
SCIence .

According to Morris , semiotic , the theory of semiosis , is not just a

theory about scientific language . It is , he maintains , applicable " to all
signs , however simple or complex " ( 1938 , 10 ) . He holds that the following
disciplines can be absorbed into semiotic in whole or in part : logic ,
mathematics , linguistics , aesthetics , epistemology , the sociology of knowl -
edge , rhetoric , psychology , and the human social sciences . This is on the
surface a heterogeneous list , but what brings the disciplines together ,
Morris claims , is that they study in whole or in part systems of signs with
common elements that can be captured by his semiotic theory . Thus ,
Morris is led to give a wide and loose account of the central notions in
 152 Davis

semiotic. As well, his taking semiotic to be the scienceof sciencepulls him

in this direction.
According to Morris , as a metasciencesemiotic has three aspectsthat
correspond to the three philosophical positions, rationalism, empiricism,
and pragmatism, that are incorporated into logical positivism. The first is
the " rationalist" feature of science.

Scienceexistsas a body of written charactersand spokenwords. It is possible

to investigatethis linguisticresidueof the scientists' activity purelyformally. . . .
Thistypeof investigationmay, in onesenseof theterm, becalledlogicalanalysis ;
becauseof the ambiguityof the term 'logical', it maybepreferableto call it . . . the
syntacticalinvestigationof the languageof science . (1955, 69)
Hence, syntax is part of the semiotic theory of the language of science.
The secondfeature has its origin in empiricism.
The signswhichconstitutescientifictreatiseshave, to someextentat least, a cor-
relationwith objects
, andtheinvestigationof all aspects of this relationconstitutes
a secondtask of metascience . . . . This may be called. . . the seman tical investiga
-
tion of the languageof science. (1955,69) .
Thus, semanticsbecomespart of semiotic. Third , there is the elementthat
has its sourcein pragmatism.
[TJhesignswhichconstitutethe languageof sciencearepartsand productsof the
activitiesof scientists
. The studyof the relationof signsto scientists
maybecalled
. . . thepragmaticalinvestigationof thelanguageof science . (1955,70)
Consequently, pragmatics is part of semiotic.
Let us look more closely at Morris 's views about what should be
included in syntax, semantics, and pragmatics. According to Morris ,
" Syntactics . . . is the study of the syntactical relations of signs to one
another" (1938, 13). But syntax, on Morris 's view, not only encompasses
what linguistics takes to fall within syntax, but also includes logic and
mathematics. The reasonfor this is Morris 's concernto construct a theory
that is adequatefor an account of the formal structure of scientific theo-
ries. And both logic and mathematics playa role in giving scienceits
formal structure. But Morris 's " syntax" goes beyond even this wide
sweep. As well, he includes within its scopean account of analytic, con-
tradictory, and synthetic properties of sentences . In addition, he claims
that " syntax" is to account for synonymy, deal with the logical para-
doxes, and " clarify the modal expressionsof necessity, possibility and
impossibility" (1938, 15). Thus, Morris 's syntax encompassesnot only
Unity of Science 153

logical theory, including modal logic, but also phenomenathat fall within
what theorists such as Jerry Katz regard to be in the domain of semantics.
For Morris , " [s]emanticsdealswith the relation of signs to their desig-
nata and so the objectswhich they mayor do denote" (1938, 21). A word
must be said about 'designatum' and 'denotation'. According to Morris ,
" The designatum of a sign is the kind of object which the sign applies
to . . ." (1938, 5). Morris assimilates'kind' to 'classof objects'. Every sign
must have a designatum, he claims; that is, it must designatea class of
objects. However, it is not necessarythat every sign denote an object.
Consider 'Santa Claus'. This sign, on Morris 's view, designatesthe Santa
Claus unit class, but the class is empty. Hence, 'Santa Claus' does not
denote anything. (In a later work , Morris replaces 'designatum' with
'significatum' and gives it a behavioristic analysis, but it is not necessary
for our purpose to follow him into this dead end.) For every sign, Morris
holds, there is a semanticalrule that " determinesunder which conditions
a sign is applicable to an object or situation" (1938, 23). The form such
rules take is " The sign vehicle 'x' designatesthe conditions a, b, c . . . . . .
under which it is applicable" (1938, 24). If an object or situation, y, meets
the conditions, a, b, c, . . . , then ' x' denotesy . Morris 's view of semanticsis
quite close in many respectsto Carnap's. In recent years, a number of
decisivecriticisms have been raised against this conception of semantics
by Kripke , Putnam, Quine, and others, but I want to pass over these
problems and turn to Morris 's views about pragmatics.
Pragmatics, Morris claims, is " the study [of ] the relation of signs to
interpreters" (1938, 6). To have an understanding of Morris 's view of
pragmatics, we must understandhis use of 'interpreter' and 'interpretant'.
Semiosis, according to Morris , involves four factors: " that which acts as a
sign, that which the sign refers to, and that effect on some interpreter in
virtue of which the thing in questionis a sign to that interpreter" (1938, 3).
This gives us the sign vehicle, the designatum, the interpretant, and the
interpreter. Interpreters are agentsthat " take account" of the designatum
of signs. " Taking account" is a central notion in Morris 's semiotic, to
which he attachesthe technical term 'interpretant'. Morris gives a behav-
ioristic account of this notion. The interpretant of a sign is " the habit of
the organism to respond, becauseof the sign vehicle, to absent objects
which are relevant to a present problematic situation as if they were
present" (1938, 31). This givesrise in pragmatics to a pragmatic rule that
is, Morris says, " the habit of the interpreter to use the sign vehicle under
certain circumstancesand, conversely, to expect such and such to be the
 154 Davis

case when the sign is used" ( 1938, 32) . Here we see the tight connection
between Morris ' s views about semiotic and behaviorism . He takes it that

" from the point of view of pragmatics , a linguistic structure is a system of

behavior " ( 1938, 33) .
Morris ' s turn to behaviorism arises because of his concern with the

unification of science. Since semiotic is supposedly the science of science,

it is itself a science and can be studied in the same way that other sciences
are studied . A question then arises for semiotic as a science. How can it be
unified with the other sciences ? Morris ' s solution is to try to show how the
various aspects of semiosis (i .e., syntax , semantics, and pragmatics ) can
be accounted for by behaviorism . He argues that his account of the pro -
cess involving signs, that is, semiosis, " lends itself to treatment from the
point of view of behaviorists , . . ." ( 1938, 5). Hence , semiotic becomes part
of psychology . Consequently , there is no independent problem , apart
from the unification of psychology , about unifying semiotic with the rest
of science .

Morris claims that the behavioristic account that he gives of the basic
notions of semiotic is not necessary ( 1938, 6). Well that it is not , since
behaviorism is inadequate for the purpose . Sentences, for Morris , are
signs that are supposed to be connected with interpretants , that is, a habit
of the interpreter to respond . For there to be a habit , there must be con-
ditioning . And for there to be conditioning with respect to a particular
stimulus , there must have been past exposure of the subject to the stimu -
lus. But there can be no such conditioning and thus no such habit , as
Chomsky has taught us, connected with sentences. The reason is that
most of the sentences we hear and use are novel . We have never heard or
uttered them before ; hence , there can be no habit connected to their
comprehension or use.
It might be thought that Morris 's behaviorism can be sustained for
subsentential units . We can imagine that there are habits connected to all
or some of the morphemes in a language . Syntax , then , would be a system
that operates over these habits and composes them into habits connected
to sentences. There are many objections to this compositional behavior -
ism . First , there are many morphemes for which it makes no sense to
connect them with a habit . What , after all , would be the habit connected
with the syncategoric expressions 'the' , 'and ' , or 'but '? Second, the mini -
mum unit of use is a sentence. Therefore , for a morpheme to be connected
with a habit , it must be possible to use it as a one-word sentence. Some
nouns and verbs , like ' red ' , ' chair ' , ' run ' , and ' horse ' , pass the test , but
Unity of Science 155

other morphemes do not - for example , 'prudentially ' , 'possible', and

'rather ' . Third , in some cases a morpheme can be used as a one-word
sentence , but it is difficult to see what the stimulus patterns could be for
establishing a habit for the morpheme behind the sentence. What stimulus
patterns are connected with 'love ', 'quark ' , or 'contract '? Understanding
these terms comes with understanding theories or sets of laws in which
they are embedded or complicated forms of life that give them their
meaning . Even if these words were used as one-word sentences, there is no
possibility of connecting them with stimulus patterns . It seems, then , that
there is no hope for saving behaviorism , even in its more sophisticated
compositional version .
So Morris 's behaviorism must be dropped , and we must return to his
" taking account of ." The problem with 'interpretant ' is that it is a very
broad notion. It includes, for Morris , " all the psychological, biological
and sociological phenomena which occur in the functioning of signs"
( 1938, 30)- that is, any response by an interpreter or effect on an inter -
preter that arises from the use of signs. Given the very broad notion of
sign, this definition of pragmatics includes many of the phenomena
studied in psychology and the social sciences. Hence , we see how Morris
is able to bring much of human behavior within the purview of semiotic ;
it falls within pragmatics .
Let me summarize the argument to this point . We have seen that
Morris 's notion of semiotic is very wide , encompassing all of science,
including the physical and the human sciences. The sciences fall within its
range, because on Morris 's view scientific theories are languages- thus ,
systems of signs. As such, they are to be studied by semiotic , which
Morris considers to be metascience . As we have seen , Morris holds that

[s]cience, as a body of signs with certain specific relations to one another , to

objects and to practice , is at once a language , a knowledge of objects and a type
of activity ; the interrelated study of syntax , semantics and pragmatics of the
language science in turn constitutes meta science- the science of science. ( 1955, 70)

It does not follow from the inclusion of the study of science within semi-
otic that the phenomena that the scientist studies are themselves signs,
in Morris ' s sense of 'sign' . For example , the solar system itself is not a
system of signs; rather , the theory that describes it is supposed to be such
a system and thus , open to study by semiotic . There is another sense in
which certain theories in the human sciences fall within semiotic . Morris

claims that many of the actual phenomena studied by the human sciences
are themselves systems of signs; for example , various sorts of ritual
 156 Da vis

behavior that anthropologists and sociologistsstudy fall within Morris 's

view of semiotic. In fact, Morris thinks that the reach of semiotic is very
broad: " Psychology and the human social sciencesmay find part (if not
the entire) basis of their distinction from biological and social sciences
in the fact that they deal with responsesmediated by signs" (1938, 56).
Hence, the range of semiotic for Morris is wide in two ways. First, it
includesall the sciences , sincethey themselvesare systemsof signs. Second,
it includes the human sciencesand psychology, since supposedly what
they study is human behavior mediated by a systemof signs.
I now turn to some criticisms of Morris 's wide view of semiotic. First,
attempts at achieving a linguistic unification of the scienceshave been
completely abandoned. One of the main doctrines underpinning this sup-
posed unification, the verificationist theory of meaning, and its correlate,
the verificationist theory of meaningfulness, have been shown to face
insurmountable problems (Hempel 1965, 101- 120). Moreover, it is now
thought that there is no common logical and semanticstructure to scien-
tific theories in the physical and human sciencesthat can throw any light
on the epistemological questions that motivated the logical positivists.
One of the central issuesfor the logical positivists was the problem of
demarcating sciencefrom nonscience. On one side of the line was what
humans can know; what was on the other side of the line was not only
beyond human knowledge, but was literally nonsense(Ayer 1946, 5- 15,
35- 44). The collapse of the verificationist projects showedthat there was
no such demarcation to be made. In addition, recent work in the philos-
ophy of the human scienceshas shown that its theories are shot through
with normativity . The normativity arises because of the ineliminable
necessityfor theories within the human sciencesto attribute to human
activity meaning that involves the construction of theories of interpreta-
tion that assign meanings to human activities. As a result, there is no
possibility of unity with the physical sciences , the theories of which are
devoid of theories of interpretation that assignmeaning to their objects
of study (Davidson 1984, 125- 141). Any attempt to bridge the gap-
currently fashionable connectionist theories, for example (Churchland
1984, 43- 5I )- ends up removing meaning from the range of phenomena
for which it must account and, with that, removing from explanatory
reach what the human scienceswished to explain in the first place. Even if
we restrict our domain of unification to the physical sciences , the current
view is that if such unification is to come about at all, it will come about
by providing a theoretical unification of different domains of inquiry .
Unity of Science 157

Thus, to unify , say, somebranch of biology (e.g., genetics) and chemistry

it must be shown that geneticphenomenain biology can be explained by
chemical theory. This sort of unification is not linguistic; instead, it
requires well-worked-out theories in microbiology and corresponding
experimentalresults. An exampleof suchunification is the discoveryof the
chemical structure of DNA . Hence, semiotic cannot provide a unification
of the sciencesby showing that scientific theories share common formal,
semantic, and pragmatic properties. Second, there is no evidencethat the
various phenomenathat Morris describesas involving semiosishave very
much in common that is of theoretical interest. Morris holds that scien-
'tific theories, natural languages, dance, rituals, gestures, body language,
facial expressions , and the like, can be accounted for by semiotic. He
claims that they all involve the relations among signs, designatum, inter-
pretant, and interpreters. Work in linguistics in the last 30 years shows-
conclusively, I believe- that natural human languages are not sets of
signs, in Morris 's senseof 'sign'. For Morris , as we have seenif a mor-
pheme is to function as a sign, it must be connectedto some object or
other. But there are many morphemes, like syncategorematicexpressions ,
for which no such connection is possible; and there are expressions , like
speechact verbs, that are connectednot to objects but to sets of rules.
There might be some extendeduse of 'sign', other than Morris 's, so that
we could count human languagesas setsof signs, but this use would be
so broad that it would be of little theoretical interest (Sperberand Wilson
1995, 3- 9). Last, and most important , Morris includes too much within
semiosis, especiallywithin pragmatics; as a result, it is impossibleto have
a coherent theory that could give a theoretically unified account of the
disparate phenomenathat he claims fall within the domain of semiotic.
This is one of the reasonsthat some pragmaticians who have followed
Morris 's lead now claim that pragmatics, which is only a subdomain of
semiotic, is not a theory on a par with syntax, but a discipline, or even
more widely yet, a perspectiveon language. I would like to suggestthat
Morris 's wide view of semiotic, and especially of pragmatics, be aban-
doned, since the underlying reason for it , the desire to have a unified
science, no longer can be sustained.2
Let me close by saying something about the way I think research
should proceed in pragmatics. Theories, perspectives , or disciplines that
cover everything explain nothing; unconstrained theories that attempt to
account for everything will not tell us anything interesting about natural
languagesand human communication. What is neededare theories that
 158 Davis

give explanations for a restricted range of data and that are constrained in

various ways by the data they are supposed to explain and by the ordi -

nary norms that govern scientific theorizing , including simplicity , rele -

vance , explicitness , and preciseness .

Notes

1 . There is some variation in the term used for the study of signs . Morris uses

semiotic , following Locke and Pierce ; Margaret Mead uses semiotics ; and Roland

Barth , following Saussure , uses semiologie , which is translated into English as

semiology . ( For a discussion of these variants , see Morris ' s quotation from Sebeok
in Morris 1971 , 9 .)

2 . It might seem that I am beating a dead horse and that Morris ' s wide view of

pragmatics is only of historical interest . I began the chapter by citing the under -

lying perspectivalism that informs the organizing principles of the International

Pragmatics Association . It is shared by a number of researchers who work on

pragmatics , including Mey , Ostman , and Verschueren ( see Mey 1993 ; Ostman
1988 ; Verschueren 1987 ).

References

Ayer , A . J . ( 1946 ) . Language , truth , and logic . London : V . Gollancz .

Carnap , R . ( 1934 ) . The unity of science . London : Kegan Paul , Trench , Trubner .

Churchland , P . ( 1984 ) Matter and consciousness . Cambridge , MA : MIT Press .

Davidson , D . ( 1984 ) . Inquiries into truth and interpretation . New York : Oxford
University Press .

Hempel , C . ( 1965 ). Empiricist criteria of cognitive significance : Problems and

changes . In Aspects of scientific explanation and other essays in the philosophy of

science . New York : Free Press .

Joergensen , J. ( 1951 ). The development of logical positivism ( The international

encyclopedia of unified science , vol . 2 , no . 9 ) . Chicago : University of Chicago Press .

Mey , J . L . 1993 . Pragmatics . Cambridge , MA : Blackwell .

Morris , C . W . ( 1936 ). Remarks on the proposed encyclopedia . In Actes du

Congres International de Philosophie Scientifique II Unite de la Science . Paris :

Hermann .

Morris , C . W . ( 1937 ) . Logical positivism , pragmatism and scientific empiricism .

Paris : Hermann .

Morris , C . W . ( 1938 ). Foundations of the theory of signs . ( The international

encyclopedia of unified science , vol . 1 , no . 2 . ) Chicago : University of Chicago

Press .

Morris , C . W . ( 1955 ) . Scientific empiricism . In Encyclopedia and unified ..\'cience

( The international encyclopedia of unified science , vol . 1 , no . 1 ) . Chicago : Univer -

si ty of Chicago Press .
Unity of Science 159

Morris, C. W. (1971). Writingson thegeneraltheoryof signs.TheHague: Mouton.

"
Ostman , J.-0 . (1988 ). Implicit involvement in interactive writing. In J.
Verschueren and M. Bertucelli -Papi (Eds.), Thepragmaticperspective: Selective
papersfrom the 1985 InternationalPragmaticsConference . Amsterdam : John
Benjamins .
Sperber, D., and Wilson, D. (1995 ). Relel-'ance: Communicationand cognition
,
Cambridge , MA: Blackwell.
Verschueren , J. (1987
). Pragmatics asa theoryof linguisticadaptation. In Work-
ing documents # 1. Antwerp: InternationalPragmatics Association
.
Chapter 9

Scientific Theories That Leslie Margaret Perrin

Unconceal Being : Intentions McPherson
and Conceptions in Their
Genesis

You must inquire into all, both the untrembling heart of well-rounded uncon-
cealment and the opinions of mortals, in which there is no unconcealing assur-
ance. Yet nonethelessthese too you shall learn, as the seemingthings one must
intend to prove to be (though, through it all, all of them transgressall assurances
),
becauseall are pervading everything.
The Goddessof Day to Parmenides

When John Macnamara agreed to be my doctoral supervisor , I had no

idea what a joyful graduate experience awaited me , for I had never met

him . But I had two good reasons for wanting him to supervise my grad -

uate work . First , out of all the psychological theories on language and

cognition , his were the only ones that really made good sense to me . His

theories mirrored my own experience of my own mind - something few

other theories did . Second , when I asked around to find out what his

former graduate students thought of him , I was told , " They adore him ."

I was intrigued .

When I began my studies at McGill , I quickly learned why John 's

students adored him . I discovered that he had a brilliant mind that

This work was inspired to an immeasurable degree by Robert H . Schmidt , who

always seems to point me in the right direction . Many of the insights are taken

wholly from him , including some of the insights regarding the original senses of

scientific concepts , which he earned through years of contemplation and research .

I acknowledge also his help in pointing me to certain important passages , in vari -

ous old works , relevant to changes in these concepts . Further , his comments and

insights regarding the ideas in this chapter proved invaluable . I am extremely

grateful to him . This chapter benefited also from useful comments from David

Alexandre Nicolas , Ray Jackendoff , and Keith Niall . And I thank my dear

departed friend and mentor John Macnamara for creating such beautiful theories

that I could not help but ask myself how he did it .

162 McPherson

instinctively aimed at only the most fundamental questions and pene-

trated straight into the core of them. He had a profound understandingof
critical philosophical and psychological problems, and he shared his
knowledge and insights with his students generously. To listen to him
speak was as entertaining as reading his wonderfully elegant prose. But
besidesall this, he was sweet, loving, kind, wise, charming, playful, and
funny- a delightful gentleman. I quickly cameto count myself among the
studentswho adored him.
It took me considerably longer to discover why his ideas and theories
make so much sense. I believe I can now explain why. This chapter is
devoted to this effort. I hope to show why the approach John took to
theorizing results in superior theories, and why the usual approachesdo
not. Though I focus on theoriesof natural languagesemanticsearly in the
discussion, the insights should apply to theory in any domain of inquiry .1

9.1 Theory

The word theory derives from the Greek theoria, 'a viewing' or 'a
beholding'. Ideally, a theory should result from a theorist's viewing or
beholding (or noetic perception) of being,2 and it should permit others
to behold that being. At a minimum, this implies that any conceptguiding
a theorist's thinking and transmitted to others along with a theory must
originate in an authentic and relatively complete conception of being of
some sort, and that the theory must be compatible with each such con-
cept. What a theorist must do (and avoid doing) to ensuretheseoutcomes
will becomeclear.

9.2 Intentions in Theories

In the context of theories or, more generally, declarative statementsin

somelanguage, an " intention" (from the Latin intentio, 'tension' or 'aim')
is a noetic tensing or aiming of the mind toward being of some sort, an
aiming concomitant with speaking or writing , and brought about by
hearing or reading, a string of signsfor which one knows the conventional
(or theorist-defined) interpretations.3 I take this to be the original senseof
intention in the context of signification; be warned that it bears little
resemblanceto modern sensesof the term (e.g., Bertrand Russell's " propo-
sitional attitude" understood as a relation of a person to a proposition,
suchas the " relation" of belief, where no relation of the personto being is
Scientific Theories That Unconceal Being 163

necessarilyimplied; for instance, those who advocate " methodological

solipsism," suchas Fodor (1980), conclude, from the fact that false beliefs
can affect a person's actions, that a propositional attitude entersinto the
explanation of behavior under an opaque construal, which is considered
to be the only construal relevant to psychological theories becausethe
explanation of behavior is now typically said to be the goal of psychol-
ogy- a naturalistic pretenseof the behaviorist and postbehaviorist eras).
The original senseof intention is more similar, in some respects, to the
modern senseof attention.
In describing the object of an intention as " being," I do not mean to
imply that its object must be " objective reality," or being as it is carved up
by a god or nature, independently of our characterization of it (e.g., a
" natural kind " ); I am not presupposingthe correctnessof what Putnam
(1981) calls the " externalist" perspective, according to which being and
beingsare necessarilydefinedor individuated objectively, and referenceto
anything extrapsychic is necessarilya connection of some sort to objec-
tively defined being or beings. Theories- especially in psychology, the
social sciences , mathematics, and semantics- often require that being
be carved up or defined in other ways (e.g., the firms, consumption, and
income of economictheory). Being that is carved out mentally by us is not
imaginary or unreal just becauseits definition has no " objective" (i .e.,
nonsocial and nonpersonal) source; as Field (1982) argues, " what we have
carved out existed independent of our carving it out, despite its non-
saliencefrom the God's-eye perspective" (p. 561).4

9.3 A Defense of Intentions

The theses of this chapter presupposethe existence of intentions qua

intentions. But the supposition of their existenceis surprisingly contro-
versial these days. So, with the aim of reducing the appeal of modern
skepticism about them, and to show why scientific theorizing would be
impossiblewithout them, I will examinesomeclaims that have beenmade
about their nonexistenceas well as someof the attempts to reducethem to
somethingelse.
The physicalist philosopher Daniel C. Dennett and his many followers
would have us regard the intentions we ascribeto ourselvesand others as
" idealized fictions in an action-predicting, action-explaining calculus"
(Dennett 1978, 30), and our ascription of them as a " stance" we take in
relation to an organism or a machine when that stance seemsuseful in
164 McPherson

explaining or predicting actions or behavior , but one irrelevant to scien-

tific theorizing ; talk of reference, thoughts , beliefs, perceptions , desires,
fears, aims , and so on , is " a sham and . . . mere word play " (Haugeland
1982, 616; Dennett claims that " intentional theory is vacuous as psychol -
ogy " (p . 15)) . But Dennett owes us an answer to this question : how can
one adopt the intentional stance, or any explanatory stance, toward
something 's actions without coming into any (genuine) intentional states
such as perceiving actions , desiring or trying to explain them , and believing
or asserting that their explanation lies in , say, beliefs and desires? (See also
Slors 1996.) As Putnam ( 1989) observes, " Explanation is a flagrantly in -
tentional notion " (p . 217).
Some physicalists (e.g., Churchland ( 1981)) would have us embrace the
goal of eliminating all talk of intentions - deemed unreal - from a science
of the mind . But it seems to me foolish to deny the reality of that which
one experiences directly in every waking moment just because averting
one' s eyes eases ontological queasiness. As Silvers ( 1991) remarks ,
" Eliminativism with respect to intentional phenomena is a way out but it
is much like the way of the ostrich " ( p . 68). If that which manifests itself
to us does not accord with what we know about nature , then our knowl -
edge of nature must be less than complete ! The fact that intentional
notions such as reference resist naturalistic explanation merely shows, as
Putnam ( 1992b) explains , " the limitations of a certain notion of 'nature '
and a certain notion of 'explanation ' " (p . 384). Instead of denying the
manifest , one might fruitfully aspire to what James Clerk Maxwell ( 1877)
called " that state of thoroughly conscious ignorance which is the prelude
to every real advance in knowledge " ( p . 245) .
It is easy to show that eliminativism is untenable . The eliminativist

cannot even state his or her thesis without making covert use of inten -
tions . When Churchland ( 1984) claims that the commonsense (or " folk
psychology " ) conception of the mind as intentional " is a false and radi -
cally misleading conception of the causes of human behavior and the
nature of cognitive activity " ( p . 43; emphasis added), he uses the essen-
tially intentional notions of falsity and a conception . When he says of folk
psychology , considered as " an empirical theory ," that " its ontology is an
illusion " ( 1981, 72; emphasis added), he implies that theories are about
(illusory or real) being and therefore intentional ; and since an illusion is a
type of intentional state (i .e., perceiving something as other than what it
is), one cannot sensibly dismiss the intentional states in the ontology of
the folk theory as illusions . Eliminativists cannot even maintain , without
Scientific
Theories That Unconceal Being 165

inconsistency , that their goal is to eliminate all mention of intentions from

scientific discourse because they are fictions , for goals are intentional .
Garfield ( 1988) sums up the problem :
The eliminativist enterprise is, in a straightforward sense, self-refuting . It commits ,
as Baker [( 1987)] has put it , " cognitive suicide " by denying what is necessary in
order to give it sense, viz ., that those who assert it believe its truth (or , for that
matter , anything at all ) or mean anything by asserting it . ( p . 123)

The whole of the scientific enterprise involves goals (e.g., the goal of
generating hypotheses from a theory , and the goal of obtaining evi-
dence in support of those hypotheses) and intentions of other kinds (e.g.,
desires for procedural rigor in experiments , perception and a belief in the
veracity of one' s sensesduring observation , and assertions about experi -
mental results) ; as Baker (1987) argues, " Without intentionality science
would be impossible " (p . 173; see also Garfield 1988, 112- 116). More -
over , I cannot see how a scientific theory could be meaningful - some-
thing other than a word game or a kind of dream - unless it permitted the
mind to intend being .
To make sense of what follows , the reader would do well to resist

another (even stronger ) modernistic temptation , namely , to try to under -

stand signification , not in terms of intentionality , but in terms of the
popular modern concept of a " mental representation ." Having changed
radically from the " representation " of German philosophers such as
Kant and Brentano (i .e., their Vorstellung , the act of presentation to
or placement in front of the conscious mind ), a representation is now
understood as a mental being (e.g., a mental symbol ) to be operated upon ,
one unavailable to consciousness, in the context of the computational (or
" information -processing" ) view of the mind . As Searle ( 1980) has dem-
onstrated with his " Chinese room " argument , operating upon symbols in
conformity with formal rules (i .e., rules that depend upon the shapes of
the symbols ) does not lead to any understanding of the symbols . Fodor
( 1986) puts it this way : " You can't get semantical properties out of
symbols just by piling up their syntactical ones" (p . 13). More to the
point , no existing theories of mental representations succeed in reducing
intentionality to something else, I will argue, or allow for the mind to
make the kind of contact with being necessary for scientific theorizing to
proceed.
Sometimes mental representations of linguistic meaning are imagined
to be symbols or symbolic formulae (in a mental dictionary , lexicon , or
encyclopedia , or in some representational " space" ) in some partially
166 McPherson

or entirely unlearned language (e.g., " Mentalese" or " the language of

thought" ). This view leads to a problem that Hamad (1990) calls the
" symbol-grounding problem" and that Johnson-Laird (1988) calls the
" symbolic fallacy" : the interpretations of symbols cannot be just more
symbols.
Some theorists seem to be blissfully unaware of the problem or
unwilling to recognizeit (e.g., Eco (1988); Schankand Kass (1988); Wilks
(1988)), or seemto think it can be solved by some sort of " encoding" of
semanticcontent by symbols in a foffi1al system, but leave in shadow the
mysteriesof the encoding and decoding (e.g., Pylyshyn (1984, 258)).
Among those who recognizethe problem and make a serious attempt
to find a solution to it , some grasp at the slenderhope that the interpre-
tation of a sign or symbol is a causal connection of some sort (e.g.,
Dretske (1981); Stampe (1979)). Fodor (e.g., 1987), for instance, tries to
explain semanticcontent in terms of the reliable causationof tokenings of
a certain symbol by perceptsof a certain kind, which are in turn caused
by observableproperties of beings of a certain kind; the covariation of
tokens of a symbol and the referents that cause them is supposed to
deteffi1inethe interpretation or content of the symbol- presumably even
when a referent is not presentto causeits tokening. This seemsto be not
so much an explanation of interpretation as a change of topic; it seems
plausible only as a partial account of why we tend to think " There's
something red there" when a red surface fills the visual field (but a
full account of that would require more, including accounts of how the
English word red came to be associatedwith red things, how one takes
one's perception of rednessas evidence for something red- cf. Silvers
1991- and how one then interprets " something red" into it ; all of these
accounts would surely involve intentions). Moreover, the causesof to-
kenings of a sign (and of beliefs about its referents) need not be referents
of the sign (or of any phrase headed by it), as when one learns about
electrons from textbooks; and though some causal chain may link elec-
trons to tokenings of electron immediately " caused" by textbooks, that
causal chain could not be specifiedunlessone already knew the referents
of electronsand could refer to them by someother means(Putnam 1981).
Contrary to the assumptions of those who espousecausal theories of
semantic content, causesand causal chains are not purely objective and
unique things. A cause or a causal chain leading to some event is not
deteffi1inedindependently of observers; an observermust decide, out of
the entirety of being- all of which somehowinfluencesthe event- which
Scientific Theories That Unconceal Being 167

things are the most important to mention in giving an explanation in

line with his or her current purposes (Putnam 1984). Causal accounts
of meaning also leave mysterious the way in which the expressioncausal
chain comes to have its content (Putnam 1984). And attempting to
explain the content of individual words in terms of causal connections
doesnothing to explain the content of complex referring expressionssuch
as thefirst baby born after the year 2500 that have referentswith which
one cannot possibly now interact causally (Putnam 1992a, 163). But per-
haps the most serious problem with causal theories like Fodor's is that,
evenwhen the referent is present, the being that causesthe sensorystimu-
lation that is supposedto lead to the tokening of a symbol is rarely being
that the symbol can be usedto signify. The being that stimulatesthe retina
so as to lead to a tokening of person, namely, parts of the exposedsurfa-
ces of a person's clothing and of the body of matter currently coincident
with the person, is not equivalent to the referent of any noun phrase
headedby the noun person. Causal accountsof referencewill only seem
plausible to those who unwittingly presupposethe intentionality such
accountsare supposedto explain (cf. Searle1980, 454).
Others hope that symbols can be grounded by nonsymbolic repre-
sentations- iconsor images (e.g., Hamad (1990); McGinn (1989)) or
mental models (e.g., Johnson-Laird (1988)) that are supposedto represent
by means of similarities, (natural, not assigned) correspondences, or iso-
morphisms with the being represented, and to intrinsically preserve
information about it .
Sometimessuch representations- or even symbolic representations-
are imagined to give signsmeaning in and of themselves. But this would
imply that all statements, including theoretical statementsin science, are
about mental representationsrather than being- which would make psy-
chology the basic science(as in T16n, the imaginary world invented by
Jorge Luis Borges, one populated entirely by idealists); as Macnamara
once remarked, " Flattering to a psychologist, but dubious!" (Macnamara
1989, 350). Moreover, the idea simply makes no sense. The fact that
we can discusswhether or not statementsput the mind in contact with
something outside the mind presupposesour ability to refer to the
extrapsychic. Macnamara (1989) made the point this way:
Onemight ask: Doesthe expression 'extramentalreality' put us in contactwith
somethingoutsidethe mentalmodels ? If it does, why do otherwordsnot? If not,
whatmightthe expression mean? And howcouldany of usconceivetheexistence
of an extramentalreality, let alonewallowin its exuberantpresence ? (p. 352)
168 McPherson

An inability to refer to the external would also make obscure the meaning
of an indexical or demonstrative such as this, that , here, or there (espe-
cially in discourse contexts in which it had no antecedent) and would
leave us with no way to account for its obvious kinship with (and frequent
concomitance with ) an indexical gesture such as pointing , nodding , or
gazing toward something so as to indicate to another person an object of
thought ; one does not (and cannot ) point , nod , or gaze in the direction of
one's own mental representations . It is apparent that , as Putnam ( 1973)
says, " 'meanings ' just ain 't in the head" ( p . 704).5
Sometimes some relation of a nonsymbolic representation to being is
imagined to ground or give meaning to a symbol or formula . But theorists
fail to characterize this relation in any way that could explain reference or
interpretation , or that would permit scientists to theorize about the being
represented.
This relation to being is most popularly described as a map or function
from the representations to states of affairs in the extrapsychic world (e.g.,
Millikan 1984). But no such function could explain reference because it
presupposes reference. .Unless the mind already had the capacity to refer ,
such a function could not be computed in the mind because, though the
arguments of the function (representations ) could be in the mind , its
values (extrapsychic states of affairs ) could not .
Some (e.g., Palmer ( 1978) seem to think that the mere (objective )
existence of a similarity , correspondence , or isomorphism will , by itself ,
do the trick - that such a correspondence with being is what we really
mean by reference or intention (or at least representation). But that is not
enough . Mental images or models could not , just by means of corre -
spondences, determine what signs signify . For one thing , they would leave
the signified indeterminate because a given one is similar to , corresponds
to , or is isomorphic to many beings or states of affairs , and is similar to ,
corresponds to , or is isomorphic to a given being or state of affairs in
many different ways, " representing " the truth -conditions of many differ -
ent sentences and the referents of many different signs (see Cummins
1996; Fodor 1990; Putnam 1981) . The correspondence to the signified
being that is relevant to reference and truth -conditions would have to
be singled out somehow . It could not be singled out by stipulation (cf .
Morris 1991), since stipulating that the correspondence between this and
this or stipulating that this correspondence between a representation and
being of a given kind is the relevant correspondence presupposes reference
(to the representa ti on , the represented being , and , perhaps, the corre -
Scientific Theories That Unconceal Being 169

spondence ), which is the very thing that correspondenceis supposedto

explain (seePutnam 1981). Nor could any causal connection fix the rele-
vant correspondence(see Fodor 1990; Putnam 1984). But even if one
correspondencecould somehow be singled out by some nonintentional
means, still more would be neededto give rise to representationof, ref-
erence to, or an intention toward being. A given correspondence(or
similarity, or isomorphism) does not a representation make (see, e.g.,
Pylyshyn 1984, 40- 41). One thing can representanother only by virtue of
an intention to represent(seeFodor 1990; Putnam 1981; and so it is fool-
hardy to try to reduce intentions to representations ) . The moon may
happen to look like a man's face, but it would be odd indeedto say that it
is a representationof a man's face; as far as we know, nobody shapedthe
surface of the moon so as to create the likenessof a man's face (or, for
that matter, the likenessof a rabbit, which somepeople seein the moon).
As Putnam (1981) points out, " Thought words and mental pictures do
not intrinsically represent what they are about" (p. 5); " believing that
some correspondenceintrinsically just is reference (. . . as an ultimate
metaphysical fact) amounts to a magical theory of reference" (p. 47)-
that is, " a theory on which certain mental representationsnecessarilyrefer
to certain external things and kinds of things" (p. 15).
A thinking subjectmust do somethingwith , say, a mental icon if it is to
bring the mind into contact with being. Three possibilities occur to me
(but all of them presupposeintentionality). The subjectcould interpret the
icon as (1) a linguistic sign of a specialsort, one that happensto look like
the being it signifies(comparable to early Sumerian cuneiform ideograms
and early Egyptian hieroglyphs), in which case it is redundant; (2) an
image of the being pictured; or (3) a representative, proxy, or counterpart
for the being represented(seealso Cummins 1996), somethingthat stands
in for it in mental activities, just as a map can stand in for a region in
planning a route of travel. The latter possibility is pertinent to mental
activities such as planning actions in the imagination, but it could not
provide a plausible generalaccount of interpreting signsor strings of signs
in sentences . An icon (one available to consciousness ) interpreted in the
third way may playa role in signification when there is no particular or
definite referent for a word or phraseembeddedin a sentenceso that the
mind cannot fall upon extrapsychicbeing in interpreting it , as when one is
interpreting an indefinite noun phrase in the predicate of a denial (e.g.,
a ball in Rick did not hit a ball, which obviously does not refer to any
particular ball, sinceit makes no senseto ask, " Which ball?" ; seeGeach
1962; McPherson 1995, app. B) or in the predicate of a sentencesuch as
170 McPherson

Youwill needan umbrellatoday(wherethe lack of a definitereferentof

an umbrellais evident, again, in the inappropriateness of the question
"Which umbrella?" ); in such cases , the mind may aim at an icon
understoodas a representative , proxy, or counterpartfor any individual
whatsoeverof the kind. If a representation is interpretedin accordance
with oneof the first two possibilities , asan imagelikesignor an image, it
may playa mnemonicrole, helpingthe subjectto rememberthe sort of
beingthe associated linguisticsignconventionallysignifies . But in any of
thesecases , we would not want to say that the icon determines what is
signifiedsothat theassociated sign(or a referringexpression headedby it)
signifiesjust those aspectsof being that are represented . That would
detenninethe beingsignifiedtoo narrowly. Representation by meansof
an icon at modelis alwayspartial (seealsoCummins1996 ), and to think
that only what is represented is signifiedwould be, in the caseof theoret-
ical terms, to missthe wholepoint of scientificinvestigation : to discover
or unconcealthe unknownaspectsof the natureof being. Moreover, the
historyof sciencedemonstrates that a word canbe usedto referto some-
thing evenwhenit hasbeenconceivedwrongly. Science couldnot proceed
if scientistscould not usea scientifictenn suchas electron , gene, or gold
to referto beingof thesamekind whiletheyentertaineddifferentconcepts
of it and differenttheoriesabout it (seealso Putnam1988 ) . (Of course ,
the current conceptcannotbe completelymistaken. One could not use
electronsto referto electronsif one's conceptof their naturewereso far
off that one did not evenknow that they are subatomicconstituentsof
matter, for instance . And thoughonecanreferto somethingnot correctly
conceived , its naturecannotyet be fully revealedto and beheldby one's
mind; full unconcealment would leadto the conceptionof a new, correct
concept.) The potential referentsof a sign are fixed by a convention
accordingto whichthe signis usedto referto, or aim the mind at, being
of a particularsort. The beingtoward which a linguisticsigncan direct
the mind is determined , not by a similarity, a naturalcorrespondence , or
an isomorphismbetweenany associatedmentalrepresentation and the
being, but by a conventionof a speechcommunity(seealsoAristotle, On
Interpretation2, 16a19).

9.4 IntentionalObjectsandSemanticTheory: Macnamara

's
Contemplationof Being

The nature of thinking can be seenonly by looking away from thinking .

Martin Heidegger, Discourseon Thinking
Scientific Theories That Unconceal Being 171

If we take the interpretation of signs or combinations of signs in sentences

to be the objects of the intentions to which the signs give rise, then a
theory of the interpretation of sentences- a theory of semantics- must
characterize or define the objects of the intentions that arise upon hearing
or reading sentences (where object is to be understood , here and else-
where, as that toward which the mind is aimed , which is not necessarily
an object in the sense of a sensible thing ) . A semanticist must characterize
or define the being signified ; to do so is to define what is meant . This is
just what John Macnamara did . He did not merely name the objects of
intentions (e.g., the word John is interpreted into John ), for to do so is
to leave their definition implicit ; he characterized or defined the being
toward which the mind aims . In his theory of the semantics of proper
names , for instance , he identified the referent of a proper noun as an
individual individuated or informed by a kind (or form or species),
some psychologically (and perhaps ontologically ) privileged kind such
as PERSON - as opposed to an individual individuated by some other
kind (e.g., WOMAN , FRIEND , DANCER , or PASSENGER ), and as
opposed to , say, a particular collection of molecules (which would not
permit the application of a proper name both to a given woman and to
the infant she once was), or a set of predicates or properties (for the
properties of the bearer of a proper name change), or something that gives
rise to a certain appearance (for that which bears a proper name- say, a
person- can change radically in appearance- for example , from infancy
to old age, through changes in hair color and hairstyle , and through
changes in clothing , which alter the appearance radically but do not
change a part of the person proper ), or that appearance itself (for that
which bears a proper name may well be largely hidden , as a person
beneath his or her own skin , and usually under clothing as well ; an
appearance does not itself bear a proper name , but announces the pres -
ence of that which does) . The identity of the named individual across
factual and counterfactual situations is traced through the individuating
kind (or form ) . (See La Palme Reyes, Macnamara , and Reyes 1994; La
Palme Reyes et al . 1993; Macnamara and Reyes 1994.)
John arrived at definitions of intentional objects by " carefully sifting
through " his own intuitions , as he put it . He would find meaningful
utterances that seemed to reveal something about the interpretation of a
sign or phrase and contemplate the nature of the being into which it is
interpreted . He once described the process to me by comparing himself to
a worm turning over the same piece of earth again and again . The anal -
ogy is apt . To turn over earth , a worm must make direct contact with the
 172 McPherson

earth. In his contemplation, John seemsto have kept his mind in direct
intentional contact with the objects of linguistic intentions. Those who
knew John will know that the earth he turned over most frequently was
the being intended in interpreting the sentenceFreddie is a dog. The late
Freddie was a large black poodle who shared his life with John and his
family . The fact that the sign Freddiecan still be usedto pick out a certain
dog, who is no longer with us, as well as a certain pup who was born in
Montreal, that it can be used to pick out the samedog even in counter-
factual situations (such as one in which Freddie was adopted by other
humans), and other such facts led John to his theory about the nature of
the being into which a proper name is interpreted. The theory accounts
well for the facts of interpretation for proper nouns, and reveals or
unconcealsor brings to light the nature of the being implicitly and sponta-
neously understood to bear a proper name in every instance of the inter-
pretation of a proper name.
John's choice of a declarative sentence(Freddie is a dog) as the focus of
his contemplation may have facilitated his discovery of the nature of that
being. A declarative sentenceis the only type of utterance that can be
judged true or false, and, when true, it unconcealsbeing of some sort to
the mind (which is why the Greeks used the word aletheia, 'unconceal-
ment', for truth ; see Aristotle , On Interpretation 1, 16a9- 18; 4, 16b33-
17a4); and so a true declarative sentencecan bring the mind into direct
intentional or noetic contact with being. Contemplation of an isolated
word may not permit such contact, for a word by itself may not actually
signify; a word uttered or written in isolation, such as " person" or
" green," doesnot seemto aim the mind at any definite or particular being
or beings; it may only potentially signify (i.e., have the potential to signify)
by virtue of one's having conceiveda concept of, or formulated a defini-
tion of, the kind of being it would actually signify in a referring expression
in accordancewith a convention of a speechcommunity. (And this may
explain the Greeks' use of the word dunamis, 'power' or 'potentiality', for
the force of an individual word.)6 When a word is out of context, one
cannot even determine on which side of the use/mention distinction it
falls, that is, whether it is to be interpreted in its signifying function or as a
string of soundsor letters (La Palme Reyeset al. 1993).

9.5 Inner Perceptionand PsychologicalMethod

John came to understandhis sifting of intuitions largely in terms of Franz

Brentano's " inner perception" (innere Wahrnehmung ; see, e.g., Macnamara
Scientific Theories That Unconceal Being 173

1994) . Brentano ( 1838- 1917) was, along with Wilhelm Wundt , one of the
first empirically minded psychologists . Neither Wundt nor Brentano
advocated the use of the methods of the natural sciencesin psychology (see
Macnamara 1992, 1993) . Brentano advocated the use of inner perception
as the proper empirical technique for a genuine science of psychology (see
Brentano 1924) .
Brentano maintained - and quite rightly , I believe- that all and only
psychological phenomena are intentional ; in any psychological phenom -
enon, the mind is turned toward - the attention is focused on- something
as its object . Psychology , then , is the science of the intentional . Brentano
argued that one can study intentions directly only by means of inner per-
ception , which he described thus : 'Only while one's attention is turned
toward another object , it happens that one also incidentally arrives at the
perception of psychical events that are referred toward it [i .e., the other
object ]' ( 1924, 41) .7 In the context of semantics, this means that when
a sign or combination of signs makes one' s mind intend something as
its object , one can perceive or notice or become aware of aspects of the
intention such as the conditions under which it arises , its character , and

the nature of its object .

Brentano took great pains to distinguish inner perception from intro -
spection or inner observation, which he regarded as impossible (see Mac -
namara 1992, 1993) . One cannot observe one's own intentions , for to do
so would require a redirecting of one's attention - of one' s mind ' s firm or
taut aiming or tensing- away from the objects of intentions and toward
the intentions themselves, which would abate or relax in one' s so doing ,
ceasing to be; but one can notice characteristics of one's intentions inci -
dentally as they occur (or , as Aristotle put it , en parerg6i , 'in by-work ' , or
'in a deed done on the side' ; Metaphysics A .9, 1074b36). One can in some
sense observe one's intentions only when they have passed out of exis-
tence, through contemplating memories of them .
The science of psychology that Brentano envisioned , though founded
on inner perception , was not fettered by the fact that one can perceive
directly only one's own intentions . Brentano recognized that we can learn
about the intentions of others through their reports about them or evi-
dence of them in actions or in facial expressions or other involuntary be-
haviors . Such indirect means of learning about intentions can be the basis
of psychological experimentation . But a theory that generates experi -
mental hypotheses can only be a genuine theory of , or permit a genuine
beholding of , intentions of some sort if the theorist uses inner perception
to gain direct evidence of the nature of those intentions - for no other
174 McPherson

direct meansof discovering their nature exists. It is the discovery of their

nature by this direct meansthat permits an interpretation of indirect evi-
denceof intentions such as facial expressionsand actions; no knowledge
about intentions could be acquired on the basisof the indirect evidenceby
itself, for , as Brentano (1924) pointed out, 'what is a perception, what a
judgment, what joy and sorrow, desire and aversion, hope and fear,
courage and despair, what a decision and an aim of the will , we would
never come to know if the inner perception of our own phenomenadid
not bring this knowledge to us' (p. 40). Psychologistscannot escapethe
needfor inner perception.
I turn now to an examination of intentions and conceptions that will
reveal someof the reasonsa theory can fail to pennit a beholding of being
when the theoretical work behind it is done without the theorist's keeping
his or her mind firmly aimed at the being of interest.

9.6 First IntentionsandSecondIntentions

Following the Scholastics, philosophersdistinguish " first intentions" and

" secondintentions," or conceptsconceivedby direct intentional contact
with being and concepts conceived by reflection upon concepts already
conceived. But the distinction seemsoriginally to have been betweentwo
types of intention rather than two types of concept. First intentions
seemto have been understood originally as aimings of the mind toward
'real beings' (entia realis) directly, and secondintentions as aimings of the
mind first toward concepts(conceivedeither first-intentionally or through
reflection upon concepts; conceptsconceivedin the latter way were called
'beings of reason',8 entia rationis) before the mind may be redirected
toward real beings. The distinction has consequencesfor theoretical
work . Theoriesare usually about being of which our conceptsare second-
intentional (i.e., beingsof reason- conceptsoriginally conceivedby reflec-
tion upon concepts), but theoretical reasoningthat proceedswithout the
mind's aiming directly at any extrapsychicbeing can result in theoriesthat
fail to reveal the nature of any such being (partly becauseof the tendency
of second-intentional conceptsto changein senseand becomedepletedin
sense - a problem addressedin the next section).
The original sensesof first intention and secondintention are evident in
expositionsof someof the RenaissanceScholastics.
Jacopo Zabarella (1533- 1589) described the original, etymologically
discoveredsenseof (first ) intention as that action in which 'the soul aims
Scientific Theories That Unconceal Being 175

[or stretches; tendit] toward an object with which it is to become

acquainted [in objectum cognoscendum ] ' (1590- 1605/ 1966, 870). The
notion of an aiming (or stretching) of the soul toward something (some
extrapsychicbeing) as its object was central to the understandingof a first
intention even when the word intentio was being used by medieval Scho-
lasticsjust for aims of the will . Aquinas, for instance, said that 'an inten-
tion, as the very name implies, means to aim [or stretch; tendere] toward
another' (1266/ 1970, 111- 112).
Eustachiusa SanctoPaulo (1573- 1640) pointed out that the name of a
secondintention revealsthe original senseto be an aiming of the mind at
real being secondarilyrather than directly as in a first intention:
A secondintention, if you look at the forceof the name, is that very operationof
the mind by which it secondarilyaimsin the directionof a real thing already
knownpreviously , insofaras it is known. (:SummaPhilosophiae Quadripartita4;
seeGilson 1979, 107)
Francisco Suarez(1548- 1617) gave a more detailed treatment of second
intentions (in De Entibus Rationis, 'On Beings of Reason'):
Secondintentions[i.e., second -intentionalconcepts ] . . . [result] from a second
intentionor attentionor consideration of the intellect, by whichnameis properly
called a reflexiveintellection[i.e., one turned back upon the intellect itself],
becauseit supposes another[intellection ] with which it is occupied[circa quam
versatur ]. . . . Properlythat operationis reflex, and (if I may put it this way) is in
accordance with itself[per se] second , whichfalls uponanothercognition, or upon
an objectaccordingas it is denominated by a prior cognition, and is from that
participatingin someproperties . Thereforebecausethoselogical relationsare
alwaysbasedon reflexcognitionof this kind, for that reasontheyin particularare
calledsecondintentions , or secondobjectivenotions, because theyareobjectsfor
a secondnotion [i.e., acquaintance ] or formal intention. . . . Theserelationsof
reasonare usuallycalledthingsof secondintention. . . because.. . [they have
their] beingonly objectivelyin [i.e., asthe objectof] a secondor reflexivecogni-
tion of the intellect.
From this moreover , it further arisesthat the intellectcould be reflexed[i.e.,
bentback] againupon secondintentionsthemselves , and consideragreements or
differences amongthem, and definethese , or reasondiscursivelyfrom these , and
thenafterthat basesimilarrelationson these ; asfrom genusandspecies it abstracts
the relationof universal , anddenominates that a genus , similarlyalsowith regard
to others; thenthosearerelationsof reasonnow not only by reasonof the other
deficiencies [e.g., not havinganybeingexceptasobjectsof secondintentions ], but
becausethey are not occupied[non versantur ] proximatelywith real things, or
existing[things], and for that reasonwhentheserelationsare saidto be in some
way basedon things, it mustbe understoodto be because of the first relationsof
176 McPherson

this series
. . . . Theserelationscanbe multipliednearlyto infinity by fictions[i.e.,
formationsor inventions ] or reflexionsof the intellect. (1597/1965, 1041
)
A first intention, then, in the original sense, is an aiming of the mind
directly at real being. A secondintention is reflex (reflexus, 'bent back',
directed toward itself ), with the mind turning toward or intending some
of its own contents, namely, concepts; if it turns toward the real being of
which they are concepts, it does so only secondarily. Secondintentions
toward or reflection upon conceptscan give rise to a new concept, which
was also called a secondintention. As Suareznoted, the being of reason
so conceived can then be taken up as the object of further reflexive
intellection to conceive further concepts, as when one conceivesa rela-
tion through reflecting upon the concept of a ratio and conceptsof com-
parative being such as being greater than and being lessthan, where some
of these concepts have themselvesbeen conceived by second intentions
(e.g., the concept of a ratio is conceivedthrough reflection upon the con-
cepts of double, triple, etc.).
How is this distinction betweenfirst and secondintentions relevant to
theorizing? If a being of reason can be the object of an intention in the
reflection that permits new conceptions, then surely it can be the same
in theoretical reflection. Now when a theorist makes use of a second-
intentional concept such as that of a relation, the theorist's mind aims at
this being of reason. If the theorist's mind is to fall upon any real being, it
must then aim toward one of the conceptsthat gave rise to the second-
intentional concept, a concept such as that of being greater than, and it
can only then aim at any being that is an instanceof a relation (just as the
mind, when thinking of the genus ANIMAL , can only fall upon a par-
ticular animal after aiming toward a concept of someparticular speciesof
animal). But the theorist's mind need never be redirected toward any real
being; the theorist may think about a relation without ever thinking about
specificinstancesof it in being, such as the ratio of the magnitude of one
of the long sidesof a certain rectangleto the magnitude of one of its short
sides. Suppose, though, that in someinstancesat least the theorist's mind
doescontinue to be redirecteduntil it comesinto intentional contact with
real being. A conceptis, by necessity, the result of a partial conception (or
a partial taking into the mind) of the real being that gave rise to it , and
especiallyso when the concept is conceivedby secondintentions. When
somereal being is viewed " through" a being of reason or concept (i.e., as
an instance of the being of which it is a concept) , the concept servesas a
kind of filter , focusing the attention so as to limit the experienceof the
Scientific
Theories
ThatUnconceal
Being 177

being to a reexperiencing of that which is already known about it - that

which has already been conceived . Second intentions thereby hamper any
new revelations about being ; they deny being any real opportunity to
unconceal or reveal itself . And any such opportunity is most certainly
denied whenever the theorist ' s mind fails to aim toward any real being at
all , standing firm in an intention toward a concept or being of reason .
This purely reflexive aiming is most likely to occur when the concept that
is its object is second -intentional , since such a concept is further removed
from being than a first -intentional concept .
If a theorist has only second intentions while theorizing , and especially
if second -intentional concepts are the sole objects of the theorist ' s inten -
tions (which is not unlikely , because the being a theory is supposed to
describe or explain is often conceived second -intentionally ) , any real
intentional contact with being is wanting , and so any real opportunity
for genuine discovery is passed over . If being is to unconceal itself and
thereby guide the formation of theories about it , a theorist ' s mind must
aim at it directly . This is not to say that second -intentional concepts
should be avoided (for many theories require the introduction of signs tied
to such concepts in order to achieve the appropriate level of generality ) ,
but just that they should be grounded in first intentions . Such concepts
may even facilitate a revelation of being if they are compared to the first -
intentional concepts upon which they are based , and to the being that
gave rise to those concepts , to see if they " fit " the being of which they are
supposed to be (secondary ) concepts ; whether or not they fit , the exercise
of comparing them to being may lead the theorist to look at the being in a
new way such that something about it is revealed .
Theorizing solely by means of second intentions toward second -
intentional concepts is especially dangerous because such concepts can
change so that , in addition to impeding the unconcealment of aspects of
being not previously conceived , they actually come to conceal the nature
of being . This kind of change is addressed next .

9 .7 Conceptual Sedimentation and Sense Depletion

[" The disciplines ," said I ,] " which we said did lay hold of being to some extent ,
geometry and also those [disciplines ] attending it , are, as we see, dreaming about
being , but a waking vision of it is impossible for them to behold so long as they
leave unstirred the suppositions being used, not being able to offer an account
of them . For when the origin is not examined , and the conclusion as well as the
intervening steps are interwoven out of that which is not examined , what means
178 McPherson

can ever transform taken -for -grantedness such as this into understanding ?"
" None , " said he .

Plato , The Republic

A scientific or philosophical concept , when originally conceived , may well

reflect the being of which it is a concept - to the degree to which it is true
to intuitions arising from first -intentional contemplation of the being
conceived . But over the course of its use in theoretical and philosophical
reasoning , and especially through its transmission to new generations of
thinkers , a concept can diverge from and come to conceal its original
sense .

Edmund Husserl ( 1859- 1938) described such conceptual change as

" sedimentation ." If we regard a concept as it was originally conceived
(ideally in a contemplation or noetic beholding of being) as the original
" deposit ," sedimentation can be understood as a process by which layers
of sense are imposed upon the concept , usually with the effect of bringing
it into line with contemporary philosophical views and contemporary
methods , techniques , and tools (e.g., symbolic logic ) . It occurs when a
concept is used without the being that originally gave rise to it ever getting
contemplated anew. The additional layers of sensealter the way we regard
that being , the perception of which is filtered through the concept ; the
mind falls upon the changed concept before it falls upon the being (if it
falls upon it at all ). As Husserl ( 1925/ 1977) put it , " Our opinions which
stem from our theoretical or practical activities clothe our experience over ,
or clothe its sense with new layers of sense" (p . 41). The superimposed
layers of sense conceal the original sense, the sense most likely to reflect
the nature of the being that originally gave rise to the concept .
To uncover the original sense of a concept (which implies uncovering
its ontic or first -intentional root or source, if it was conceived through first
intentions ), and to distinguish this sense from the acquired senses that
cover it up , one must engage in what Husserl called " genetic analysis ."
Harvey ( 1989) describes it as follows :

Genetic analysis is the investigation backwards (Zuriickuntersuchung ) from the

given , into the layers of accumulated sense that constitute the given . Or , more
colorfully stated, it is the archaeological dig into the sense-chains of understand -
ing . By uncovering these sense-chains , genetic phenomenology can restore meaning
to knowledge domains that have become severed from their intuitive , intentional ,
and semantic roots . (pp . 52- 53)

For a theory to be truly meaningful , the theorist must conduct such a dig
through the layers of sensein any concept that guides the interpretation of
Scientific
Theories
ThatUnconceal
Being 179

a sign used in stating the theory and , having uncovered the original sense,
examine that sense to make sure it is true to carefully sifted intuitions
arising from first intentions - for even the original sense may carry con-
ceptual baggage inherited from earlier thought . And so Husserl (1954/
1970) said,
A theoretical task and achievement . . . can only be and remain meaningful in a
true and original sense if the scientist has developed in himself the ability to inquire
back into the original meaning of all his meaning -structures and methods , i .e., into
the historical meaning of their primal establishment , and especially into the mean-
ing of all the inherited meanings taken over unnoticed in this primal establishment ,
as well as those taken over later on . ( p . 56)

Few theorists currently see the need to inquire back into the original
sense and the acquired sensesof any given concept . A concept 's contem -
porary sense is typically regarded as evident and not in need of examina -
tion , and it is imagined to have been immutable across generations . The
concept gives the impression of what Husserl called Selbstverstiindlichkeit ,
literally 'self-understandability ' or 'self-intelligibility ' . Harvey ( 1989)
describes this " obviousness" as the " taken -for -grantedness" of inherited
ideas. H usserl argued tha t taking a concept for gran ted leads to 'sense
depletion ' (Sinnentleerung); the concept comes to reflect less and less the
being from which it originated .
To the degree to which a concept is sedimented and depleted of sense,
theories that make use of a sign interpreted in keeping with that concept
fail to reveal the nature of being , and the science that permits itself to be
guided by those theories is not sound, for it is not anchored to being . For
this reason, Husserl cautioned scientists against adopting and using sci-
entific concepts and terms in a thoughtless way :
It is easy to see that even in [ordinary ] human life , and first of all in every
individual life from childhood up to maturity , the originally intuitive life which
creates its originally self-evident structures through activities on the basis of sense-
experience very quickly and in increasing measure falls victim to the seduction of
language. Greater and greater segments of this life lapse into a kind of talking and
reading that is dominated purely by association ; and often enough , in respect to
the validities arrived at in this way , it is disappointed by subsequent experience .
Now one will say that in the sphere that interests us here- that of science, of
thinking directed toward the attainment of truths and the avoidance of falsehood
- one is obviously greatly concerned from the start to put a stop to the free play
of associative constructions . In view of the unavoidable sedimentation of mental

products in the form of persisting linguistic acquisitions , which can be taken up

again at first merely passively and be taken over by anyone else, such construc -
tions remain a constant danger . ( 1954/ 1970, app . 6, 362)
180 McPherson

Husserl pointed out that the incorporation of a concept into mathema-

tized theories, or into mathematics itself, and the facile manipulation, in
formal operations, logical or otherwise, of a symbol for the being of
which it is a concept increasesthe likelihood that it will become sedi-
mented and sense -depleted. There is a tension betweenmathematics as a
techne(i.e., a practical art) and mathematics as a science; proficiency at
calculation and innovations that facilitate calculation may obstruct the
unconcealmentof being.
When the original senseof a concept that was conceivedthrough first-
intentional contemplation of being is discovered(e.g., through a careful
and unprejudiced examination of ancient texts in the original languages,
or through an etymological study of the roots of the word for the being of
which it is a concept) , it should seemnatural and authentic, and, in my
experience, it is readily intelligible- perhaps becauseit arises naturally
from a contemplation of being that is entered into at will , or perhaps
because, as Jacob Klein (1940) claimed, it somehow exists within the
current senseand is understood semiconsciouslyor latently:
No science , in its actualprogress , canescape the " seduction " emanatingfrom the
spokenand written word. For the signifyingfunctionof a word has, by its very
nature, the tendencyto loseits revealingcharacter . The morewe becomeaccus -
tomedto words, the lesswe perceivetheir original and precise" significance ": a
kind of superficialand " passive " understanding is the necessary resultof the in-
creasingfamiliarity with spokenand writtenwords. The originalmentalactivity,
andproductionof significance , embodiedin soundsandsigns,is not reproduced in
the courseof actual communication . Yet it is there, in everyword, somehow
" forgotten" but still at the bottom of our speakingand our understanding , how-
evervaguethemeaningconveyed by our speech mightbe. Theoriginal"evidence "
hasfadedawaybut hasnot disappeared completely . It neednot be "awakened "
even, it actuallyunderliesour mutual understanding in a " sedimented " form.
" Sedimentation is alwayssomehowforgetfulness " [(HusserI1939 , 212)]. And this
kind
. of forgetfulness accompanies , of necessity
, the development and growthof a
SCIence.
He added,

The " sedimentation of significance

" can reachsucha degreethat a particular
science, and science
in general, appearalmostdevoidof " significance
." (pp. 155-
156)

These critical problems of sedimentation and sensedepletion can be

truly understood only through an examination of instancesof them. And
so I devote most of the remainder of this chapter to genetic analysesof
three concepts: that of a number, that of a relation, and that of a predi-
Scientific
Theories
ThatUnconceal
Being 181

cate. An " archaeological dig" into the layers of sensein these concepts
will , moreover, be of benefit to theorists becauseof their wide usein scien-
tific theories. These three concepts are examined together becausesedi-
mentation in the number concept was accompaniedby sedimentationin
the concept of a relation, and the latter contributed to sedimentation in
the concept of a predicate.
I will not attempt to give a complete or fully documented history of
changes in each concept. (See McPherson 1995, apps. A and B, for
somewhat fuller treatments.) I aim only to give the reader a feel for the
ways in which scientific conceptscan changeand what sorts of forces can
drive the changes. I begin, in each case, with a description of the concept
as it was understood in classical Greece, for I have concluded (on the
basisof my studiesof relevant Greek texts) that when theseconceptswere
conceivedby Greek scholars (some of the first thinkers to treat of them
thoughtfully), they arose from these thinkers' own reflections upon con-
cepts arising from first intentions. In each case, the original senseof the
concept will be seen to come from conceptions of things present in or
somehow depending upon real being, whereas the current senseof the
concept is such as to lead to the conclusion that its conception did not
involve any real being; it is, rather, an ensrationis in the most literal sense,
something that can reside only in the mind.

9.7.1 Numbers
Naturalnumbersweremadeby God; the restis the work of man.
LeopoldKronecker

In ancient Greek thought, numbers were distinct from the continuous

magnitudesof geometry; what we would call positive whole numberswere
the whole of number. Moreover, numbers were always tied to collections
of individuals of some kind (see Klein 1934- 1936/ 1968). As Aristotle
revealed, a number, in the Greek way of thinking , is always a number of
particular things (MetaphysicsN .5, 1092b19 - 20), whether apples or fin-
gers or coins; it is the number of individuals of one speciesor genusin a
definite collection (MetaphysicsN .l , 1088a8- 14).
In the Greek science of numbers (i .e., arithmetic), the concept of a
monad figured prominently. Some modern scholars claim that monads
were understood as abstract units to be counted, which may be taken to
imply that numbers were not always tied to real individuals. But there
is evidence that a monad was understood as a boundary or partition
182 McPherson

between items being counted , where the items are individual beings of one
species; a monad may have been understood to separate or partition off
one individual as it was added to a collection of individuals already sep-
arated out in counting (as Robert H . Schmidt has argued , in conversa-
tion ; alternatively , a monad may have been understood as a resting -place
or stopping -place, mone, for the movements (say, of a finger or hand )
making up the act of separating out an individual as items are counted in
succession- as the telos or end of one cycle of movements during count -
ing ). This thesis is compatible with Aristotle 's description of counting as
'by parts ' or 'by portions ' (kata meridas), which implies separating some-
thing out and setting it apart or allotting it , and as 'takings in [or addi -
tions ]' (proslambanontes ; Metaphysics M .7, IO82b34- 36), as well as his
description of a count (i .e., a number ;9 arithmos ) as 'a delimited multi -
tude ' (plethos peperasmenon; Metaphysics !! .13, 1020a13) and Eudoxus 's
definition of a count as 'a multitude marked [or separated] out by bound -
aries' (plethos horismenon; see Klein 1934- 1936/ 1968, 51). It is also com -
patible with Aristotle 's characterization of monads as being one after
another or in succession (ephexes; Physics E .3, 227a29- 30), his descrip-
tion of a monad (along with a point , a line , and a plane) as something by
which a body gets marked (or separated) out by boundaries (or bounded ,
defined , or delimited ; ( toutoij horistai to soma; Metaphysics B.5, 1002a4-
6), and his description of a count or counting (arithmos ) as that which
marks (or separates) out by boundaries (or bounds , defines, or delimits )
all things (horizein panta ; Metaphysics ~ .8, 1017b20- 21). The thesis
accords as well with a statement of Nicomachus : 'By means of the monad ,
all the counts having begun from the dyad , being separated out together
one after another , generate the orderly species of that which is multiple in
the proper sequence' (see Klein 1934- 1936/ 1968, 52, for the Greek ).
If monads were, in addition or instead , understood by some mathema -
ticians to be the items of counting themselves, they nonetheless seem not
to have been regarded as separate from real beings, but rather as present
in real beings. 10A monad (monas, 'aloneness' , derived from monos, 'alone '
or 'solitary ') seems sometimes to have been understood as the oneness in
an atomic being as isolated by abstraction in the Greek sense, that is,
what is left over in an atom (e.g., a cat) when all its particular properties
other than its oneness- a oneness concomitant with its indivisibility as
something informed by its eidos or species (or " basic-level kind " ; see
Aristotle , Metaphysics B .3, 999al - 6; ~ .6, 1016b3- 6; M .3, 1078a23- 25)-
are mentally stripped away from it . (See Metaphysics K .3, 1061a29- 35,
Scientific
Theories
ThatUnconceal
Being 183

and Categories 7, 7a31- 7b9, for examples of abstraction as the Greeks

understood it - aphairesis, 'taking away from ' , 'carrying off ' , or 'removal ' ,
or periairesis , 'taking away that which surrounds ' or 'stripping off ' .) This
thesis (and the partition thesis as well ) is in line with Euclid 's definition of
a monad : 'A monad is that because of which [or in accordance with
which ] each of the beings gets called one' (Elements, bk . 7, def. 1; see
Heath 1956) . Also compatible with the abstraction thesis are Aristotle 's
arguments that mathematical entities (presumably including the monad )
do not exist separately from beings (Metaphysics M .2), even though they
are sometimes considered separately by arithmeticians and geometricians
(Metaphysics M .2; M .3, 1078a21- 23; On the Soul r .7, 431bI5 - 17), and
that arithmoi in particular , which are counts of monads , do not exist
separately (Metaphysics M .8) .! ! The understanding of a monad as the
oneness in an atom that is concomitant with the indivisibility deriving
from its species is complemented by the understanding as a partition , for
to delimit or define the items of counting by means of a partition is to
reveal which species is the source of oneness for each of them . Counting
individuals delimited in this way may even , in some sense , effect the
abstraction that leaves only the oneness in a being , for to separate out an
individual and count it as one is to ignore everything about the individual
except its being an atom or monad ; in being counted (and in having been
counted ), an individual is understood as one- as a monad .
It seems, then , that a number (or count ) was normally understood to
exist only by virtue of a collection of beings of the same kind , as the count
of individuals in that collection . For the ancients , numbers could not exist

independently of collections of beings.

This view was preserved by Roman mathematicians , but they created
a numeric notation that allowed them to use collections of strokes on

papyrus as the objects of their calculations (much as the Babylonians

had used collections of slim wedge-shaped impressions in clay in their
calculations ) . These exemplars of collections could take the place of other
collections to facilitate reckoning . In the Roman numeral system (as in
the Babylonian numeral system), the symbols for the numbers one through
three are I , II , and III . A manipulation of these symbols is a manipulation
of collections - collections of strokes . These collections can stand for , or

stand in for , collections of individuals of any kind whatsoever (Schmidt

1986). Schmidt calls them " counterparts ," and I will use this term here. A
counterpart is not a sign for a collection , but an example of one that can
stand in for another so that operations on it stand in for operations on the
184 McPherson

collection for which it is a counterpart . Schmidt gives as another example

of a counterpart a nautical chart . A course is planned on the chart ,
whereupon the planned movements are executed in the place that the
chart represents. In a comparable way , manipulations of numeric coun -
terparts lead to results that can be transferred to any other collections
having the same numbers of individuals : if two strokes counted together
with three strokes gives a collection of five strokes, then two coins counted
together with three coins will give a collection of five coins . The Roman
numeral for the number five , V , is not itself a collection of strokes , but it
was understood to be a sign for , or rather shorthand for or a proxy for ,
five strokes; it meant 11111
. And so some Roman numerals (e.g., V, X , L )
-
were not themselves counterparts , but stood in the place of counterparts
and were operated upon just as if they were the counterparts they stood
for .

When the Hindu -Arabic numeric notation became popular in the West ,
late in the twelfth century , the symbols seem initially to have been
understood in just the same way as the Roman numeral V, namely , as
shorthand marks or proxies for collections of strokes on paper that
were counterparts for collections of individuals of any kind whatsoever .
( Leonardo of Pisa- now known as Leonardo Fibonacci - called his book
of 1202 on Hindu -Arabic numeric symbols and algebra Liber Abbaci ,
'Book of the Abacus ' . The new numeric symbols would have brought to
mind an abacus because both the counters on an abacus and the digits
of a Hindu -Arabic numeral were understood as counterparts in a place-
value system.) Alternatively , the symbols may have been understood as
proxies for Roman numerals , which were either counterparts for collec -
tions of individuals or shorthand or proxies for such counterparts
(Schmidt 1986).
Over time , this understanding of the symbols was lost ; mathematicians
began to forget that their symbols were originally intended to be proxies
for counterparts . As a result , the symbols began to lose their conceptual
tie with collections of individuals , and the number concept began to
change correspondingly , losing its basis in multitude .
With collections of individuals no longer consciously guiding the con-
ception of number , the concept became more inclusive . Proper fractions
were reinterpreted as numbers ; ratios (newly interpreted as quotients ) of
whole numbers and of commensurable magnitudes were reinterpreted as
(rational ) numbers ; and ratios of incommensurable magnitudes were
reinterpreted as (irrational ) numbers (or at least this seems to be the way
Scientific Theories That U nconcealBeing 185

in which irrational numbers were originally understood ; in modern times ,

irrationals are understood as " real " numbers invented so as to fill in gaps
in the " real number line ," where the filling of gaps is accomplished by
means of Dedekind cuts or Cauchy sequences) . For the number concept
to include fractions of units , quotients of any two numbers , and quotients
of any two magnitudes , whether commensurable or not , a number had to
have been newly conceived as something similar in nature to a continuous
magnitude ; and so it had , as John Dee ( 1527- 1608) explained :
Practice has led Numbers farther , and has framed them , to take upon them , the
show of Magnitudes ' property : Which is Incommensurability and Irrationality .
( For in pure Arithmetic , a Unit , is the common Measure of all Numbers .) And ,
here , Numbers are become , as Lines , Planes and Solids : sometimes Rational ,
sometimes Irrational . ( 1570/ 1975, 5)

The change in the concept likely came about because mathematicians

thoughtlessly began to perfoffi1 operations that the syntaxes of their
mathematical languages permitted (such as multiplying and dividing , not
just numbers , but also ratios ) but that were not meaningful until the
concept of num "ber changed. This may be what Dee meant by " practice "
framing numbers to take on a property of magnitudes .
Though the new concept of number as continuous like magnitude
seems implicit in mathematical writings of the late Middle Ages and the
early Renaissance, the first mathematician to whom an explicit espousal
of it is attributed is Simon Stevin ( 1548- 1620) . He viewed number as a
property of magnitude comparable to the wetness of water :
Number is something in magnitude comparable to wetness in water , for as this
extends everywhere and in each part of the water , so number tied to some mag -
nitude extends everywhere and in each part of its magnitude : Just as to a water
continuum there corresponds a wetness continuum , so to a magnitude continuum
there corresponds a number continuum . Just as the wetness continuum of the
entirety of the water undergoes the same division and separation as its water , so
the number continuum undergoes the same division and separation as its magni -
tude , so that these two quantities cannot be distinguished by continuity and dis-
continuity . ( 1585/ 1958a, 502)

Stevin may have been influenced by his own adoption of decimal

notation for fractions (see Stevin 1585j1958b ) . He was one of the first
mathematicians in the West to use decimal fractions , and the first to suc -

cessfully promote their use. Acceptance of this notation could lead one to
believe that there is a number for every possible decimal expansion , so
that the numbers can fill up a continuum .
186 McPherson

Stevin even went so far as to argue that the notion of discontinuous

(or whole ) numbers is merely a product of the imagination :

All that is just a quantity is not at all a disjunct quantity . Sixty , according as it is
number , is a quantity (that is to say a number ) . Sixty , then , according as it is
number , is not at all a disjunct quantity . As for that which you divide by your
imagination , this proposed unique and whole quantity , into sixty unities [or units]
(which you could divide , by equal reason, into thirty dualities , or twenty trinities ,
etc.), and which then you next define [to be] the divided up , that is not definition of
the proposed of which there is some question : you could similarly divide the pro -
posed magnitude , by the imagination , into sixty parts , and then by equal reason
define it to be discontinuous quantity , which is absurd . ( 1585/ 1958a, 501- 502;
emphasis added)

So for Stevin , at least, numbers had become completely divorced from

collections of beings and multitude . But this means that numbers had
(ironically ) moved completely into the realm of imagination , as entia
rationis alone (despite the reality Stevin attributed to the " number " he
imagined to be inhering in magnitudes ).
Stevin 's view of number as continuous found easy acceptance, perhaps
because of its practical utility , or the utility of the associated decimal
fractions , and perhaps also because it had been implicit in mathematical
reasoning for some time (despite explicit rejections of the irrational num -
bers it permitted ; e.g., Frater Fredericus said, c. 1460, 'An irrational
number is not a number . For a number is that which a unit measures ' ; in
1544 , Michael Stifel said that ' an irrational number is not a true number ';
see Klein 1934- 1936/ 1968, 251, n . 176; in 1560, Jacques Peletier argued
that irrational numbers are not numbers in the absolute sense , but that
they are nonetheless necessary in practice , even though unintelligible ; he
said that what they truly represent 'conceals itself in perpetual darkness';
see Klein 1934- 1936/ 1968, 290, n . 302) . The new number concept may
also have been welcomed because it facilitated contemporaneous attempts
to submit both numbers and magnitudes to algebraic treatment ; doing so
presupposed either interpreting geometric objects as special cases of
numbers (as John Wallis seems to have done12) or interpreting arithmetic
objects (i .e., numbers ) as special cases of geometric magnitudes (as Des-
cartes may have done13). But not all mathematicians understood the new,
continuous number in the same way as Stevin . John Wallis (1616- 1703)
assimilated his concept of number to the concept of a ratio of continuous
magnitudes . In Greek geometry , continuous magnitudes acquired some-
thing comparable to values only through comparison with other (com -
mensurable ) continuous magnitudes ; for instance , the magnitude of one
Scientific Theories That Unconceal Being 187

side of a triangle was understood quantitatively in terms of its magnitude

relative to the magnitude of another side of the triangle - that is, the ratio
that magnitude had to the magnitude chosen as a standard . ( For the
Greeks , this ratio was always understood in relational , not numeric ,
terms ; as will be seen in the next section, it was only much later under -
stood to be a number .) For Wallis , any number was a ratio of some
quantity to a unit quantity , or rather the quotient of the antecedent and
consequent of that ratio (which he called an 'index ' or 'exponent ') :
The whole of arithmetic itself , if looked at more closely , seems scarcely other than
a doctrine of ratios . And numbers themselves [are] indices of just so many ratios
whose common consequent is I, unity [i .e., the unit ] . For when I, or unity , is re-
garded as [the] quantity set forth [as the reference unit ], all the rest of the numbers
(whether whole , or broken , or even irrational ) are indices or exponents of just so
many ratios of other [quantities ] to [the] quantity set forth [i .e., to unity ]. (Mathe -
sis Universalis ; for the Latin , see Wallis 1657/ 1695, Opera, vol . 1, as cited in Klein
1934- 1936/ 1968, 220- 221)

In this interpretation of numbers , Wallis attempts to understand the

strange new continuous numbers in terms of the more familiar geometric
concept of a ratio of magnitudes .
When even quotients of incommensurable magnitudes came to be
understood as numbers , mathematicians could imagine that to every point
along a continuous magnitude there corresponds a number . 14 To facili -
tate calculation , they even came to include negative numbers among the
so-called real numbers . With no possible tie to multitude or even geo-
metric magnitude , numbers were now fully divorced from any entia rea/is
(making ironic the appellation " real numbers " ). In the modem view ,
numbers correspond to points on the " number line ," an imaginary entity
that allows number to be understood to be continuous and extending to
infinity in two directions from an origin .
With numbers divorced from entia realis , mathematicians got into great
muddles trying to describe the interpretation of a numeral . Gottlob Frege
(1813- 1890), for instance , took a numeral to be a proper name for a
number , which is, according to him , an invisible and locationless but self-
subsistent individual of some mysterious sort defined in terms of one-
to -one correspondence (see Frege 1884/ 1980). This odd view may again
reveal an effect of notation . The interpretation of a numeral as a proper
name for an individual of some sort follows naturally from the notation
that mathematicians had come to use. In natural language , numerals
enter mostly as quantifiers with count nouns for individuals (e.g., three
188 McPherson

cats), in keeping with the Greek (and presumably our first-intentional)

understanding of numbers; but in modern mathematical notation,
numeralsappearin isolation, just asproper namesdo in noun phrases. (The
Greeks always placed a count noun next to a numeral, becausea number
was always a number of something; in arithmetic, the noun was monas-
'monad'. The disappearancefrom mathematics of count nouns is a con-
sequenceof number moving into the realm of the purely imaginary, where
numbers are no longer counts of beings; instead, they are points on the
number line.)
Some mathematicians steered clear of the muddles by abandoning
attempts to interpret a numeral. Mathematicians in the formalist camp
came to view the written or typed marks on paper that are instancesof
numerals as the objects of mathematics, and operating on symbols, or
syntactic manipulation, as the whole of number science. The symbols are
not interpreted at all, in this way of thinking .
The concept of number seemsto have become one of what Husserl
(1954/ 1970) called " ideas emptied [of sense ] which have been obscured
and have becomemere word-concepts; ideas burdened, through attempts
at exposition, with false interpretations" (p. 74). The concept has become
so weirdly distant from anything involving real being that numeric sym-
bols have becomelittle more than tokens- in the senseof vestigesof the
numeric proxies for counterparts that mathematicians used into the
Renaissance(for a numeral is no longer understood as shorthand for or a
proxy for a counterpart for a collection of beings), as well as in the sense
of formal marks that signal the permissibility of certain operations upon
them. The original senseof the concept survivesin only the vaguestforms,
namely, in the fact that definitions of number begin with the definition of
the natural numbers (though theseinclude zero and one; the former is not
a count, and the latter was originally the measurefor counting) and in
the fact that these numbers are defined in terms of succession , which is
understood (if not initially defined) in terms of successiveadditions of one
(but not of one thing) as an imperfect reflection of counting in the Greek
mode. But as soon as the integers (which include negative numbers), the
rationals, and the reals enter the picture, the original senseis tossedaside.
This is not to deny the utility of the modern number concept in practical
applications, but that utility dependsupon interpreting a positive number
that is the result of calculations as the measure of the multitude of the
beingsin a collection, or as a multiple of unit magnitudes(such as inches
or centimeters).
Scientific
Theories
ThatUnconceal
Being 189

Changes in the concept of number appear to be attributable , in a great

measure , to changes in notation that set up conditions for forgetfulness

about the original sense ( see also Schmidt 1986 ) . The Hindu - Arabic

numeric notation made the counterpart function of numerals invisible ,

opening the door to the imposition of a new layer of sense upon the

number concept ( with the new sense apparently seeping into mathematics

through initially meaningless extensions of mathematical operations that

were permitted by the syntaxes of mathematical languages ) . The modern

sense of the number concept , with number regarded as continuous like

magnitude , fully covered over the original sense after the introduction

into widespread use of decimal fractions .

9 . 7 . 2 Relations

A new interpretation of the ratio accompanied the new interpretation of

number as continuous . The ratio was the prototypical relation for mathe -

maticians , and so the new understanding of it may be responsible for

observable change in the concept of a relation in general .

Greek mathematicians understood a ratio to be an aspect of the being

of one magnitude or number as considered relative to another . So a

magnitude - say , one side of a given triangle ( understood as a magnitude ,

i . e . , as a length ) - having a ratio of 2 : 1 to another magnitude - say , one

of the other sides of the triangle ( understood as a magnitude ) - was seen

to have as a part of its being doubleness in relation to the other ; the

greater magnitude was understood to be the subject or substrate for the

doubleness , which was determined or defined by reference to the lesser

magnitude , the object of the relation . A relation of any sort was under -

stood in this way , as an aspect of the being of one individual , its subject ,

defined with reference to another individual , its object .

Oddly enough , ratios came to be interpreted as numbers . The ratio 6 : 3

is now interpreted as the common fraction 6 / 3 , which is equal to the

number 2 , the result of dividing the numerator by the denominator

( i . e . , the quotient ) . We call the numbers that are identified with the new

" ratios " rational numbers .

The reinterpretation of ratios as numbers seems to have been tied to the

new concept of number as something along a continuum . When numbers

were no longer understood as measured multitudes , but as exponents or

indices of ratios of magnitudes ( for Wallis ) , properties of magnitudes

( Stevin ) , or multiples of unit magnitudes ( Descartes ) , the result of the

division of the antecedent of a ratio by the consequent could be interpreted

190 McPherson

as a number whether or not the result was a whole number . Even before

numbers were explicitly interpreted as continuous , the potential reinter -

pretation of ratios as numbers may have been at least partially actualized
through forgetfulness concerning the use of what were called " denomi -
nations ." In use from the late Middle Ages, denominations are names
given to ratios , taken from their " base" ratios . A ratio - say, 6 : 3- is de-
nominated by the ratio of the smallest numbers for which the antecedent
is in the same relation to the consequent - or 2 : 1 in this case . But how did
denominations contribute to a reinterpretation of ratios as numbers ? It
seems that denominations themselves came to be understood as numbers .

Probably the first to be so understood were denominations of the proto -

typical ratios , " multiples " such as double (2 : 1) and triple (3 : 1), which are
ratios to the unit ; such denominations were especially likely to come to
be understood as numbers because they were given as the antecedent
alone (e.g., 2 for 2 : 1, 3 for 3 : 1) . But it seems that the denomination for
any ratio of commensurables came to be understood as a number ;
Thomas of Bradwardine ( 1290- 1349) said that any ratio of commensur -
abIes 'is immediately denominated by some number, just as in the case of
the ratio double , and triple , even so in the cases of the others ' (1328/ 1955,
66; emphasis added). Moreover , from the thirteenth century onward ,
denominations were used in mathematical computations just as if they
were numbers .

When ratios came to be interpreted as numbers , mathematicians '

understanding of relations seems to have changed concomitantly . The
number to which a ratio is reduced- say, the quotient 2 as the reduction
of the ratio 6 : 3- is not an aspect of the being of the antecedent defined
by reference to the consequent; it is, rather , a separate individual number.
( The 2 is a measure of parts of the antecedent, where each part is equal to
the consequent; but unlike the ratio of that measure to unity , or 2 : 1, it is
not relational .) A relation in general likewise came to be viewed as sepa-
rate from the two individuals involved , not present in either individ ,ual - a
third " thing " of some sort . For this reason, scholars ceased speaking of
the relation olone thing to another (where the of implies that the relation
belongs to the first thing , and the to implies that the relation is defined
with reference to the second thing ) . They began to speak instead of the
relation between two things . This way of talking seems to imply that a
relation came to be viewed as something that exists in neither individual ,
but that exists between them somehow , connecting them with one another .
With a relation newly conceived as a connection of some sort , it can
Scientific Theories That Unconceal Being 191

involve more than two individuals, in which caseit is sometimesspoken

of as a relation among individuals.
Gottfried Wilhelm Leibniz (1646- 1716) wrote what is perhapsthe first
and maybeeventhe only explicit description of the modern " relation." He
characterizedit as a type of relation distinct from the usual and described
it as an 'ideal thing', which meant, for him, an idea in the mind of God
(e.g., 1714/ 1902, 262). His description of the new " relation" appearsin a
letter written to Isaac Newton's disciple SamuelClarke:
The ratio or proportionbetweentwo linesLand M can be conceivedin three
ways: asa ratio of the greaterL to the lesserM, asa ratio of the lesserM to the
greaterL, and finally assomething abstracted from the two, that is to sayas the
ratio between Land M, withoutconsidering whichis the antecedent or theconse -
quent, the subjector the object. . . . In the first consideration, L the greateris the
subject ; in the second , M the lesseris the subjectof that accident , that thephilos-
opherscall relationor rapport. But whichof themwill be the subjectin the third
sense ? Onecouldnot saythat both of them, Land M together , arethe subjectof
suchan accident , for that way wewouldhaveoneaccidentin two subjects , which
wouldhaveonelegin theone, andtheother[leg] in theother, whichis contraryto
the notion of accidents . Thereforeonemustsay, that this rapport[or relation] in
this third sense is reallyoutsideof thesubjects ; but that beingneithersubstance nor
accident , it mustbe a purelyidealthing, the consideration of whichis nonetheless
useful. (1717j1965b , 401; emphasis added )
By the " ideal" relation " abstracted" from the two lines, Leibniz likely
meant the order of the magnitudes of the two lines, for elsewherehe
describedtime and spaceas orders that he called " relations" (seeLeibniz
1717/ 1965b), and, in the following passage, he again describesorders, as
well as various other arrangementsof interrelated entities, as individual
relations:
Thereare. . . examples of a relationamongseveralthingsat once, asthat of order
or that of a genealogical tree, whichexpresses
the rank andthe connectionof all
the termsor members ; andevena figurelike that of a polygoncontainsthe rela-
tion of all the sides
. (1765/1965a , 211)
A " relation" in the new sense(which was, for Leibniz, an idea in the
mind of God) cannot be presentin any ensrea/is, and so it can be nothing
more than an ensrationis in the minds of modern thinkers.
In mathematics, a relation (or what is called a " binary relation" ) has
come to be defined and understood as a set of orderedpairs. The order is
such that the first coordinate in a pair is always in the relation to the sec-
ond coordinate in the pair. The set of all the first coordinates and the set
of all the secondcoordinates are called the " domain" and the " range" of
 192 McPherson

the relation, respectively. A relation defined as a set of ordered pairs of

individuals resemblesLeibniz's favorite " ideal" relation, namely, an order
of individuals; but it is not the order itself; it is, rather, ordered individuals.
It is a poor reflection of a relation in the original sense,for both coordinates
are individuals- say, the numbers 6 and 3 (or delimited multitudes of
6 and 3 individuals, if numbers are interpreted more traditionally )-
whereas a relation- say, double- is just one aspect of the being of an
individual (or of a delimited multitude of individuals), which is defined
with referenceto another individual (or to another delimited multitude of
individuals)- say, of 6 (defined with referenceto 3); the doublenessof 6
as compared to 3 is not equivalent to 6 or to 3, or to 6 and 3 considered
together, or to their order. The order of two individuals in an orderedpair
is a conventional way of signaling which individual bearsthe relation and
which individual it is related to, which reveals the stamp of the original
senseof the concept. But since the inverse of a relation, or the set of
ordered pairs with the order reversed, is usually identified with the rela-
tion, and sincesetsof ordered triples, ordered quadruples, and so on, are
also consideredto be relations, the modem senseof a relation as some-
thing separatefrom and somehowconnectingthe individuals involved has
evidently beenlaid over the original sense. And since a relation is defined
as a set of ordered pairs, as opposedto an aspectof the being of the first
coordinate as consideredrelative to the second coordinate, the original
senseis covered over with yet another layer of sense. The original sense
nonetheless peeks through the formalization in the ordering of the
individuals in a pair (with the bearer of the relation first in the order). The
original senseof relation somehowlies at the bottom of the mathematical
concept, but the Leibnizian notion and another idea, that of individuals in
a certain order, are laid over the original sense.
The concept of a relation had its original sensecovered over, it seems,
primarily becauseof a new conception of the prototypical relation, the
ratio- a new conception tied to change in the concept of number. An
additional layer of sensethat concealsfurther the original sensecame to
be laid upon the concept as a result of its being incorporated into set-
theoretical mathematics, which deals only with individuals- individuals
qua individuals- so that relational being can be accommodated only
through placing individuals possessingit and the individuals to which
they are referred into sets. Set theory provides no meansfor dealing with
relational being in a way true to its nature as an aspect of the being of
some substrate, dependent for its existence upon that individual and
ScientificTheories That Unconceal Being 193

someotherindividual to which the foffi1eris referredso as to definethe

relation.

9.7.3 Predicatesand Predication

Aristotle 's teachings on logos in the Organon reveal a highly developed
understanding of predicatesand predication. Though his view was prob-
ably unique to him, his conceptsof predicatesand predication appear to
be basedon first intentions, so they are worthy of careful consideration-
especially becausethey are so well hidden in modem translations and
treatments of his works.
The words Aristotle used for a predicate and predication are clues to
his conceptsof them. The noun he usedfor a predicateis kategoria, which
is derived from the verb he used for predication, kategoreD. This verb
means'accuse', or literally 'speakout against publicly'; it is formed from
the prepositional prefix kata, 'against' , and the verb agoreuo, 'speakin the
agora (i .e., the assemblyof the people)'. In common language, the noun
kategoria was usedfor the speechfor the prosecutor in front of a tribunal
(and it was opposedto apologia, the speechfor the defendant) . Aristotle
sometimesusedanother expressionfor predication, one synonymouswith
kategoreo: lego kata tinos, 'speakagainst someone/something'. Aristotle 's
choice of expressionsfor a predicate and predication15 suggeststhat he
regardedpredication as an attribution of responsibility to the subject for
whatever the predicate signifies (cf. Heidegger 1927/ 1962). In accusing
someoneof some deed, we hold that person responsiblefor it ; we claim
that the person gave rise to it . (In Athenian law, an accusationcould be
made against an animal or something inanimate if , say, it had played a
part in a death by unnatural causes(seeMcPherson 1995). So the use of
accusationas a metaphor for predication would not carry the implication
that predication is restricted to predications about persons.)
Aristotle seemsto have based his ascription of responsibility to the
subject on its being the origin or arche of that which the predicate signi-
fies. He used the verb hupaicho to describewhat a predicate doesin rela-
tion to a subject. This word contains, as a prefix, the preposition hupo,
'under', and the verb archo, 'originate'. It meansto originate under in the
senseof being dependent upon or under the power of.16 So Aristotle
seemsto have understood the being a predicate signifies to originate in
dependencyupon a subject. (In On Interpretation 5, 17a23- 24, he said
that 'the simple declarative sentence[apophansis ] is significant spoken
sound about whether something [in the caseof an affifl1lation] originates
 194 McPherson

in dependency [huparchei ] or [in the case of a denial ] does not originate

in dependency' . In Prior Analytics A .37, 49a6- 8, he explained that this
dependent origination is to be apprehended differently for each of the 10
schemata of predication , or kategoriai .) Much the same idea is implied by
the word chosen for a subject: hupokeimenon, the present (middle ) parti -
ciple of the verb hupokeimai , 'underlie ', here used as a nominal , means
'that which is underlying ' . The subject is the being that supports the being
signified by the predicate ; it is its substrate and its source.
Laying an accusation against someone also brings something about
that person to light ; it reveals something brought into being by that per-
son (say, an unlawful deed), something that was hidden . Aristotle seems
also to have understood predication to reveal , or make noetically mani -
fest, something about the subject .17 This view is revealed, again , in his
terminology . His word for a declarative sentence, which combines a
predicate with a subject and a copula , is apophansis; the preposition apo
that appears as a prefix means 'away from ' or 'forth ', and phansis is
equivalent to phasis, a noun derived from the verb phaino , this verb
meaning 'bring to light ', 'cause to appear ', 'show' , 'uncover ' , or 'reveal ' ,
but also 'denounce ' or 'inform against ' . The primary meaning of the word
apophansis seems to have been 'showing forth ' . In contrasting the effect of
an apophantic utterance with the effect of an isolated word , Aristotle said,
'N ow the noun as well as the rhema [i .e., verb or adjective ], let it be
[called ] only an utterance , for one is not , in speaking in this way , making
something [noetically ] manifest with that spoken sound so as [for it ] to
show itself forth [apophainesthai ]' (On Interpretation 5, 17aI7- 19). His use
of the middle -voice form of the infinitive of the verb apophaino reveals
that the subject is showing itself forth , from itself (see Heidegger 1927/
1962); for much the same reason, Aristotle described the sorts of being
signified in the different schemata of predication - the kategoriai - as
kath ' hauta , or 'because of (or in accordance with ) themselves' (Meta -
physics L\ .7, IOI7a22- 23).18 The nature of the showing can be discerned
through a consideration of the word ph asis that is a part of the word
apophansis, and that is also a part of the words for an affirmation and a
denial , that is, a positive and a negative declarative statement , namely ,
kataphasis and apophasis.19 Phasis (as derived from phaino ; phasis was
derived fromphemi as well ) had one meaning that seems to be particularly
enlightening in this context : it was used for the appearance of a (fixed or
wandering ) star as it rose above the horizon , and especially for its first
such appearance after a period of concealment under the sun's beams
Scientific Theories That Unconceal Being 195

(i .e., for its heliacal rising ). So the word strongly suggests something that
was hidden coming to light or becoming manifest . The same suggestion
comes from another source. As Aristotle pointed out , apophansis is the
only kind of utterance that can be judged true or false; neither an indi -
vidual word or phrase nor a sentence of any other type (e.g., a question , a
prayer , a request, or a command ) can be said to be true or false (see On
Interpretation 1, 16a9- 18; 4, 16b33- 17a4). But the Greek word translated
as 'true ' is alethes (a-leth-es), 'taking out of hiding ' or 'unconcealing ' (or
'unconcealed ') . The privative or negative prefix a- expresses want or
absence , loss , or forcible removal , and the verbal stem contained in this
adjective is that of the verb letho or lanthano , which means 'escape
notice ' , ' be unseen ' , ' be unnoticed ' , or ' be unknown ' ; in the middle and
passive voices, this verb means 'forget ' . So the word usually translated
as 'true ' suggests something failing or ceasing to escape one's notice or
remain unseen; it suggests something that was hidden or forgotten - or
something unknown - being brought to light or revealed. (The Greek
word for false is pseudes, 'deceitful ' or 'lying '; it means giving the im -
pression of unconcealing something without doing so.)
That which predication brings to light is being. This is evident in the
fact that the copula , in any language that has it , is the verb meaning
'be' . In declarative statements, the copula (whether or not it is explicit )-
which is conjugated to agree with the subject (in person and number )-
indicates that being is unconcealed from the subject.2o (When a predicate
and the copula combine with a subject, the being that is shown forth or
unconcealed from the subject- the being that the predicate signifies- is
being that depends for its existence upon the subject, its support and its
origin . When the copula combines with a subject alone , as in Socrates
was, the being unconcealed is that of the subject as a whole .) And so
Aristotle said that the 'is' in an affirmation signifies that it is true or
unconcealing (Metaphysics Ll.7, 1017a31- 34) , that being is the true or
the unconcealed , and that a true affirmation is unconcealing because the
being signified by the predicate is lying together with the being signified
by the subject (Metaphysics 0 .10, 105la34 - 1051b13) .
The nature of the experience of an unconcealment of being is revealed
in the use of the word episteme for the understanding that comes from
hearing a true apophantic utterance or a syllogism or dialectical argument
in which such utterances are properly combined . This noun is derived
from the verb epistamai , which is believed to be an old middle -voice form
of ephistemi; the latter is composed of the prepositional prefix epi, mean-
 196 McPherson

ing 'upon ' , and the verb histemi , which has among its intransitive mean-
ings 'stand ', 'stand still ', and 'stand firm ' . And so the most literal meaning
of the nominal episteme (formed with the suffix -me) would seem to be
'standing upon ' , 'standing still upon ', or 'standing firm upon '; what we
call understanding (or standing under ) was called overstanding (or stand-
ing over or upon ). The standing would seem to be the mind 's standing
upon unconcealed being . Aristotle regarded this standing as a stillness or
resting of the mind - resting not in the sense of being slack, but in the
sense of being arrested by a tension in one direction : 'Mental perception
[noesis] is more like a sort of rest [or stillness] and stopping than motion '
( On the Soul A .3, 407a32- 33); 'When the discursive thought [dianoia ] has
come to rest and has come to a standstill , we are said to stand upon [or
understand ; epistasthai] and to intend [or attend ; phronein ]' (Physics H .3,
247bll - 12). The experience Aristotle described is likely the aiming of the
mind that is concomitant with an unconcealment of being ; in intending or
aiming toward being , the mind is necessarily standing firm for so long as
the being is unconcealed - for when it ceases to stand still , it will turn in
other directions , and the being toward which it was aimed will slip out of
awareness, entering into or returning to a state of escaping one's notice or
being hidden or forgotten .
The Aristotelian account of predication and of the apophansis or
showing forth that results from it implies that hearing or reading or
saying to oneself " Cats are furry " makes the being of cats' furriness -
being that originates in and depends upon cats- appear noetically to
the mind as it aims toward that being , which was escaping one's notice ,
hidden from consciousness or forgotten , and which now comes to light
(and remains in the light until the mind ceases to stand firm , aimed
toward it , and turns elsewhere) .
After the Aristotelian theory of predication was forgotten , concepts of
predication were swept every which way with the currents of philosoph -
ical movements , like shifting sediment . For a partial history of the various
views that arose, see McPherson 1995. Briefly , here are a few of those
views : According to nominalists such as William of Ockham , the subject
and the predicate are names for the same thing (and Geach ( 1962) has
thoroughly trounced this view ). In the opinion of Thomas Hobbes , they
are marks that bring to mind thoughts of the same thing . James Mill ,
influenced by Hume ' s and Hartley 's associationism , asserted that they are
names for ideas experienced in succession or concurrently . The theories of
Hobbes and Mill are examples of one general trend . Under the influence
Scientific Theories That Unconceal Being 197

of materialism- a tenet that precludesany serious consideration of first

intentions- many thinkers came to deny that linguistic signs can bring
the mind into any relation to being. The force of the copula 'be' was for-
gotten, and predication came to be seenas a relation of one mental entity
(e.g., a concept or an idea) to another. As such, it could not involve any
attribution of responsibility for the origination of being, or any uncon-
cealment of being.
A view revealing another trend had becomethe " establishedopinion"
by the mid-nineteenthcentury (Mill 1851, 103) and still prevails: the idea
that the subject is included in a class of things that the predicate is sup-
posedto signify; Cats arefurry is supposedto mean that cats are included
in the class of furry things. John Stuart Mill (1851) describedthis theory
as " a signal example of a logical error very often committed in logic, that
of . . . explaining a thing by something which presupposesit " (p. 104);
predication in the original senseis presupposedin the class-inclusion
approach becausethe class of furry things cannot be defined unless one
can determine which things are furry - that is, of which things the predi-
cate furry can be truly predicated. This class-inclusion approach is in
keeping with a trend toward extensional versus intensional thinking . As
modern times approached, the extensional (e.g., spatial dimensions of
bodies) came to be deemedmore and more real as compared to the in-
tensional (i.e., that which intensifies and remits, such as a form or a
quality; more and more qualities came to be considered" secondary," or
primarily subjective). This trend is evident in Descartes's assertionsthat
only extension and its concomitants (e.g., figure and motion) can be
deemedreal with any certainty, the reality of qualities being questionable
(e.g., Meditations on First Philosophy3), and that extensionis the essence
of bodily substance(e.g., Principles of Philosophy3). (Robert H . Schmidt
has argued, in conversation, that this extensional bias arose becauseof
a desire to follow Galileo in seeking a mathematical reality behind the
appearancescombined with a continuing failure to adequately measure
intensivemagnitudesso that intensionscould be mathematized- a failure
that led to the abandonment of this measurementproblem.) With the
reality of intensional being (such as a quality) qua intensional widely
denied, such being came to be reinterpreted as a collection, a class or a
set, so that it could be imagined to be extensionalrather than intensional.
As modern symbolic logic developed, the class~inclusion interpreta-
tion of predication underwent a change so as to include classesor sets
of ordered couples, triples, and so on. In symbolic logic, a " predicate" is
 198 McPherson

understood to signify , in the case of a monadic or one-place predicate , a

property , or , in the case of a polyadic or many -place predicate , a relation
in the modern sense (i .e., an order or other " connection " of any number
of interrelated individuals ), and the usual interpretation is extensional
rather than intensional , with a property identified with the class or set of
individuals possessing that property , and with a relation identified with
the class or set of ordered pairs (or ordered triples , etc.- ordered n-tuples)
for which the relation holds . The idea has its origin in Frege's many -place
function (even though Frege did not understand predication in terms of
class inclusion ) . In a Fregean function , variables symbolized by letters are
substituted for some or all of the noun phrases that are the arguments of a
verb heading the predicate of a declarative sentence. This historical source
alone suggests that the concept of a predicate in symbolic logic is irrele -
vant to a theory of predication , for Frege did not design his symbolic
language to take account of the distinction between a subject and a
predicate , and so predicates did not find their way into his language :
A distinction of subject and predicate finds no place in my way of representing a
judgment . . . . [Take ] the two propositions 'the Greeks defeated the Persians at
Plataea ' and 'the Persians were defeated by the Greeks at Plataea ' . . . . Even if a
slight difference of sense is discernible , the agreement in sense is preponderant .
Now I call the part of the content that is the same in both the conceptual content .
Only this has significance for our symbolic language ; we need therefore make no
distinction between propositions that have the same conceptual content . ( 1879/
1952a, 2- 3)

Frege found that retaining the distinction between a subject and a predi -
cate in a formalized language was " obstructive of [his] special purpose "
(p . 4), namely , to model a language of thought on the formalized lan -
guage of mathematics (in which , he said, " subject and predicate can be
distinguished only by doing violence to the thought " (p . 3))- a language
to be used for drawing valid inferences. He therefore dismissed as " slight "
the difference in the sense of an active and a passive statement - a differ -
ence he seems not to have understood , for he thought that the subject is
merely that which occupies the place in the word order that the speaker
wants the hearer " to attend to specially " (p . 3). Frege failed to seethe real
difference in sense: an active predicate signifies action , and a passive one
signifies an undergoing of action . Aristotle characterized these as distinct
types of kategoria , and distinct types of being- and so they are, for to hit
and to be hit , or to defeat and to be defeated, are surely different things .
Unfortunately , Frege's view that active and passive utterances have (more
Scientific Theories That Unconceal Being 199

or less, for his purposes) the same conceptual (or semantic) content came
to be shared by many later thinkers (who saw no difference at all in sense,
irrespective of any purpose). This erroneous way of thinking likely came
to prevail when the modern concept of a relation was incorporated into
the concept of a predicate . Since a relation , in the modern sense (which
includes transitive actions), is not an aspect of the being of any particular
one of the individuals involved (but something outside all of them and
somehow connecting them ), any given relation is often assumed to be
unchanging in nature regardless of which individual involved in it is
chosen as the subject ofa sentence (though the unchanging nature attributed
to it always comes from choosing - sometimes unconsciously or tacitly
- one individual as the subject; with verbs for actions , that individual is
the one acting , so that the nature attributed to the " relation " or " predi -
cate" is active , even when the associated verb is in its passive form ). And
so the sense of defeated can be imagined to be the same in the two sen-
tences The Greeks defeated the Persians and The Persians were defeated by
the Greeks ; the second sentence , like the first , can be taken to be about an
action rather than an undergoing of action . The following consideration
may help highlight the distinct sense of the passive. The passive voice
developed out of the middle voice in ancient languages; the distinct mean-
ing of a passive utterance can thus be seen more clearly through consider -
ing an utterance with a meaning comparable to that of an utterance in the
middle voice : The Persians let themselvesbe (or got themselves) defeated by
the Greeks clearly has a different meaning than The Greeks defeated the
Persians .

The concept of a predicate came to subsume the modern concept of

a relation through David Hilbert ( 1862- 1943) and his collaborators ,
who were the first mathematicians to use the term predicate (Pradikat )
for polyadic relations and Fregean polyadic functions (see Hilbert and
Bernays 1934; Frege ( 1891/ 1952b) called grammatical predicates " func -
tions ," but he did not call polyadic functions " predicates " ). Hilbert and
Ackermann ( 1938/ 1950) described such " relations " as " predicates having
several subjects" (p . 45) , revealing a confusion of subjects with arguments
and a failure to see that the predicate must change with the subject (e.g.,
from defeated the Persians to defeated by the Greeks). They attempted to
justify the new concept of a predicate as follows :
The Aristotelian formalism [based on the structure of a declarative sentence, i .e.,
a predicate attributed to a subject] turns out to be inadequate even in quite
simple logical situations . It is basically insufficient for dealing with the logical
200 McPherson

foundations of mathematics . It fails , specifically , whenever a ,"elation among sev-

eral objects is to be represented symbolically .
This may be clarified by a simple example . Consider the statement : " If B lies
between A and C, then B also lies between C and A ." . . . In the . . . [monadic
predicate calculus , the statement ] may in fact be formulated thus : " If an ordered
triple of points has the property that the second point lies between the first and
third , then it also has the property that the second point lies between the third and
first ." This formulation , however , fails to express the logical essence of the state-
ment , namely , the symmetry with respect to A and C of the relation " between ."
Therefore , it cannot be employed to derive the mathematical consequences of the
statement under consideration ....

. . . Since the foregoing calculus has turned out to be inadequate , we are forced
to seek a new kind of logical symbolism . For this purpose we return to that point
in our discussion at which we first went beyond the sentential calculus . The deci-
sive step there was the division of sentences into subject and predicate . . . . [ We
now ] separate in the rendering of a sentence the objects (individuals ) from the
properties (predicates ) attributed to them and . . . symbolize both explicitly .
This is done by employing functional symbols with argument places (n-adic func -
tional symbols where n is the number of argument places) for the symbolic ren-
dering of predicates , in which symbols representing objects are to be substituted in
the argument places. . . . If the relation of the smaller to the greater is expressed by
the two -place functional symbol , ), then 2, 3) is the symbolic rendering of
the sentence " 2 is less than 3 . " Likewise , the sentence " B lies between A and C "
may be rendered by Z (A , B , C ) .
All mathematical formulas represent such relations among two or more quan -
tities . For example , to the formula x + y == z there corresponds a triadic predicate
S(.,"\", y , z) . The truth of S(x , y , z) means that x , y , and z are connected by the
relation .,"\" + y == z. ( 1938/ 1950, 55- 57)

They pointed out the novelty in their use of the word predicate for the
modern " relation " in a footnote and without explanation :

Hitherto it has been customary in logic to call only functions with one argument
place predicates , while functions with more than one place were called relations .
Here we use the word " predicate " in a quite general sense. ( 1938/ 1950, 57, fn . 1)

Hilbert and Bernays first introduced the new usage of the term as follows :

We understand " predicate ," here and in what follows as well , always in a wider
sense [than the usual], so that even predicates with two or several subjects are
included . Depending on the number of subjects, we speak of " one-place," " two -
place" . . . predicates . ( 1934, 7)

This extension of the term predicate , so casually introduced , was un-

fortunately accepted by later mathematical logicians , apparently without
question , and it has even influenced views on predicates and predication
outside of mathematics (e.g., in philosophy of language , linguistics , and
Scientific Theories That U nconceal Being 201

psychology). As a result, many modern scholarsthink a predicateis pred-

icated of all the arguments of a verb heading an actual (i .e., a gram-
matical) predicate. And so we find Rispoli (1995) saying that in the
sentenceI took a spoonfrom the drawer, " took from is a predicate pred-
icated of three arguments: I , spoon, and drawer" (p. 333). In reality, a
different predicate is predicable of each of the three noun phrases, for
the type of being that can be shown forth from each of the individuals
involved in the event is different in each case; the predicate took a spoon
from the drawer is predicable of I , takenfrom the drawer b}' me is predi-
cable of a spoon, and theplacefrom which I took a spoonis predicable of
the drawer.
Modem linguists have a number of different opinions about the nature
of predication. Somedescribeit simply as " saying something" about the
subject, and others claim that " the speakerannouncesa topic and then
says something about it " (Hockett 1958, 201), a view sometimesunder-
stood in terms of the subject's or topic's expressionof given information
(i .e., information already given in the discoursecontext) and the predi-
cate's introduction of new information . The idea that a predicate " says
something about" a subject is far too vague to qualify as a theory. The
reduction of the subjectand the predicateto a topic and a comment upon
it seemsmisguided, for in the sentenceThat idiot wrecked my car, the
subject noun phrase seemsto comment on the character of the subject,
not just to point to an individual about whom somethingwill be said. But
perhaps the most seriousproblem with this view, as well as with the view
that a subject provides infoffilation given in the discoursecontext and a
predicate provides new infoffilation , is that it does not account for the
facts of predication, such as the fact that subjectstend to be ousiai, that is,
'beingnesses ' or substances, that primary beingnessesor individuals can-
not be predicated of anything other than themselves , that generaare pre-
dicated of speciesbut speciesare not predicated of genera, and that being
that is dependentupon a substratefor its existence, such as a quality, a
relation, or an action, can be predicated of beingnessesbut beingnesses
cannot be predicated of any such dependentbeing.
The modem concepts of a predicate and predication are so sense -
depleted that modern linguists rarely find any place for them in their
theories, and when predicatesdo find a place in linguistic theories, they
are typically treated as purely syntactic entities.
In postclassicaltimes, new conceptions of predicates and predication
seemto have been shapedlargely by philosophical trends rather than the
202 McPherson

nature of the being a predicatesignifiesor its relation to the being a subject

signifies. The concept acquired a new layer of sediment when it was
incorporated into symbolic logic. The aim of modeling a logical language
after a mathematical language, such as the language of arithmetic or
algebra (Boole and Frege) or the calculus of sets(Peano), and the aim of
creating a languagefor the deduction of mathematical consequences (e.g.,
Whitehead and Russell, Hilbert ) blurred the distinction betweena declar-
ative sentenceand a mathematical formula; and a rash, ill -considered
decision to use the term predicate for a Fregeanfunction followed by the
almost universal, unquestioning adoption of this use of the term led to
widespreadforgetting of the distinction between a predicate and both a
Fregean function and a relation in the modern, sedimentedsenseand
of the distinction betweena subject and an argument. Contemplation of
predicateshas been conspicuouslyabsentin the creation and adoption of
new concepts of a predicate. Geach (1962), regarding a class-inclusion
interpretation of predicatesin statementsof a certain kind, commented
that " writers hurry over the topic, as over a thin patch of ice" (p. 15); the
samecan be said about discussionsof predication in general.

9.8 SomeConsequences
of Sedimentationfor Scientific Theories

It is one thing for people who do no scientific work to let sedimented

conceptsguide their thinking ; it is another thing for a theoretical scientist
to use words or signs tied to sedimentedconceptsin the construction of
theoretical statements. When a concept becomessedimentedso that it no
longer reflects the nature of the being that originally gave rise to it , any
theory shapedby that concept will fail to permit a beholding or uncon-
cealmentof that being and will actively concealit . The theory may reflect
well the thinking of the theorist, but it will be of little value if the being it
revealsis only that of entia rationis in the minds of the theorist and his or
her followers.
A consideration of some specificconsequencesof using the sedimented
conceptsanalyzed above will help to show how sedimentedconceptscan
impede the formation of true theories- that is, theories that unconceal
being.
Psychologistsand philosophers of mind often aim to distinguish un-
learnedfrom learnedconcepts(or mental representations ) . But discussions
about whether a conceptis unlearnedor learnedcan only be meaningful if
the original senseof the concept is known- for this is the only sensethat
Scientific Theories That Unconceal Being 203

could be unlearned. Consider, for instance, the concept of number. The

modern concept of number as continuous emerged in human thought
only within the last few centuries. It is highly unlikely that it is unlearned,
or that we have somespecialunlearnedcapacity just for understandingor
using this sedimentedconcept. For this reason, arguments in favor of
unlearned mental representationsof numbers that are continuous, with
any whole number representedas an extent or magnitude that is a multi -
ple of a unit magnitude (e.g., Meck and Church 1983), would seemto
require careful examination.
Becausesedimentation hides the original senseof a concept, or the
sensemost likely to have been conceivedthrough first intentions and to
reflect the nature of being, doors that would be open to a theorist familiar
with the original senseare closed. But removing sedimentedlayers from
one's conceptto reveal the original sensecan open thosedoors. For exam-
ple: if numbers are understood first-intentionally as counts of individuals
of the same kind, then unlearned representationsof numbers (if there
are any) may be counterparts for collections of individuals (much like
Roman numerals), with collections of like symbols representingcollec-
tions of individuals of any kind whatsoeverwith the same cardinalities.
(Wynn (1992) presentsevidenceconsistentwith the existenceof unlearned
representationsof numbers that are counterparts of this sort.)21
Sedimentedconcepts can lead theorists down blind alleys. Take, for
instance, the sedimented concept of a predicate adopted by modern
mathematiciansand those influenced by them, which subsumesthe sedi-
mented concept of a relation. This conceptled somelinguists down a path
that is now fairly widely recognizedas a dead end. The mathematician's
(polyadic) " predicate" is a relation in the modern sense, so that it is
imagined to apply equally and simultaneously to all the arguments of
a verb that heads a grammatical predicate (i .e., to be something that
connects them together somehow, with a nature that remains the same
regardlessof which argument is chosen as the subject of a sentence ).
Linguists who adopted this sedimentedconcept were blinded to the dif-
ferencein senseof an active and a passivesentencecontaining the same
verb and the same noun phrases but with a different noun phrase in
subjectposition and possibly a different verb form. They thought that any
sentencewith a predicateheadedby the verb defeat (even when it is in its
passiveparticipial form) must be about the action of defeating (and so
they made a distinction betweenthe " grammatical" or actual subjectof a
sentenceand the " logical" subject, which is the noun phrasefor the indi-
 204 McPherson

vidual carrying out an action - but this reveals an inconsistent vestige of

the Aristotelian view according to which a predicate signifies being that
originates in the subject) . They failed to see that a passive utterance is
about the undergoing or suffering of an action , which is being of a distinct
type . For this reason, they regarded a passive sentence as an active sen-
tence that has undergone a " transformation " in which the words were
rearranged , the verb may have changed form , and a few elements were
added (e.g., is or was and by in English ) without any accompanying
change in meaning . This way of thinking is tied to the invention of two
notions : ( 1) transformations in language processing (and so Chomsky
( 1965) said that sentences containing passive participles , or sentences in
which the " logical " subject and object differ , in this way of thinking , from
the " grammatical " subject and object , " provide the primary motivation
and empirical justification for the theory of transformational grammar "
( p . 70)), and (2) a " deep structure " (later " D -structure " ), which contains
the untransformed version of a sentence- a notion supplemented by the
" Katz -Postal principle ," or the idea that the meaning of the sentence in
deep structure is unaffected by the transformation that produces the
" surface-structure " sentence. Though the notion of deep structure was
rejected by some linguists not long after its emergence in Chomsky 's
( 1965) theories (e.g., in some papers first circulated in 1967: Lakoff and
Ross 1976; McCawley 1968), and though the accompanying Katz -Postal
principle was soon rejected by Chomsky and his followers (e.g., Jackendoff
1972), deep structure (as D -structure ) remained a part of Chomsky 's
popular theories for many years (e.g., Chomsky 1991), albeit with pro -
gressively less prominence (see, e.g., Lakoff 1980) and a progressively
more restricted role with regard to meaning ; Chomsky (e.g., 1995b) only
recently eliminated it from his theories . The notion of a transformation ,
though also questioned early on (e.g., in 1967 by Postal ; see Postal 1976),
still survives in Chomsky 's theories (e.g., Chomsky 1995b), even though it
has a greatly diminished role in the grammar (and has largely fallen out of
favor with other linguists as proposed alternatives to a transformational
grammar have come into favor ) . Much ink will have been spilled in vain
over this notion and the notion of deep structure before they are both
finally abandoned .

9.9 Generating Theories That V nconceal Being

Let us not fool ourselves. All of us, including those who think professionally , as
it were, are often enough thought -poor ; we all are far too easily thought -less.
Scientific
Theories
ThatUnconceal
Being 205

Thoughtlessnessis an uncanny visitor who comesand goeseverywherein today's

world. For nowadayswe take in everything in the quickest and cheapestway, only
to forget it just as quickly, instantly. . . .
. . . Man today is infiightfrom thinking. This flight-from-thought is the ground
of thoughtlessness .
Martin Heidegger, Discourseon Thinking

If theoretical statementsare to be genuinely apophantic, unconcealing

being and permitting a noetic beholding (theoria) of being, theorists must
adopt and use scientific conceptsand terms much more consciously and
deliberatelythan is usual nowadays. They must not allow themselvesto fall
an easyprey to the pervasiveforgetfulnessof sensedepletion and sedimen-
tation- to that which only seemsto be- but must pursue its opposite, the
unforgetting of well-rounded unconcealmentor full understanding.
Concepts should not be ~dopted passively. Rather than trusting other
scholarsto have properly conceiveda second-intentional concept through
reflection upon first-intentional concepts, and rather than having faith
that the concept remains unsedimented, a theorist should compare any
second-intentional concept used with the first-intentional concepts that
originally gave rise to it to seewhether the former adequatelyreflectsthe
latter, and should compare the first-intentional concepts with the being
that gaverise to them. If a theorist is using the second-intentional concept
of a relation, for instance, concepts of individual types of relation (e.g.,
the ratio double, the relation taller than) can be examined to seeif the
theorist's concept of a relation reflects something in the nature of a rela-
tion of each type as it is conceived- and, more importantly , as it resides
in being.
In addition to this sort of reflection, a theorist can attempt to unearth,
through historical or etymological investigation, the original senseof any
concept guiding the thought behind a theory, and to dig up whatever
layers of sensehave been depositedover it to conceal it . But the theorist
should also determine whether the original sensereflects authentic con-
ceptionsof being originating in first intentions, and this determination can
be made only through a contemplation of being.
Contemplation of being servesa purpose other than restoring meaning
to sedimentedand sense -depleted concepts: it can also lead to new con-
ceptions so that theories are freed from the limitations of conceptsthat
happen to be ready to hand.
Becauseapophantic (or declarative) sentencescan unconceal being, a
theorist attempting to conceivebeing anew through first-intentional con-
206 McPherson

templation might benefit from embeddinga word for the being of interest
in true declarative sentences . And sinceapophantic sentencesare the only
linguistic entities that can unconcealbeing, semanticiststrying to discover
the meaning of a word or the schema of meaning for a class of words
(e.g., a class of words belonging to one part of speech) should always
embed the word (or an exemplar of the part of speech) in declarative
sentencesthat they judge to be meaningful (such as Freddie is a dog) and
try to becomeaware of the nature of the being signified.
Becausestatementsin standardmodern mathematicalnotation (whether
algebraic or logical or of another sort) are not (or are not properly)
modeled after apophantic utterances, theoretical work should begin with
nonmathematical statementsand reasoning. When the being under inves-
tigation has been conceived, properly speaking, and a theory developed,
mathematical formulations can then be introduced- but they should be
introduced only with great caution, with the theorist making sure that in
formalizing a theory, he or sheis not importing sedimentedconceptsinto
the theory by virtue of the formalization. Further, if contemplations of
being reveal inadequaciesin existing mathematical conceptsand notation,
mathematical languagescan be revised or constructed so as to provide
adequateformal tools for theorists (as Macnamara (1994) argued).
All of theserecommendationspoint in the samedirection. Becausethe
mind is only in direct (noetic) contact with being in first intentions, these
must be the foundation of theoretical work . If theories are not anchored
to being by first intentions, they will drift in whateverdirections the imagi-
nation and philosophical currents take them. The result is usually a series
of theoretical fads (especiallyin psychologyand the social sciences - in the
Geisteswissenschaften - where the being under investigation- intentions,
as well as the actions that follow from them- is not, strictly speaking,
sensible, and so it does not force one's mind to turn toward it ; it even
resistsits own unconcealmentbecauseintentions are directed away from
themselves , and, being the \JTerystuff of consciousness , their constant
presencein one's waking psyche leads one to take them for granted so
that they typically escapeone's notice).22Faddish theoretical frameworks
often stimulate fruitless debates about issues irrelevant to actual phe-
nomena, which retreat into concealmentto the degreeto which the posi-
tions taken in the debates are accepted as guides to one's thinking . A
genuineresolution of such debatesis rarely found (even though one posi-
tion may come into favor by virtue of the strength of the personalities
supporting it); far more frequently, the debated issuesare forgotten as it
becomesfashionableto follow new currents.
Scientific Theories That U nconcealBeing 207

If theories are to be about anything other than human imaginings ,

theorists must develop their capacity for first intentions and make a habit
of them . Having done so, they will be far more likely to adopt theoretical
terms , and concepts of the being they signify , that peffi1it being to be
unconcealed ; moreover , they will be far more likely to combine those
terms in structures that reflect the constitution of the unconcealed being .

Notes

1. Edmund Husserl and Jacob Klein made use of many of these insights , either
explicitly or implicitly , in evaluating theories in modern physics and mathematics
(see Husserl1954 / 1970; Klein 1934- 1936/ 1968, 1932/ 1985) .
2. I use being as a general term for that which is, including existent and sub-
stantial beings (where beings is comparable to the Greeks ' ta onta ; it does not
necessarily refer to living things ) as well as matter and the being that inheres in ,
originates in , or otherwise depends upon beings or matter , such as qualities ,
quantities , relations , actions and passions, locations in space and time , and so on
(and being is here understood as a mass noun , comparable to the Greeks ' to
on) . What I refer to as being is not identical with the " Being " of Heidegger and
the Greek philosophers who inspired him . That Being is the proper subject of
philosophy (see Aristotle , Metaphysics r .1, 1003a21- 22; Heidegger 1927/ 1962);
theories in the special sciences deal only with parts of Being ( Metaphysics r .l ,
1003a22- 26). Wherever I use being as a count noun , it implies an individual
informed or individuated atomically by some form (or species), and where I use
being as a mass noun , the context should make clear whether I mean matter or
dependent being alone or whether I mean beings, matter , and dependent being
together .
3. For other mental phenomena , an intention arises under different conditions ,
and it is a qualitatively different kind of aiming at something as an object , origi -
nating not through activation of the noetic faculty , but of some other faculty of
the psyche. In a state of desire, for instance , a person intends something or aims
his or her mind at something by longing for it . In an instance of the perception of
sensible things , he or she intends something by perceiving it sensibly :---"
The notion of an intention entered into Scholasticism through the Islamic phi -
losophy of Al -Farabi and Avicenna . Like intentio , the word these philosophers
used for it , ma 'na (or sometimes ma 'qul ; see Engelhardt 1976), means concretely
the drawing or stretching back of a bow (which brings it into a state of tension ) in
aiming an arrow at a target . This word brings together in one concept the ideas
of aiming and tensing . The tensing that accompanies the aiming might best be
understood as the mind 's standing firm in its aiming - its tensing in that sense. But
the characterization of an intention as a tensing also brings out the fact that it can
vary in intensity , intensifying and remitting ; the standing firm in the aiming can
be more or less firm . Desires , fears , loves , angers , and so on , can be more or less
intense, and thoughts too can in some sense intensify and remit , being entertained
more or less intently , or , one might say, involving more or less attention .
208 McPherson

4. Some theorists (e.g., Chomsky ( 1992, 1995a)) seem to think that a carving up
that involves a human mind necessarily implies that what is carved up cannot be
outside the mind . But the involvement of the mind in the carving no more implies
that the stuff being carved is internal to the mind than the involvement of arms ,
hands, and an axe in chopping up the wood of a tree (an act that violates the tree's
natural boundaries ) implies that the stuff being chopped up does not exist in the
world outside the instruments used in chopping - the arms , the hands , and the
axe. If the members of a speech community collectively delimit , bound , or define
certain beings- say, the beings called DANCERS (defined as people who
dance)- this does not imply that those beings exist in our minds rather than in the
extrapsychic world ; dancers, like the dancing that makes them dancers, and like
the people who underlie them , exist outside our minds (and , unlike mental beings,
may be found performing on a stage) . TIle fact that dancers (as well as firms , con -
sumption , and income ) are not beings with metaphysical primacy (in the sense of
Aristotle 's prole ousia, 'primary beingness' or primary substance)- that is, beings
that underlie all other being - does not mean that they are not extrapsychic beings.
The example of dancers serves to highlight another point . In comparing the carving
up of being to chopping wood , I do not mean to imply that the being to be carved
up is necessarily matter . A dancer is carved from a person , coming into being
when the person takes up dancing , and passing away when the person ceases to
dance. A person is not a portion of matter (though a portion of matter coincides
with a person at any given moment ); the matter in a given person 's body changes
throughout the lifespan . Nor is a person the mereological sum of time -slices of all
the matter that is, was, or will be in his or her four -dimensional (space-time ) body
(a view that Putnam ( 1994) attributes to David Lewis ), for the summation of the
portions of matter that happen to coincide with a given person at different times
presupposes the person , the identity of whom guides the summation .
Being can also be defined relative to other being without implying that it is
unreal or imaginary . If Patricia is taller than Katherine , the fact that reference to
Katherine is necessary to determine Patricia 's being taller than her does not mean
that Patricia 's relative tallness is not really an aspect of her being . And being can
be defined with respect to individual observers and yet reside outside their heads .
That beauty is imaginary is not implied by the fact that beauty is relative to
observers (" in the eye of the beholder " ). If some people judge a painting to be
beautiful , then the aspects of its being that give rise to that judgment are the
components of the beauty they perceive . ( This idea is reminiscent of Locke 's
( 1694) view that to sayan object is red is to say it has the power , by virtue of the
nature of its surface, to give rise to a certain sensation in us. But unlike Locke , I
would not deem redness unreal and the physical property that makes something
look red real ; I would identify the two instead . As Berkeley ( 1710/ 1982) pointed
out , all qualities - not just those that Boyle and Locke called " secondary " - are
experienced the way they are because of properties of the observer . But I do not
conclude , as Berkeley did , that all qualities are secondary and there is no reality
outside our experiences; I conclude , rather , that most qualities are " primary " and
the peculiar and mutable nature of our experiences of them is irrelevant to their
reality . The fact that the appearances we experience may not resemble or be simi -
Scientific Theories That Unconceal Being 209

lar to the qualities that give rise to them does not means that those qualities are
not bona fide qualities of being (cf . Putnam 1981) .)
5. Most of the semanticists who adopt the " It ' s all in your head" view (or what
Jerry Fodor calls " psychosemantics
" )- the view that all linguistic meaning arises
somehow from mental representations - seem, like eliminativists , to be guided pri -
marily by skepticism about the mind 's manifest ability to intend being or beings ,
skepticism motivated by faith in a physicalistic ontology , which , as it is currently
conceived , excludes this ability (see also Baker 1987; Morris 1991) . It is skepticism
of this sort that led Fodor ( 1987) to say, " If the semantic and the intentional are
real properties of things , it must be in virtue of their identity with (or maybe
of their supervenienceon?) properties that are themselvesneither intentional nor
semantic . If aboutness is real , it must be really something else" (p . 97)- which I
would answer with Joseph Butler 's observation that " everything is what it is and
not another thing " (see Kripke 1972, 94) .
Fodor attests to the fact that many cognitive scientists and philosophers " are
repulsedby the idea that intentionality is a fundamental property of mental states
(or , indeed , of anything else- talk about your ontological dangers!)" ( 1990, 313).
He points out that " the deepest motivation for intentional irrealism derives . . .
from a certain ontological intuition : that there is no place for intentional catego-
ries in a physicalistic view of the world ; that the intentional can't be naturalized "
( 1987, 97) .
Such ontological dogmatism seems inappropriate in science, where being comes
to be understood by means of observation and evidence. Evidence from inner
perception and observation of people 's use and interpretation of language clearly
reveals the capacity of the mind to intend being outside the mind in its thoughts ,
perceptions , desires, and so on . If , for instance , a girl asks a salesclerk for an ice-
cream cone, both the girl and the salesclerk understand perfectly well that the girl
desires an ice-cream cone, not a mental representation of one.
6. The dependence of an isolated word 's potential to signify upon a concept or
definition may be one source of the belief of many semanticists that utterances of
all sorts get their meanings solely by virtue of mental representations(where
mental representation seems to be understood as a general term for mental entities
of various sorts, including concepts and definitions ) . Semanticists typically take a
compositional view of sentence meaning , seeing the meanings of sentences as
structured aggregations of the meanings of the words in the sentences. If words
considered individually do not permit the mind to intend being , but have a
" sense" or " meaning " only by virtue of a concept or definition , then a composi -
tionalist could be led to believe that the meanings of sentences are combinations of
mental representations .
But even if a theorist is only interested in the " senses" of individual words - the
mental representations (or concepts or definitions ) of the being they have the
potential to signify - the theorist must nonetheless characterize the being the words
would signify in declarative statements (or , more generally , in referring expres-
sions), for the simple reason that mental representationsare representationsof
being ; one cannot characterize the representations until one has characterized the
being or beings they represent .
 210 McPherson

7 . This translation and the other translation of Brentano ' s German are my own ,
as are all the translations in this chapter from classical Greek (i .e., of Aristotle ,
Euclid , Eudoxus , Nicomachus , Parmenides , Plato , and Porphyry ), Latin (i .e.,
of Aquinas , Bradwardine , Descartes , Eustachius a Sancto Paulo , Fredericus ,
Peletier , Stifel , Suarez, Wallis , and Zabarella ), and French (i .e., of Leibniz
and Stevin) . Translations of German authors other than Brentano (i .e., Dede -
kind , Frege , Heidegger , Hilbert and Ackermann , Husserl , and Klein ) are from
the published English editions cited , except for the translation of the passage from
Hilbert and Bernays , which is my own . Wherever a quotation that I have trans -
lated into English is incorporated into the text , I have enclosed it in single quota -
tion marks to distinguish it from direct quotations from works published in
English , which are enclosed in double quotation marks .
8. 'Being of reason' is the standard (and literal ) translation of the Scholastic ex-
pression ens rationis . It refers to a being , such as the concept of a universal , that
exists in the mind (and perhaps only in the mind ) .
9. The Greeks had no word meaning 'number ' per se; their word arithmos is best
translated as 'count ' or 'counting ' (depending on the context ). The Greeks did not
understand numbers to fall under a single idea corresponding to our number
concept , for they did not believe that things that have an order can fall under
a common idea, definition , or genus (see Aristotle , Nicomachean Ethics A .6,
1096a17- 19; On the Soul B .3, 414b22- 28; Metaphysics B .3, 999a6- 10) . And ,
indeed , if numbers are understood as counts , it does not make sense to think of
them as belonging to a genus, since each one contains within it , or is a successor
of , all smaller ones. To count 5 beings, one must first count 2 beings, then 3 beings,
then 4 beings, successively adding 1 being to the collection of beings already
counted . Two is not separate from 3, or from 5 , but rather a part or antecessor of
it . And so even though each number is a species (with , e.g., all counts of 3
belonging to one species), the different species of number are not independent
species that can be included under a common genus.
10. For Plato , monads were separate, nonsensible , noetically perceived beings,
real beings in the realm of the mathematicals ; but , as Klein ( 1934- 1936/ 1968)
pointed out , " the emphasis with which the thesis of 'pure ' monads is propounded
is indicative of the fact that arithmoi were ordinarily , and as a matter of course,
understood only as definite numbers of sensible objects " ( p . 70).
11 . The idea that a monad is that which remains in an atom when all but its oneness

is abstracted away is also consistent with Aristotle ' s comment that a monad , like a
point , is indivisible in any dimension , but whereas a point has position , a monad
has none (Metaphysics A .6, 10I6b24 - 31; and so he calls a monad a point without
position , stigme athetos; Metaphysics M .8, IO84b26- 27); a point is what remains
in a being when its species, its sensible qualities , its matter , and its extension in
every direction are mentally stripped away (Metaphysics K .3, IO6Ia29 - 35; and so
Aristotle describes the objects of geometry as physical , e.g., physical points or
lines drawn on papyrus , but not qua physical ; see Physics B .2, 194a9- 11) .
12. Wallis said, 'Geometry is more or less subordinated to arithmetic , and to that
extent it applies universal assertions of arithmetic specially to its objects ' (Mathesis
Scientific
Theories
ThatUnconceal
Being 211

Univer salis; for the Latin , see Wallis 1657/ 1695, Opera, vol . 1, as cited in Klein
1934- 1936/ 1968, 216) .
13. Descartes ( 1596- 1650) stated that the geometric figures used in his algebraic
geometry as counterparts for magnitudes or quantities of any sort whatsoever (see
Schmidt 1986) 'must exhibit at one time continuous magnitudes , at another time
multitude or number also' ( 1701/ 1966, 452) . By introducing a unit magnitude with
which other magnitudes could be compared , he was able to interpret magnitudes
as numbers : 'Thanks to the unit we have assumed, continuous magnitudes can in
some instances be reduced in their entirety to multitude ' (pp . 451- 452); he also
said, 'If number be the question , we imagine a particular subject [i .e., a figure ]
measurable by a multitude of units ' ( p . 445), and , 'That very division into a plu -
rality of equal parts , whether it be real , or only intellectual , is properly a [mode of ]
measuring in accordance with which we count things ' (pp . 447- 448) .
14. In a later era, Richard Dedekind ( 1831- 1916), who regarded numbers as " free
creations of the human mind " ( 1888/ 1963, 31), stated explicitly that the goal of
introducing irrational numbers is to permit numbers , as points , to fill up a straight
line . After comparing the fact of one rational number ' s being greater than another
to a point ' s being to the right of another on a straight line , and of its being less
than another to a point ' s being to the left of another , and similarly , he said,
To every rational number a, i .e., to every individual in R [the domain of rational numbers ],
corresponds one and only one point p , i .e., an individual in L [the points in a line ]. . . .
Of the greatest importance , however , is the fact that in the straight line L there are
infinitely many points which correspond to no rational number . . . . The straight line L
is infinitely richer in point -individuals than the domain R of rational numbers in number -
individuals .
If now , as is our desire, we try to follow up arithmetically all phenomena in the straight
line , the domain of rational numbers is insufficient and it becomes absolutely necessary that
the instrument R constructed by the creation of the rational numbers be essentially improved
by the creation of new numbers such that the domain of numbers shall gain the same com -
pleteness, or as we may say at once , the same continuity , as the straight line . (1888/ 1963, 8- 9)
The use of the points on a line as a model for numbers was so customary by his
time that Dedekind felt no need to justify the assumption that for every point on a
straight line there must be a corresponding number .
15. Aristotle used these words as if they were standard terms , not commenting on
his use of them , so he may have been borrowing from and developing an existing
theory . But by the third century C.E. at the latest , the words for predicates and
predication that Aristotle used were deemed to have been of his own choosing .
Porphyry (232- 309), in his commentary on the Categories , asked,
Why , since kategorias in customary language is said , in respect of the forensic speeches, of
the [kategorias ] in accusation , to which is opposed the speech for the defendant , did Aristotle ,
not having undertaken to instruct us about how to speak against opponents in the courts of
justice , but [ about ] something else, which very thing is not called by this name among the
Hellenes , choose to make strange [use of language ] by entitling the book kategorias ? (265/
1887, 55)

16. Modem English editions of Aristotle 's logical writings often give 'belong to '
as the translation of huparch6 . Alternative and even more unrevealing translations
212 McPherson

are 'apply to ' and 'hold of ' , which do not correspond to any meanings the word
ever had for the Greeks .

17. As Plato pointed out (Sophist 262d- 263a), a simple declarative sentence such
as Theaetetus sits has two aspects, which he captured in the expressions peri hou
esti and hotou esti, 'whom /what it is about ' and 'whosever it is' : ( 1) Viewed as a
rhema (a verb or an adjective ) mingled (kekramenon ) or intertwined (sumplekes)
with a noun (or as a predicate intertwined with a noun phrase), the simple sen-
tence reveals something about something (i .e., the subject , or Theaetetus in this
case)- some determination (logos) of the subject . (Plato said that intertwining a
rhema with a noun tiperainei , 'brings something to an end', 'limits something ', or
'determines something ' , and that for this reason we call this combination of in -
tertwined word a logos (see 262d) . Robert H . Schmidt has argued , in conversation ,
that the original meanings of lego as 'lay (to rest)' and 'gather (together )' suggest
that a logos was understood as a gathering together that lays to rest. The gathering
together of words in an intertwining of a verb with a noun (through conjugation
for person and number ) gives a sense of completion and thus brings the mind to
a rest or standstill , whereas a word or a phrase by itself does not do so. A verb
intertwined with a noun determines or brings to an end in that sense. The inter -
twining of a verb with a noun , or of a predicate with a noun phrase, also limits or
defines the being of the subject that is shown forth from the subject .) (2) Viewed as
that which has laid down together (suntheis) a noun (or a noun phrase) with a
rhema (or a predicate ), a logos (or 'determination ') is seen to belong to something ,
or be of something (i .e., the subject , Theaetetus) . The two aspects correspond to
two roles as answers to questions of two sorts, exemplified by " What is Theaetetus
doing ?" (or " What is it about Theaetetus ?" ) and " Who sits?" (or " Who is it who
sits?" ) . Both can be answered with " Theaetetus sits" ; this utterance both reveals
something about Theaetetus (namely , that he sits) and lets it be known that the
particular determination of the subject (namely , the sitting ) is the determination of
Theaetetus (the possessor of the sitting ). But since predication , or affirming or
denying something about a subject , is the intertwining of a predicate with a (pre-
supposed) noun phrase, it brings into being an utterance with the former role or
aspect- one that reveals something about the subject .
18. This description reveals the error in the claim of scholars , at least since
Boethius (480- 524), that the categories other than ousia- literally 'beingness' but
conventionally translated as 'substance'- are supposed to be categories of " acci-
dental " being , for Aristotle explicitly contrasted the being of the categories , or
being that is kalh ' hauto , with accidental being , or being that is kala sumbebekos-
'because of (or in accordance with ) concomitance ' (see Metaphysics 11.7).
19. Here , phasis seems to mean primarily 'denunciation ' or 'a charge laid '- a
bringing to light in that sense
; the prepositional prefix kata in kataphasisimplies a
charge being laid against someone (or something ), and apo, which appears also in
apologia and apologeomai ('speak in one's defense'), as well as in apopsephizomai
('vote a charge away from ' or 'acquit '), implies a charge being removed .
(Alexander of Aphrodisias , in his commentary on Aristotle 's Topics , said that
apophasis is anairesis, which means 'taking up ' or 'taking away ' ; in legal contexts ,
Scientific Theories That Unconceal Being 213

it means the quashing of an indictment .) Aristotle said that 'kataphasis is apo -

phansis [i .e., showing forth ] of something [laid ] against something , and apophasis
is apophansis of something [i .e., some charge] removed from something ' ( On Inter -
pretation 5, 17a25- 26).
20. When sentences of other types contain the verb meaning 'be' (explicitly or
implicitly ), it plays some other role regarding being . In a question , it indicates that
being can be unconcealed in the response the question is intended to elicit . In a
command (e.g., " Be kind " ; " Don 't be cruel " ) , a request (e.g., " Please be quiet " ),
or a prayer (e.g., " Be merciful to him " ), where there is no subject , it indicates that
being of the type desired (or desired to be free of ) can get brought into existence
(or taken or kept out of existence) if the addressee complies , thereby becoming (or
ceasing to be, or ensuring he or she is not ) the subject or origin of the being (so
that , in the case of desired being , the verb meaning 'be' can be conjugated to agree
with that subject in an affirmative declarative statement unconcealing the being
from that subject). Among clauses with the structure of a declarative sentence, in
some (such as those that follow because after verbs such as quit ) the copula in -
dicates that being is unconcealed from the subject of the clause, but in others (such
as those that follow that , whether, or ifin the complements of verbs such as hope,
believe, wonder, say, and report ) the copula indicates that being would be uncon -
cealed from the subject of the clause if it were true .
21. It is possible that we have no unlearned representations of numbers per se, but
rather an unlearned ability to detect sameness and difference in the cardinalities of
collections of beings, combined with an unlearned ability to understand something
as a counterpart for something else (which is similar to but not identical with the
ability to understand an image as an image , i .e., as transparent rather than opaque ,
and let one's mind fall upon that of which it is an image through assenting to the
image 's announcement of it , rather than being spellbound by the image so that
one's mind stands firm in an intention toward the image itself ; the Greeks called
this ability eikasia) . These abilities would permit us to use any image of a collec -
tion of individuals of any kind held in the imagination as a counterpart for other
collections with the same cardinality , much as when we use our fingers , in count -
ing , as counterparts for individuals in a collection of beings of some other kind .
The ability to understand something as a counterpart for something else ( which
permits us, for example , to use a map as a counterpart for a region ) seems to be
present from an early age, for young children learning common nouns (along with
the adults teaching them ) seem to understand pictures of , say, animals in books as
counterparts for real animals , so that the children take the nouns applied to the
pictures by the adults to be names for species of real animals .
As the examples of a map , pictures in books , and fingers used in counting show ,
representations used as counterparts need not be internal to the mind (see also
Houghton 1997), as is usually supposed.
22. Perhaps some of the best examples of such fads are to be found in psy'chology ,
where the views of the mind that have dominated theories are , for the most part ,
metaphorical : the mind as chemical compounds (e.g., Locke 's complex or com -
pound ideas formed out of simple ideas, Wundt 's and Titchener 's structuralism ) ,
214 McPherson

the mind as a mechanical system (e.g., Hume 's associationism , where " ideas" are
governed by laws of " attraction " or association analogous to Newton 's law of
gravitation ), the mind as " Poof !" (behaviorism , or the mind 's retreat into com -
plete concealment ), the mind as a computer (the " information -processing " model
of the mind , functionalism ), and the mind as a brain (" neural -network " models ).
Psychology can only begin to be a genuine science when theoretical approaches
based on such metaphors - crude attempts to naturalize psychology - are aban -
doned and we begin in earnest to study the mind as a mind - when the nature of
the being in psychology 's domain (namely , intentional being) is allowed to guide
our choices in classifying phenomena , our choices of terminology , our choices of
problems to address, and the ways we choose to go about solving them , including
methods and theoretical approaches , as that being gradually unconceals its nature
to us (as Husserl ( 1954/ 1970) urged ) .

References

Aquinas , St. T . ( 1970). Summa theologiae [The sum of theology ]: Vol. 17. Psy-
chology of human acts ( T . Gilby , Ed . and Trans .). Cambridge : Blackfriars . ( First
published 1266)
Baker , L . R . ( 1987). Saving belief A critique of physicalism . Princeton , NJ :
Princeton University Press.
Berkeley , G . ( 1982) . .4 treatise concerning the principles of human knowledge ( K . P.
Winkler , Ed .). Indianapolis , IN : Hackett . ( First published 1710)
Bradwardine , T . ( 1955) . Tractatus proportionum sue De proportionibus veloc -
itatum in motibus [Treatise on ratios , or On the ratios of the speeds of motions ].
In H . L . Crosby , Jr . (Ed . and Trans .), Thomas of Bradwardine , his Tractatus
de proportionibus [Treatise on ratios ]: Its significance for the development of
mathematical physics ( pp . 64- 140) . Madison , WI : University of Wisconsin Press.
(Written in 1328)
Brentano , F . ( 1924). Psych% gie vom empirischen Standpunkt [Psychology from
an empirical standpoint ] (Vol . 1, 2nd ed., O . Kraus , Ed .). Leipzig : Verlag van
Felix Meiner .

Chomsky , N . ( 1965) . Aspects of the theory of syntax . Cambridge , MA : MIT Press.

Chomsky , N . ( 1991). Linguistics and cognitive science: Problems and mysteries . In
A . Kasher ( Ed .), The Chomskyan turn (pp . 26- 53) . Cambridge , MA : Blackwell .
Chomsky , N . ( 1992). Explaining language use. Philosophical Topics, 20, 205-
231 .

Chomsky , N . ( 1995a). Language and nature . Mind , 104, 1- 61.

Chomsky , N . ( 1995b). The Minimalist Program . Cambridge , MA : MIT Press.
Churchland , P. M . ( 1981). Eliminative materialism and the propositional atti -
tudes. The Journal o/ Philosophy , 78, 67- 90.
Churchland , P. M . ( 1984). Matter and consciousness: A contemporary introduction
to the philosophy of mind . Cambridge , MA : MIT Press.
Scientific Theories That U nconcealBeing 215

Cummins , R . ( 1996) . Representations, targets, and attitudes . Cambridge , MA :

MIT Press .

Dedekind , R . ( 1963) . Essays on the theory of numbers (W . W . Beman , Trans .) .

New York : Dover . (Original work published 1888)
Dee, J. ( 1975). The mathematicallpraeface to the Elements ofgeometrie afEuclid
of Megara . New York : Science History Publications . ( First published 1570)
Dennett , D . C . ( 1978). Brainstorms : Philosophical essays on mind and psychology .
Cambridge , MA : MIT Press.
Descartes, R . ( 1966) . Regulae ad directionem ingenii [Rules for the direction of
the mind ] . In C . Adam and P. Tannery ( Eds.), Oeuvres de Descal'"tes [Works of
Descartes] (pp . 349- 469). Paris : Librairie Philosophique J. Vrin . ( First published
posthumously in 1701)
Dretske , F . I . ( 1981) . Knowledge and the flow of information . Cambridge , MA :
MIT Press .

Eco , U . ( 1988) . On truth . A fiction . In U . Eco , M . Santambrogio , and P. Violi

( Eds.), Meaning and mental representations ( pp . 41- 59) . Bloomington , IN : Indi -
ana University Press.
Engelhardt , P. ( 1976) . Intentio [Intention ]. In J. Ritter and K . Grunder ( Eds.),
Historisches Worterbuch der Philosophie [Historical dictionary of philosophy ]
(Vol . 4, pp . 465- 474). Basel: Schwabe.
Field , H . ( 1982) . Realism and relativism . The Journal of Philosophy , 79, 553-
567 .

Fodor , J. A . ( 1980) . Methodological solipsism considered as a research strategy in

cognitive psychology . Behavioral and Brain Sciences, 3, 63- 109.
Fodor , J. A . ( 1986). Banish DisContent . In J. Butterfield ( Ed .), Language , mind
and logic ( pp . 1- 23) . Cambridge : Cambridge University Press.
Fodor , J. A . ( 1987). Psychosemantics: The problem of meaning in the philosophy of
mind . Cambridge , MA : MIT Press.
Fodor , J. A . ( 1990) . Psychosemantics or : Where do truth conditions come from ?
In W . G . Lycan ( Ed .) Mind and cognition ( pp . 312- 337) . Cambridge , MA :
Blackwell .

Frege, G . ( 1952a) . Begriffsschrift : A formalized language of pure thought modelled

upon the language of arithmetic (Chapter 1, P. T . Geach , Trans .) . In Translations
from the philosophical wl'itings of Gottlob Frege ( P. T . Geach and M . Black , Eds .,
pp . 1- 20). Oxford : Blackwell . (Original work published 1879)
Frege , G . ( 1952b) . Function and concept ( P. T . Geach , Trans .) . In Translations
from the philosophical writings of Gottlob Frege ( P. T . Geach and M . Black , Eds .,
pp . 21- 41). Oxford : Blackwell . (Original work presented 1891)
Frege , G . ( 1980) . The foundations of arithmetic : A logico -mathematical enquiry
into the concept of number (rev . ed., J. L . Austin , Trans .). Oxford : Blackwell .
(Original work published 1884)
216 McPherson

Garfield , J. L . ( 1988). Belief in psychology : A study in the ontology of mind . Cam -

bridge , MA : MIT Press.
Geach , P. T . ( 1962). Reference and generality : An examination of some medieval
and modern theories. Ithaca , NY : Cornell University Press.
Gilson , E . ( 1979). Index Scolastico - Cartesien [Scholastic -Cartesian index ] (2nd
ed.). Paris : Librairie Philosophique J. V rin .
Hamad , S. ( 1990). The symbol grounding problem . Physica D , 42, 335- 346.
Harvey , C . W . ( 1989) . Husserl 's phenomenology and the foundations of natural
science. Athens , OH : Ohio University Press.
Haugeland , J. ( 1982) . The mother of intention [Review of the book Brainstorms ).
No us , 16 , 613 - 619 .

Heath , T . L ., Sir ( Ed . and Trans .) . ( 1956) . The thirteen books of Euclid 's Elements :
Vol. 2. Books III - IX (2nd ed.) . New York : Dover .
Heidegger , M . ( 1962) . Being and time (J. Macquarrie and E . Robinson , Trans .).
San Francisco : Harper & Row . (Original work published 1927)
Hilbert , D ., and Ackermann , W . ( 1950). Principles of mathematical logic (R . E .
Luce , Ed ., L . M . Hammond , G . G . Leckie , and F . Steinhardt , Trans .). New York :
Chelsea. (Original work , as a 2nd edition , published 1938)
Hilbert , D ., and Bernays , P. ( 1934) . Grundlagen der Mathematik [Foundations of
mathematics ] (Vol . 1). Berlin : J. Springer .
Hockett , C . F . ( 1958) . A course in modern linguistics . New Yark : Macmillan .
Houghton , D . ( 1997) . Mental content and external representations . The Philo -
sophical Quarterly , 47, 159- 177.
Husserl , E . ( 1939). Die Frage nach dem Ursprung der Geometrie als intentional -
historisches Problem [The inquiry into the origin of geometry as an intentional -
historic problem ] ( E . Fink , Ed .). Revue lnternationale de Ph ilosoph ie, 1, 203-
225 .

Husserl , E . ( 1970). The crisis of European sciences and transcendental phenomen-

ology : An introduction to phenomenological philoso]Jhv ( D . Carr ~ Trans .). Evan -
- - ~ ~ , , /

ston, IL : Northwestern University Press. (Original work published posthumously

in 1954)
HusserI , E. ( 1977). Phenomenological psychology (J. Scanlon , Trans .) . The Hague :
Nijhoff . (Original work presented 1925)
Jackendoff , R . S. ( 1972). Semantic interpretation in generative grammar . Cam -
bridge , MA : MIT Press.
Johnson -Laird , P. N . ( 1988) . How is meaning mentally represented? In U . Eco ,
M . Santambrogio , and P. Violi ( Eds.), Meaning and mental representations
( pp . 99- 118) . Bloomington , IN : Indiana University Press.
Klein , J. (1940). Phenomenologyand the history of science. In M . Farber (Ed.),
Philosophical essays in memory of Edmund Husser! (pp . 143- 163) . Cambridge ,
MA : Harvard University Press.
 A11s1 ~
SS~ld
U!~g IU~JUOJU
fl ~ul{.l S~!lO;}l{.l JY!~U~!JS
~Alun pl 'BA1
' BH :YW '~~PI1qUIUJ 'cCJJssaJau
puv .8UJUlVN'(ZL61) 'Y 's '~~dll ) l

' -
 218 McPherson

Macnamara , J . , and Reyes , G . E . ( 1994 ) . Foundational issues in the learning of

proper names , count nouns and mass nouns . In J . Macnamara and G . E . Reyes

( Eds .) , The logical foundations of cognition ( pp . 144 - 176 ) . New York : Oxford

University Press .

Maxwell , J . C . ( 1877 ) . The kinetic theory of gases [ Review of the book A treatise

on the kinetic theory of gases ] . Nature , 16 , 242 - 246 .

McCawley , J . D . ( 1968 ) . Lexical insertion in a transformational grammar without

deep structure . Papers from the Regional Meeting , Chicago Linguistic Society , 4,
71 - 80 .

McGinn , C . ( 1989 ) . Mental content . Oxford : Blackwell .

McPherson , L . M . P . ( 1995 ) . Identifying verbs early in language learning : The roles

of action and argument structure . Unpublished doctoral dissertation , McGill Uni -

versity , Montreal .

Meck , W . H . , and Church , R . M . ( 1983 ). A mode control model of counting and

timing processes . Journal of Experimental Psychology : Animal Behavior Processes ,

9 , 320 - 334 .

Mill , J . S . ( 1851 ). A system of logic , ratiocinative and inductive , being a connected

view of the principles of evidence , and the methods of scientific investigation ( Vol . 1,
3rd ed .) . London : John W . Parker .

Millikan , R . G . ( 1984 ) . Language , thought , and other biological categories : New

foundations for realism . Cambridge , MA : MIT Press .

Morris , M . ( 1991 ) . Why there are no mental representations . Minds and Machines ,
1 , 1 - 30 .

Palmer , S . E . ( 1978 ) . Fundamental aspects of cognitive representation . In E . Rosch

and B . B . Lloyd ( Eds . ) , Cognition and categol "ization ( pp . 259 - 303 ) . Hillsdale , NJ :
Erlbaum .

Porphyry ( 1887 ) . In Aristotelis Categorias [ On Aristotle ' s Categories ] . In A . Busse

( Ed .) , Commentaria in Aristotelem Graeca : V 01 . I ~ Pars L Porphyrii Isagoge

et In Aristotelis Categorias commentarium [ Greek commentaries on Aristotle : Vol .

4 , Part 1 . Porphyry 's commentary , Introduction and On Aristotle 's Categories ]

( pp . 53 - 142 ) . Berolini : Typis et Impensis Georgii Reimer . ( First published c . 265 )

Postal , P . M . ( 1976 ) . Linguistic anarchy notes . In J . D . McCawley ( Ed .) , Syntax

and semantics : Vol . 7 . Notes / rom the linguistic underground ( pp . 201 - 225 ) . New
Yark : Academic Press .

Putnam , H . ( 1973 ) . Meaning and reference . The Journal of Philosophy , 70 , 699 -

711 .

Putnam , H . ( 1981 ) . Reason , truth and history . Cambridge : Cambridge University

Press .

Putnam , H . ( 1984 ) . Is the causal structure of the physical itself something physical ?

In P . A . French , T . E . Uehling , Jr . , and H . K . Wettstein ( Eds .) , Midwest studies in

philosophy : Vol . 9 . Causation and causal theories ( pp . 3 - 16 ) . Minneapolis , MN :

University of Minnesota Press .
Scientific Theories That Unconceal Being 219

Putnam , H . ( 1988). Representation and reality . Cambridge , MA : MIT Press.

Putnam , H . ( 1989) . Model theory and the 'factuality ' of semantics . In A . George
( Ed .) , Reflections on Chomsky (pp . 213- 232) . Oxford : Blackwell .
Putnam , H . ( 1992a). Renewing philosophy . Cambridge , MA : Harvard University
Press .

Putnam , H . ( 1992b) . Reply to Noam Chomsky . Philosophical Topics, 20, 379-

385 .

Putnam , H . (1994). Logic and psychology : Comment on 'Logic and cognition .'
In J. Macnamara and G . E . Reyes ( Eds .), The logical foundations of cognition
( pp . 35- 42). New York : Oxford University Press.
Pylyshyn , Z . W . (1984) . Computation and cognition : Toward a foundation for
cognitive science. Cambridge , MA : MIT Press.
Rispoli , M . ( 1995) . Missing arguments and the acquisition of predicate meanings .
In M . Tomasello and W . E . Merriman ( Eds .), Beyond 'Namesfor things ': Young
children 's acquisition of verbs ( pp . 331- 352). Hi1lsdale , NJ : Erlbaum .
Schank , R ., and Kass , A . ( 1988) . Knowledge representation in people and ma -
chines. In U . Eco , M . Santambrogio , and P. Violi ( Eds.), Meaning and mental
representations ( pp . 181- 200) . Bloomington , IN : Indiana University Press.
Schmidt , R . H . ( 1986) . On the signification of mathematical symbols: Repairing the
number/numeral breach with the restored Renaissance counterpart concept. Annap -
olis , MD : Golden Hind Press .

Searle, J. R . ( 1980). Minds , brains , and programs . Behavioral and Brain Sciences,
3 , 417 - 457 .

Silvers, S. (1991). On naturalizing the semanticsof mental representation. The

British Journal for the Philosophy of Science, 42, 49- 73.
Slors, M . ( 1996) . Why Dennett cannot explain what it is to adopt the intentional
stance. The Philosophical Quarterly , 46, 93- 98.
Stampe, D . W . ( 1979) . Toward a causal theory of linguistic representation . In
P. A . French , T . E . Uehling , Jr ., and H . K . Wettstein ( Eds .), Contemporary per -
spectives in the philosophy of language ( pp . 81- 102) . Minneapolis , MN : University
of Minnesota Press .

Stevin , S. (1958a) . L 'arithmetique [Arithmetic ] . In D . J. Struik ( Ed .), Theprincipal

works of Simon Stevin: Vol. lIB . Mathematics ( pp . 477- 551) . Amsterdam : C . V .
Swets & Zeitlinger . ( First published 1585)
Stevin , S. ( 1958b) . De thiende [On the t-enth] . In D . J. Struik ( Ed .), The principal
\vorks of Simon Slevin : Vol. II Mathematics ( pp . 386- 425). Amsterdam : C . V .
Swets& Zeitlinger. (First published 1585)
Suarez, F . ( 1965) . Disputationes metaphysicae [Metaphysical disputations ] ( Vol . 1).
Hildesheim , Germany : Georg DIms Verlagsbuchhandlung . ( First published 1597)
Wilks , Y . ( 1988) . Reference and its role in computational models of mental rep-
resentations . In U . Eco , M . Santambrogio , and P. Violi ( Eds .), Meaning and
220 McPherson

mental representations
(pp. 221- 237). Bloomington
, IN : Indiana University
Press
.
Wynn, K . (1992). Evidenceagainstempiricistaccountsof the originsof numerical
knowledge . Mind & Language , 7, 315- 332.
Zabarella,J. (1966). De rebusnaturalibus
,. In Aristotelislibrosdeanima[On natu-
ral things; On Aristotle's bookson the soul]. Frankfurt: Minerva. (Works first
publishedposthumously in 1590and 1605 )
Chapter 10

The Nature of Human Steven Pinker and

Concepts : Evidence from an Alan Prince

Unusual Source

10 . 1 Classical and Family Resemblance Categories

This chapter is about an extensive parallel we have discovered between a

part of language and a part of cognition , and about the possibility that

the parallel is not a coincidence . The parallel involves the difference

between a classical category and a prototype or family resemblance cate -

gory , a topic of controversy for many years in cognitive psychology , phi -

losophy , linguistics , and artificial intelligence .

Classical categories are defined by necessary and sufficient criteria , and

membership in them is all - or - none . Examples include squares , grand -

mothers , odd numbers , and the vertebrate class Aves . Family resemblance

categories differ from classical categories in a number of ways :

. They lack necessary and sufficient conditions for membership . For

example , the category " chair " includes objects that have legs and that

lack them ( e .g . , beanbag chairs ) , and objects that can be sat upon and that

cannot ( e .g . , delicate museum pieces ).

. They have graded degrees of membership . A robin is a better example of

the family resemblance category " bird " than an eagle is ; and a penguin is

a worse example .

John Macnamara was a man of extraordinary wit , wisdom , and warmth . His
insights on the relations among cognition , logic , and the world greatly influenced
the thinking that went into this chapter . It is an honor to contribute to a volume in
his memory .
The order of authors is arbitrary . We thank Ned Block , Paul Bloom , Ray
J ackendoff , and Ed Smith for comments . Preparation of this chapter was sup -
ported
- by- National Institutes of Health grant HD 18381 .
A longer version of this chapter may be found in Communication and Cognition ,
29 , 307 - 361 .
 222 Pinker and Prince

. The category can be summarized by an ideal member or prototype ,

sometimes but not always an actual exemplar of the category . The more

similar other members are to the prototype , the better examples they are .

The sparrow , which is used to illustrate the entry for " bird " in many dic -

tionaries , might be a prototype of the bird category .

. There can be unclear cases - objects that mayor may not be members

of the category at all . One example is the fossil genus Archaeopteryx ,

characterized by one paleontologist as a " poor reptile , and not very much

of a bird " ( Konner 1982 ) . Garlic is an unclear example of the category

" vegetable , " as is ketchup , as we saw in the famous controversy that fol -

lowed the Reagan administration ' s proposal that ketchup be classified as

a vegetable in meeting nutritional guidelines for school lunch menus .

. They often display a family resemblance structure ( Wittgenstein 1953 ) .

The members of a family of people generally do not have a single feature

in common . Instead , a pool of features such as hair color , mouth shape ,

or nose size is shared by various sets of family members . Similarly , the

members of family resemblance categories have different features that run

through different subsets : green color is shared by spinach , celery , and

broccoli , but not carrots or cauliflower ; stems and bunches of florets are

shared by broccoli and cauliflower but not carrots .

. Good members tend to have characteristic nondefining features . For

example , gray hair and a domestic lifestyle characterize many grand -

mothers , but someone can be a grandmother without possessing either

property , such as Elizabeth Taylor .

10 . 1 . 1 Evidence for Family Resemblance Categories

Human concepts pick out categories of objects ; what kind of category do

they pick out ? There is a large body of evidence ( summarized in Smith

and Medin 1981 ; Rosch 1973 , 1978 , 1988 ) that has been taken to show

that human concepts correspond to family resemblance categories . First ,

semanticists and philosophers have generally failed in their attempts to

find necessary and sufficient conditions for most natural concepts that are

labeled by words ( see Fodor et al . 1980 ) . Second , psychologists have

found that subjects can give ratings of the goodness of membership of a

list of exemplars with respect to a category that are reliable and in close

agreement with one another . Similarly , there is good agreement about

prototypes and unclear cases . Third , these judgments are not unanalyz -

able gut feelings but can be predicted in a systematic way using a feature

calculus , in which the features possessed by a given exemplar ( assessed

Natureof HumanConcepts 223

independently, for example, by asking subjects to list the attributes of

the object) are compared with those possessedby the other members of
the category. Fourth , judgments of goodnessof membershiphave strong
effectson perfoffi1ancein many psychologicaltasks. For example, people
can verify that prototypical members belong to a category faster and
more accurately than they do with peripheral members, and when asked
to recall instancesof a category, they name prototypical membersfirst.
Fifth , developmental psychologistshave found that children often learn
the names for prototypical exemplars of a category before learning the
namesfor other exemplars, and that they apply superordinateterms such
as bird to its prototypical membersfirst. Sixth, linguists have found that
certain adverbials called hedgesare sensitiveto prototypicality : one can
say that a sparrow, but not a penguin, is a bird par excellence, and that a
penguin, but not a sparrow, is technically or strictly speakinga bird .

10.1.2 EvidenceagainstFamily ResemblanceCategories

On the other hand, there is also evidencethat certain aspectsof human
concepts do not correspond to family resemblancecategories. Some of
the empirical effectsthat have been interpreted as demonstrating family
resemblanceclassesalso occur for categoriesthat people clearly treat as
being classical. Armstrong, Gleitman, and Gleitman (1983) have found
that subjectsshow a great deal of agreementwith one another in rating
the degreeof membership of exemplars of categorieslike ' 'female" and
" odd number." For example, they agreethat a mother is a better example
of a female than a comedienneis, and that 13 is a better example of an
odd number than 23. Similarly, Armstrong, Gleitman, and Gleitman
found that peopletake lesstime and are more accurateat deciding that 13
is an odd number than that 23 is, and that a mother is a female than that
a comedienneis. Sincethesesubjectssurely knew that " female" and " odd
number" in reality have sharp boundaries and all-or-none membership
(and Armstrong, Gleitman, and Gleitman discovered, in an independent
questionnaire, that their subjectsbelieved as much), it calls into question
whether the analogousresults that Rosch and others obtained for " bird "
or " tool" really tell us anything about people' representationsof those
concepts.
Moreover, most judgments of membershipin family resemblancecate-
gories based on characteristic features are highly corrigible when people
are asked to engagein careful reasoning about it . For some purposes,
peopleare willing to considera penguin asa full -fledgedbird and Elizabeth
 224 PinkerandPrince

Taylor a full - fledged grandmother . In fact , characteristic nondefining fea -

tures can be quickly abandoned , even by young children . Children say that

three - legged dogs are dogs , and that raccoons with stripes painted down

their backs are raccoons , not skunks ( Rey 1983 ; Armstrong , Gleitman ,

and Gleitman 1983 ; Kei11989 ; Gelman , Coley , and Gottfried 1994 ) .

Similar demonstrations with adults have shown that inference is often

not driven by the similarity criteria that define family resemblance cate -

gories ( see Murphy 1993 ; Medin 1989 ; Kelly 1992 ; Smith , Langston , and

Nisbett 1992 ; Rips 1989 ; Rey 1983 ) . For example , when people are asked

which two out of three belong together - white hair , gray hair , black

hair - they say that black is the odd hair out , because aging hair turns

gray then white . But when asked about a white cloud , a gray cloud , and a

black cloud , they say that white is the odd cloud out , because gray and

black clouds give rain . In another experiment , subjects were asked

whether a three - inch disk is more similar to a quarter or a pizza , and

whether it is more likely to be a quarter or a pizza . Most said it is more

similar to a quarter but more likely to be a pizza , presumably because

quarters have to be standardized but pizzas can vary . Most people , upon

being presented with a centipede , a caterpillar that looks like it , and a

butterfly that the caterpillar turns into , feel that the caterpillar and the

butterfly are " the same animal , " but that the caterpillar and the centipede

are not , despite appearances to the contrary .

10.1.3 PossibleResolutions
This conflicting evidencecan be resolvedin severalways.
First, human concepts could basically pick out family resemblance
categories. Classical categorieswould be special casesor artifacts result-
ing from explicit instruction, such as in formal schooling. Alternatively ,
human concepts could basically pick out classical categories. Family
resemblancecategorieswould be artifacts of experimental tasks asking
subjects for graded judgments or asking them to make categorization
decisionsunder time pressure. A third , compromise position would say
that human conceptscorrespondto both classicaland family resemblance
categories. Classical categories are the " core" of the concept, used for
reasoning. Family resemblancecategoriesare " identification procedures"
or " stereotypes," usedfor identification of categoryexemplarson the basis
of available perceptual information , or for rapid approximate reasoning.
Although most theorists have tended toward compromise positions,
something close to the mainly-family-resemblanceview can be found in
Nature of Human Concepts 225

Lakoff 1987, Rosch 1978, and Smith , Medin , and Rips 1984; something
close to the mainly -classical-category view can be found in Rey 1983,
Fodor 1981, and Armstrong , Gleitman , and Gleitman 1983; and tentative
proposals favoring the core-plus-identification -procedure compromise can
be found in Smith and Medin 1981, Armstrong , Gleitman , and Gleitman ,
and Osherson and Smith 1981 .

This leads to several open questions . ( 1) Is one type of category psy-

chologically real , the other an artifact or special case? (2) If both are
psychologically real , can they be distinguished by function (e.g., reasoning
vs. categorization )? (3) If both are psychologically real , are they handled
by the same kind of computational architecture ? (4) If either or both are
psychologically real , do they correspond to ontological categories? That
is, are classical (or family resemblance) categories incorrectly imposed
by people on the world because of limitations of the way the mind works ,
or is there some sense in which the world contains classical (or family
resemblance) categories, which people can accurately represent as such,
presumably because the mind evolved to grasp aspects of the world
accurately ?
We will attempt to shed light on these questions by examining an un -
usual source of evidence: English past tense forms .

10.2 An Unexpected Test Case: English Past Tense Forms

English verbs come in two types: those that have regular past tense forms ,
and those that have irregular past tense forms . Consider them as two
categories : " regular verbs," such as walk /walked, talk /talked, jog /jogged ,
pat /patted , kiss/kissed, and play /played , and. " irregular verbs," such as hit /
hit , go/went, sleep/slept, make/made, ring /rang, bring /brought , stink /stank ,
andfiy /fiew .
In fact , the irregular verbs are not a single class but a set of subclasses,
which can be subdivided according to the kind of change that the stem
undergoes to form the past tense (see Pinker and Prince 1988 for a full
list ). Here are some examples:
. Lax the vowel : bleed, breed, feed , lead, mislead, read, speed, plead, meet,
hide, slide, bite, light , shoot
. Lax the vowel , add a -t: lose, deal, feel , knee/, mean, dream, creep, keep,
leap, sleep, sweep, weep, leave
. Change the rhyme to -ought: buy, bring , catch, fight , seek, teach, think
226 PinkerandPrince

. Change / 1/ to / ref or / AI: ring, sing, spring, drink, shrink, sink, stink,
sM-'im, begin, cling, fling , sling, sting, string, swing, wring, stick, dig, win,
spin, stink, slink, run, hang, strike, sneak
. Changethe vowel to /u/ : blow, grow, know, throw, draw, withdraw, fly ,
slay
Let us consider someproperties of the irregular subclasses
.

10.2.1 Propertiesof the Irregular Subclasses

10.2.1.1 Characteristic Nondefining Features The irregular subclasses

tend to be characterizedby phonological properties other than those that
define the change from stem to past form . Consider the subclassthat
changeson 0 or similar vowel to u: blow, grow, know, throw, draw, with-
draw, fly , slay. In principle, any verb with an 0 or similar vowel could be
included in the subclass. In fact, all the verb roots in the subclassend in a
vowel, usually a diphthong, and most begin with a consonantcluster.
Similarly, the subclassthat changes layl to law I - bind, find, grind,
wind- could include any verb with the vowel lay I , but in fact, all the
verbs happen to end in -nd. The subclassthat changesa final d to t- bend,
send, lend, rend, build- could include any word ending in d, but in fact,
most of the verbs rhyme with -end. Finally , the subclassthat changesthe
vowel leyl to Jut- take, mistake, forsake, shake- could include any word
with an leyI , but in fact, all the verbs roots rhyme with -ake and begin
wi th a coronal consonant.
Note that the characteristic nondefining features are arbitrary, not
lawful, with respectto the soundpattern of English. No rule of phonology
excludesloon as the past tenseof loan or choudas the past of chide.

10.2.1.2 Family Resemblance Irregular subclassesdisplay a family

resemblancestructure (Bybeeand Slobin 1982a; Bybeeand Moder 1983).
Consider the subclassthat changesan / 1/ to an / A/ . Most of the verbs end
with a velar nasal consonant: shrink, sink, stink, cling, fling , sling, sting,
string, swing, slink. Some end in a consonant that is velar but not nasal:
stick, dig, sneak, strike. Others end in a vowel that is nasal but not velar:
win, spin, swim, begin.
Similarly, within the subclassthat changesa final diphthong to /u/ ,
some begin with a <consonant-sonorant> cluster and contain the diph-
thong low / : blow, grow, throw. But one member, know, contains the fowl
Nature
of Human
Concepts 227

diphthongbutdoesnotbeginwitha consonant cluster. Othersbeginwith

a consonant
clusterbuthavea differentdiphthongor nodiphthong at all:
draw,withdraw
, fly, slay.

10.2.1.3 Prototypicality Bybeeand Moder(1983 ) point out that for

manyof thesubclasses , onecancharacterize
a prototype, basedon the
kinds of characteristic
phonologicalpropertiesthat definethe family
resemblancestructure
. Accordingto BybeeandModer, theprototypeof
theing-+ ungsubclass is

s C C i [Velar nasal ]
whereC standsfor a consonant . Thisprototypeis maximallysimilarto
mostmembers of theexistingsubclass , but moreinterestingly , it predicts
subjects' generalization of the/1/ -+ /AIchange to novelverbs . Bybeeand
Moderaskedsubjects to ratehownaturala varietyof putativepasttense
formssounded for eachof a setof noncestems . Theindependent variable
wasthesimilarityof thestemto theprototypelistedabove . Theyfound
that subjects wereextremely likelyto acceptthevowelchange for stems
like spling, strink, andskring , whichmatchtheschema for theprototype
exactly. Theywereonly slightlylesswillingto acceptstruckandskrum
asthepastof strickandskrim, whichdifferfrom theprototypein one
feature. Somewhat lowerin acceptability werespruvfor spriv,andsimilar
pastformsfor sking , smig, piing, andkrink. Glick, krin, plim, shinkwere
evenlesslikelyto admitof thevowelchange , andtrib, vin, andsid, the
formsfurthestfrom theprototype , weretheleastacceptable of all. The
resultshavebeenreplicatedby Prasadaand Pinker(1993 ), and with
analogous Germanformsby Marcuset al. (1995 ).

10.2.1.4 Graded Goodnessof Membership Within most of the sub-

classes, there are some verbs that clearly accept the irregular past tense
form, but there are others, usually of low but nonzero frequency, for
which the specifiedpast tenseform is lessthan fully acceptableand which
are felt to be unusual or stilted. In (1) we contrast some" good examples"
of the past tenseform with " poor examples" of the samekinds of forms;
the intuitions vary from person to person, as is true for nonprototypical
exemplarsof conceptual categories. The judgments for theseand similar
forms are documentedquantitatively by Ullman (1999).
228 Pinker and Prince

( 1) Good examples Poor examples

hit , split spit, forbid
bled , fed pled, sped
burnt , bent learnt, lent, rent
dealt , felt , meant knelt, dreamt
froze , spoke wove, hove
got , forgot begot, trod
wrote , drove , rode dove, strove, smote~strode

10.2.1.5 Unclear Cases For someverbs associatedwith a subclass, the

mandated past tense form is so poor in people's judgment that it is
unclear whether the verb can be said to belong to the subclassat all.
Sometimestheseverbs are restrictedto idioms, cliches, or other specialized
usages. For example, the expressionforgo the pleasureof, as in You will
excuseme if I forgo the pleasureof readingyour paper until it 's published,
soundsfairly natural. Becausethe verb has a transparent morphological
decompositionas [for + go], the formforgoed is clearly unacceptable, but
the irregular past tense form, as in Last night I forwent the pleasure of
grading studentpapers, is decidedlypeculiar if not outright ungrammatical
(this intuition has been corroborated by ratings from subjectsin Ullman
1999). Similarly, That dressreally becomesyou is a natural English sen-
tence, but Whenyou were thinner, that dressreally becameyou is almost
unintelligible.
In other cases,grammatical phenomenaconspireto make the past tense
form of a verb extremely rare. The transitive verb stand meaning 'to tol-
erate' is fairly common, but becauseit is usually used as the complement
of a negatedauxiliary, as in Shecan't stand him, the verb is almost always
heard in its stemform. In constructionswherethe past is allowed to reveal
itself, the verb soundsquite odd: compareI don't know how shestandshim
with I don't know how she stood him, and I don't know how she bears it
with I don't know how shebore it .

10.2.1.6 Conclusions about the Irregular Subclasses Subclasses of

irregular verbs in English have characteristic nondefining features, family
resemblancestructures, prototypes, gradations of goodnessof member-
ship, and unclear or fuzzy cases. Sincetheseare exactly the properties that
define family resemblancecategories, we conclude, in agreement with
Bybee and Moder (1983), that the irregular subclassesare family resem-
blance categories.
Nature
ofHuman
Concepts 229

This is a surprising conclusion. Linguistic rules are traditionally

thought of as a paradigm caseof categorical, all-or-none operations and
might be thought to correspond to classical categoriesif anything did.
The fact that entities subject to grammatical operations can have a clear
family resemblancestructure thus has far-ranging implications for some
theorists. For example, for Rumelhart and McClelland (1986), this phe-
nomenon is part of an argument for a radically new approach to studying
language, based on a computational architecture in which rules play no
causal role. For Lakoff (1987), it is part of a call for a radically new way
of understandinghuman cognition in general.

10.2.2 Propertiesof the Regular Class

Before we acceptthe claims of Rumelhart and McClelland and of Lakoff ,
we must ask, Do all linguistic objects fall into family resemblancecate-
gories? Indeed, does the class most associated with English irregular
verbs- namely, English regular verbs- fall into family resemblancecate-
gories? One answer, favored by Bybee (Bybee and Moder 1983; Bybee
1991) and by Rumelhart and McClelland (1986), is yes: the regular class
just has more members, and more general characteristic features. Let us
examine this possibili ty.
The regular and irregular classesinteract in a specific way, and it is
necessaryto take account of this interaction so that the properties of the
irregular subclassesdo not confound our examination of the properties of
the regular class. The interaction is governedby what has beencalled the
Blocking Principle (Aronoff 1976) or the Unique Entry Principle (Pinker
1984): if a verb has an irregular past tenseform, its regular form is pre-
empted or blocked. Thus, the fact that go has an irregular past went not
only allows us to usewent; it preventsus from using *goed. The verb glow,
in contrast, does not have an irregular past *glew, so its regular past
glowedis not blocked.
We have seenhow someirregular past forms are " fuzzy" or marginal in
their grammaticality. As a result of blocking, these gradations of good-
nesscan causethe appearanceof complementary gradations of goodness
of the correspondingregular. Thus, becausepled is a marginal past tense
form for plead but one that we nonethelessrecognize, the regular form
pleadedguilty soundsfairly good but may be tinged with a bit of uncer-
tainty for some speakers. Conversely, wept is a fairly good past tense
form of weep, though not maximally natural (cf., e.g., kept for keep). As a
result weepeddoes not sound terribly good, though it is not perceivedas
230 Pinker and Prince

being completely ungrammatical either (cf. *keeped). This effect has been
documentedby Ulhnan (1999; seealso Pinker 1991and Pinker and Prince
1994), who asked subjectsto rate the naturalnessof irregular and regu-
larized past tense fonns for verbs whose irregular pasts are somewhat
fuzzy in goodness. The two setsof ratings were negatively correlated.
We now put aside this reciprocity effect due to blocking and try to
detennine whether the regular class has family-resemblance -category
properties independent of those of the irregular subclasseswith which it
competes.

10.2.2.1 Independenceof the Phonology of the Stem The first salient

property of the regular classis that it has no sensitivity to the phonologi-
cal properties of its stems. As a result, it has no phonologically charac-
terized prototype, gradations of membership, or characteristicfeatures.
First, the phonological conditions that govern the irregular subclasses
can be entirely flouted by regular verbs. In the extremecase, homophones
can have different past tense forms: ring/rang versus wring/wrung, hang/
hung (suspend) versushang/hanged(execute), lie/lay (recline) versuslie/lied
(fib), fit /fit (what a shirt does) versusfit /fitted (what a tailor does). More
generally, there are regular counterexamplesto the membership criteria
for each of the irregular subclasses .
(2) shut/ shut jut /jutted
bleed/bled need/needed
bend/bent mend/mended
sleep/ slept seep/ seeped
sell/ sold yell/yelled
freeze/froze seize/seized
grow/grew glow/glowed
take/took fake/faked
stink/ stunk blink /blinked
ring/ rang ring/ ringed
This shows that the phonologically defined fuzzy boundaries of the
irregular subclassesdo not create complementary phonological fuzzy
boundaries of the regular classes . The effect of the Blocking Principle is
that specific irregular words block their corresponding regulars. Though
most of those words come from regions of phonological space whose
neighbors are also often irregular, those regions do not define comple-
mentary fuzzy " holes" in the spacefrom which the regulars are excluded;
a regular form can occupy any point in that spacewhatsoever. Moreover,
Nature
of Human
Concepts 231

it is not just that there already exist regular verbs in the language that
live in irregular phonological neighborhoods; the regular class can add
membersthat violate any irregular membershipcriteria. The reason has
been spelled out by Kiparsky (1982a,b), Pinker and Prince (1988, 1994),
Kim et al. (1991), and Kim et al. (1994). Irregular forms are verb roots,
not verbs. Not all verbs have verb roots: a verb that is intuitively derived
from a noun (e.g., to nail ) has a noun root. A noun or an adjectivecannot
be marked in the lexicon as having an " irregular past," becausenouns
and adjectivesdo not have past tenseforms at all; the notion makes no
sense.Therefore, a verb createdout of a noun or adjectivecannot have an
irregular past either. All suchverbs are regular, regardlessof their phono-
logical properties.
(3) He braked the car suddenly. i = broke
He flied out to center field. i = flew
He ringed the city with artillery . *rang
Martina 2-settedChris. *2-set
He sleigheddown the hill . *slew
He de-ftea'd his dog. *de-fled
He spitted the pig. *spat
He righted the boat. *rote
He high-sticked the goalie. *high-stuck
He grandstandedto the crowd. * grandstood
This makes it possible, in principle, for any sound sequencewhatsoever
to becomea regular verb. There is a lexical rule in English that convertsa
name into a verb prefixed with out, as in Clinton hasjinally out-Nixoned
Nixon. Like all verbs derived from nonverbs, it is regular. Since any lin-
guistically possiblesound can be someone's name, any linguistically pos-
sible sound can be a regular verb, allowing there to be regular homo-
phonesfor any irregular. For example:
(4) a. Mary out-Sally-Rided Sally Ride.
*Mary out-Sally-Rode Sally Ride.
b. In grim notoriety, Alcatraz out-Sing-SingedSing-Sing.
*In grim notoriety, Alcatraz out-Sing-Sang Sing-Sing.
*In grim notoriety, Alcatraz out-Sing-Sung Sing-Sing.
This effect has been demonstrated experimentally in several popula-
tions. Kim et al. (1991) asked subjectsto rate the regular and irregular
past tense foffi1s of a set of verbs that were either derived from nouns
that were homophonous with an irregular verb or were derived directly
 232 Pinker and Prince

from the irregular verbs . For verbs with noun roots , the regular form was

give higher ratings ; for verbs with verb roots , the irregular form was

given higher ratings . Similar effects have been demonstrated in non -

college - educated subjects ( Kim et ale 1991 ), children ( Kim et ale 1994 ),

and German - speaking adults ( Marcus et ale 1995 ) .

Perfectly natural - sounding regular past tense forms exist not only when

the verb root is similar to an irregular , but also when it is dissimilar to

existing regular roots and hence lacks a prototype that would serve as

the source of an analogical generalization . Prasada and Pinker ( 1993 )

replicated Bybee and Moder ' s ( 1983 ) study but also presented novel reg -

ular words of differing similarity to existing English regular words . For

example , plip is close to one of the prototypes for regular verbs in English ,

because it rhymes with slip , flip , trip , nip , sip , clip , dip , grip , strip , tip , whip ,

and zip , whereas smaig rhymes with no existing verb root , andploamph is

not even phonologically well formed in English . Nonetheless , people

rated the prototypical and peripheral forms as sounding equally natural

( relative to their stems ) and produced the prototypical and peripheral

forms with the same probability when they had to produce them .

10 . 2 . 2 . 2 No Prototypes , Gradation of Membership , or Unclear Cases

Caused by Low Frequency or Restricted Contexts Unlike irregular past

tense forms , regular past tense forms do not suffer in well - formedness on

account of frequency , familiarity , idiomaticity , frozenness , or restricted

syntactic contexts . In earlier work ( Pinker and Prince 1988 ) , we noted that

though the verb perambulate may be of low frequency , it is no worse -

sounding in its past tense form than it is in its stem form ; there is no

feeling that perambulated is a worse past tense form of perambulate than

walked is of walk . In fact , a verb can be of essentially zero frequency and

still have a regular past tense form that is judged as no worse than the

verb itself . Though fleech , fleer , and anastomose are unknown to most

speakers , speakers judgefleeched , fleered , and anastomosed to be perfectly

good as the past tense forms of those verbs . These observations have been

confirmed experimentally by Ullman ( 1999 ) : subjects ' ratings of regular

pasts correlated highly with their ratings of the corresponding stems , but

not with the frequency of the past form ( partialing out stem rating ) . In

contrast , ratings of irregular pasts correlated less strongly with their stem

ratings but significantly with past frequency , partialing out stem rating .

Unlike what happens with irregular verbs , when a regular verb gets

trapped in a frozen or restricted expression , putting it into the past tense

Natureof HumanConcepts 233

makes it no worse. For example, the verb eke is seldom used outside
contexts such as She ekes out a living, but She eked out a living, unlike
forwent thepleasureof, doesnot suffer becauseof it . Similarly: He crooked
hisfinger; Shestinted no effort,' I broachedthe subjectswith him,. The news
auguredwellfor his chances . The regular verb to afford, like the irregular
verb to stand, usually occursas a complementto can't, but when liberated
from this context its past tenseforn1 is perfectly natural: I don't know how
she afforded it . Similarly, both She doesn't suffer fools gladly and She
neversufferedfools gladly are acceptable.
These phenomenashow why the apparent gradednessof acceptability
for regular forn1slike pleaded or weepedcan be localized to the graded-
nessof the correspondingirregulars becauseof the effectsof the Blocking
Principle and are not inherent to the regular verbs per se. The gradedness
of certain irregulars generally comesfrom low frequency combined with
similarity to the prototypes of their subclasses(Ullman 1993). But for
regular verbs that do not compete with specific irregular roots, there is
no complementary landscapeof acceptability defined by phonology and
frequency; all are equally good.

10.2.2.3 Default Structure As we have seen, the regular past tense

alternation can apply regardlessof the stem's phonological properties,
verb-root versusnon-verb-root status, frequency, listedness(familiarity ),
and range of contexts. Apparently, the regular class is the default class.
More generally, there is a sensein which the category of regular verbs has
no properties; it is an epiphenomenonof the scopeof application of the
regular rule.

10.2.2.4 Conclusionsabout the Regular Class Thesephenomenainvite

the following conclusion: the classof regular verbs in English is a classical
category. Its necessaryand sufficient conditions are simply the conditions
of application of the regular rule within English grammar. Those con-
ditions for membership can be stated simply: a verb, unless it has an
irregular root.

10.3 Psychological
Implications

We have shown that by standard criteria the irregular subclassesare pro-

totype or family resemblancecategories, and the regular classis a classical
category. This conclusion has severalimplications.
234 PinkerandPrince

10.3.1 PsychologicalReality
First, both family resemblancecategoriesand classicalcategoriescan be
psychologically real and natural. Classical categories need not be the
product of explicit instruction or formal schooling: the regular past tense
alternation doesnot have to be taught, and indeed every English-speaking
child learns it and begins to use it productively in the third year of life
(Marcus et al. 1992). The fact that children apply the regular alternation
even to high-frequency irregular stems such as come and go, which they
also use with their correct irregular pasts much of the time, suggeststhat
children in some way appreciate the inherently universal range of the
regular rule. And like adults, they apply the regular suffix to regular verbs
regardlessof the degree of the verbs' similarity to other regular verbs
(Marcus et ale 1992), and to irregular-sounding verbs that are derived
from nouns and adjectives (Kim et ale 1994). Gordon (1985) and
Stromswold (1990) have shown that children as young as 3 make quali-
tative distinctions between regular and irregular plural nouns related to
their different formal roles within the grammar, without the benefit of
implicit or explicit teaching inputs (seeMarcus et ale 1992and Kim et ale
1994for discussion).
The regularization-through-derivation effect (flied out, high-sticked)
provides particularly compelling evidencethat classicalcategoriesdo not
have to be the product of rules that are explicitly formulated and delib-
erately transmitted. The use of the regular rule as a default operation,
applying to any derived verb regardlessof its phonology, is a grass-roots
phenomenon whose subtleties are better appreciated at an unconscious
level by the person in the street than by those charged with formulating
prescriptive rules. Kim et al. (1991) found that non-college-educatedsub-
jects showed the effect strongly, and in the recent history of English and
other languagesthere are documented casesin which the language has
accommodatedsuchregularizationsin the face of explicit opposition from
editors and prescriptive grammarians. For example, Mencken (1936)
notes that the verb to joyride, first attaining popularity in the 1920s, was
usually given the past tenseform joyrided, as we would predict given its
derivation from the noun a joyride . Prescriptive grammarians unsuccess -
fully tried to encouragejoyrode in its place. Similarly, Kim et al. (1994)
showed that children display the effect despite the fact that most have
rarely or never heard regularized past tenseforms for irregular-sounding
verbs in the speechof adults.
Natureof HumanConcepts 235

On the other side, family resemblancecategories are not necessarily

artifacts of reaction time studies or rating studies, as Fodor (1981) and
Armstrong, Gleitman, and Gleitman (1983) have suggested . Children
generalizefamily resemblancepatterns of irregular subclassesto inappro-
priate regular and irregular verbs in their spontaneous speech, as in
brang for brought and bote for bit, and their generalizationsappear to be
sensitive to the frequency and family resemblancestructure of the sub-
classes(Xu and Pinker 1995; Bybee and Slobin 1982a; Rumelhart and
McClelland 1986; Pinker and Prince 1988). The irregular subclassstruc-
ture also affects dialectal variation and historical change in adult speech
(Bybeeand Slobin 1982b; Mencken 1936; Prasadaand Pinker 1993), with
new irregular forms occasionally entering the languageif their stemsare
sufficiently similar to existing irregular stems.

10.3.2 PsychologicalFunction
A further corollary is that classical categories and family resemblance
categories do not have to have different psychological functions such
as careful versus causal reasoning, or reasoningversus categorization of
exemplars. What is perhapsmost striking about the contrast betweenthe
regular and irregular verbs is that two kinds of entities live side by side in
people's heads, servingthe samefunction within the grammar as a whole:
regular and irregular verbs play indistinguishable roles in the syntax and
semanticsof tense in English. There is no construction, for example, in
which a regular but not an irregular verb can be inserted or vice versa,
and no systematic difference in the temporal relationships semantically
encodedin the past tenseforms of regular and irregular verbs.
More specifically, it is difficult to make senseof the notion that family
resemblancecategoriesare the product of a set of identification proce-
dures used to classify exemplars as belonging to core categorieswith a
more classical structure. The suggestionthat " irregulars are used in per-
ceptually categorizingmembersof the regular class" makesno sense.The
irregulars are a classof words that display one kind of category structure;
the regulars do not display it .
Perhapsa closer analogy would be betweenmembershipconditions for
the irregular subclassesand the operation on the stem that generatesthe
past tenseform . One might say that a family resemblancestructure char-
acterizesthe membershipof each subclass, but once an item is a member
(for whatever reason), it is transformed into a past tenseform by a clas-
236 PinkerandPrince

sical all-or-none operation such as laxing the vowel. But even here, the
core/identification distinction does not easily apply, becausethe changes
that the member stemsof a class undergo, and not just the properties of
the stems, have a heterogeneousstructure. Within the subclassof irregu-
lars ending in ing/ink, sing goesto sangwhile sting goesto stung and bring
goesto brought. Similarly, within the subclassthat adds a / d/ to the past
tense form, some verbs have their vowellaxed (e.g., hear/heard), some
have their final consonant deleted (e.g., make/made, have/had), some
undergo the e - 0 ablaut that is frequent across the various subclasses
(e.g., sell/sold), and one undergoesa unique vowel change (do/did). In
sum, both the membershipconditions and the operations of the irregular
subclassesdisplay family resemblancecategoryeffects. Later we will show
that the core/identification distinction does not work well for conceptual
categorieseither.

10.3.3 Underlying PsychologicalMechanism

Though classical and family resemblancecategories, in the case of the
past tense, do not differ in psychological function- what they are used
for- they do differ in psychological structure- what mental processes
give rise to them. Our main claim is that the psychological differencebe-
tween regulars and irregulars is a fundamental one, and is of a piecewith
the psychological difference between classical and family resemblance
categoriesin general, including conceptualcategories.
As we have seen, the classical category consisting of regular verbs is
defined completely and implicitly by the nature of a rule in the context of
a formal system, in this case, a rule within English grammar that applies
to any word bearing the part-of-speech symbol " verb" (unless it has
an irregular root). The categoryis not a generalizationor summary over a
set of exemplars; indeed, it is blind to the properties of the exemplarsthat
fall into the category. It falls out of the combinatorial rule system that
allows humansto communicatepropositions (including novel, unusual, or
abstract propositions) by building complex words, phrases, and sentences
in which the meaning of the whole can be determinedfrom the meanings
of the parts and the way in which they are combined.
Family resemblancecategories, in contrast, are generalization of pat-
terns of property correlations within a set of memorized exemplars. Con-
sequently, factors that affect human memory affect the composition of the
irregular class. A well-known example is word frequency. Irregular verbs
tend to be higher in frequency than regular verbs (Ullman 1993; Marcus
Natureof HumanConcepts 237

et al. 1995), and if an irregular verb's frequencydeclinesdiachronically, it

is liable to becomeregular (Hooper 1976; Bybeeand Slobin 1982b; Bybee
1985). Presumablythis is becauseirregulars are memorized. To memorize
an item, one has to hear it ; if opportunities for hearing an item are few, its
irregular form cannot be acquired and the regular rule can apply as the
default. This is also presumably the cause of the fuzziness of the past
tensesof irregular verbs that are used mainly in nonpast forms, such as
forgo or the idiomatic meaningsof stand or become.
A related account could help explain the genesisof irregular verbs'
family resemblance structures. Rosch and Mervis (1975) found that
peoplefind lists of letter strings that display family resemblancestructures
easierto rememberthan oneswith arbitrary patterns of similarity . Just as
frequency affectsthe memorizability, hencecomposition, of the irregular
subclasses , so might family resemblancestructure. The current subclasses
may have emergedfrom a Darwinian processin which the irregular verbs
that survived the generation-to-generationmemorization cycle were those
that could be grouped into easy-to-rememberfamily resemblanceclusters.
In sum, the properties of the regular and irregular classesof verbs in
English show that both classicalcategoriesand family resemblancecate-
gories can be psychologically real, easily and naturally acquired, and not
subject to a division of labor by function along the lines of reasoning
versusidentification of exemplars. Rather, they differ becausethey are the
products of two different kinds of mental processes : a formal rule system
and a memorized, partially structured list of exemplars. We now point
out two less obvious implications: classicaland prototype categoriesare
suited to different kinds of computational architectures, and the mental
mechanismsgiving rise to the two kinds of categoriesare suited to repre-
senting different kinds of entities in the world .

10.3.4 ComputationalArchitecture
The acquisition of English past tensemorphology has been implemented
in a widely discussedcomputer simulation model by Rumelhart and
McClelland (1986; see also Pinker and Prince 1988, 1994; Lachter and
Bever 1988; Sproat 1992; Prasada and Pinker 1993; Marcus et al. 1992,
1995). The Rumelhart- McClelland (RM ) model makes use of a pattern
associator architecture, which is paradigmatic of the parallel distrib-
uted processing (PDP) or connectionist approach to cognitive science
(Rumelhart and McClelland 1986; McClelland and Rumelhart 1986;
Pinker and Mehler 1988). Two properties of pattern associators are
238 Pinker and Prince

crucial in understanding their behavior: items are representedby their

properties, and statistical contingenciesbetweenevery input property and
every output property acrossa setof items are recordedand superimposed.
Before being applied to the caseof learning past tenseforms, pattern
associatorshad been studied in detail, including their ability to learn and
identify members of conceptual categories(McClelland and Rumelhart
1985), and they are known to do certain things well. They can often
reproduce a set of associationsin a training set and generalizeto new
caseson the basis of their similarity to existing ones. They are sensitive
to input pattern frequenciesin ways similar to humans. Furthermore,
they reproduce many of the effects displayed by people when dealing
with family resemblancecategories. McClelland and Rumelhart (1985)
and Whittlesea (1989) have devisedpattern associatorsthat are fed pat-
terns of data concerningproperties of a set of nonlinguistic objects. They
found that the models do fairly well at duplicating the effects of fre-
quency, prototypicality , family resemblance , gradations of membership,
and influence of particular exemplarson human classification times and
error rates. Since such effects are known to be related to cooccurrence
frequencies
.. among objects' features (Smith and Medin 1981), this is not
surpnsmg.
Thanks to these abilities, the pattern associator that Rumelhart and
McClelland applied to learning past tense forms handled the irregular
verbs with some success . After training on a set of regular and irregular
verbs, the model was able to approximate the past tenseforms for all of
them given only the stem as input . Furthermore, it was able to generalize
to new irregular verbs by analogy to similar onesin the training set, such
as bid for bid, clung for cling, and wept for weep. In addition, it showed
a tendencyto extend some of the subregularalternations to regular verbs
on the basis of their similarity to irregulars, such as kid for kid and slept
for slip, showing a sensitivity to the family resemblancestructure of the
irregular subclasses . Finally , its tendenciesto overgeneralizethe regular
/ dj ending to the various irregular subclassesis in rough accord with
children's tendenciesto do so, which in turn is based on the frequency
and consistencyof the vowel changesthat the verbs within each subclass
undergo (Pinker and Prince 1988; Sproat 1992).1
However, pattern associatorsdo not seemto perform as well for other
kinds of mappings. In particular, they are deficient in handling regular
verbs. For one thing, their uniform structure, in which regulars and
irregulars are handled by a single associative mechanism, provides no
Nature
of Human
Concepts 239

explanation for why the regular class has such different properties from
the irregular classes ; it falsely predicts that the regular classshouldjust be
a larger and more generalprototype subclass.
Moreover, the pattern associatorfails to acquire the regulars properly.
In earlier work (Pinker and Prince 1988), we pointed out that the model is
prone to blending. Competing statistical regularities in which a stem par-
ticipates do not block each other; they get superimposed. For example,
the model produced erroneousforms in which an irregular vowel change
was combined with the regular ending, as in seppedas the past of sip or
brawnedfor brown. It would often blend the It I and lidl variants of the
regular past tense fonn , producing stepted for step or typted for type.
Sometimesthe blends were quite odd, such as membled for mailed or
tourederfor tour.
Furthermore, we noted that in contrast to the default nature of the
regular rule, the RM model failed to produce any past form at all for
certain verbs, such asjump, pump, glare, and trail . Presumably this was
becausethe model could not treat the regular ending as an operation that
was capable of applying to any stem whatsoever, regardlessof its prop-
erties; the ending was simply associatedwith the features of the regular
stemsencounteredin the input . If a new verb happenedto lie in a region
of phonological spacein which no verbs had previously been supplied in
the training set (e.g., jump and pump, with their unusual word-final con-
sonant cluster), no coherent set of output features was strongly enough
associatedwith the active input features, and no response above the
background noise could be made. Our diagnosis was tested by Prasada
and Pinker (1993), who presentedtypical-sounding and unusual-sounding
verbs to the trained network. For the unusual-soundingitems, it produced
odd blends and chimeras such as smairf/sprul'ice, trilbltreelilt , smeejf
leefloag, andfrilgffreezled.
Finally , the model is inconsistent with certain kinds of developmental
evidence. Children first use many irregulars properly when they use them
in a past tenseform at all (e.g., broke), then begin to overregularizethem
occasionally(e.g., broke and breaked) before the overregularizationsdrop
out years later. Sincepattern associatorsare driven by pattern frequency,
the only way the RM model could be made to duplicate this sequence
was first to exposeit to a small number of high-frequency verbs, most of
them irregular, presenteda few times each, followed by a large number of
medium-frequency verbs, most of them regular, presentedmany times
each. Only when the model was swampedwith exemplarsof the regular
240 PinkerandPrince

pattern did it begin to overregularize verbs it had previously handled

properly. However, the onset of overregularization in children is not
causedby a suddenshift in the proportion of regular verbs in the speech
they hear from their parents; the proportion remains largely unchanged
before, during, and after the point at which they begin to overregularize
(Pinker and Prince 1988; Slobin 1971; Marcus et al. 1992). Nor is it
causedby a rapid increasein the proportion of verbs in their vocabulary
that is regular; the percentageof children's vocabulary that is regular
increasesquickly when they are not overregularizing, and increasesmore
slowly when they are overregularizing (Marcus et al. 1992).
The results support the traditional explanation of overregularization,
which appeals not to frequency but to different internal mechanisms.
Children at first memorize irregular and regular pasts. Then they discover
that a regularity holds betweenmany regular stemsand their past forms
and create a rule that they apply across the board, including instances
in which a memorized irregular form does not come to mind quickly
enough. The rule is available to fill the gap, resulting in an overregular-
ization. Consistentwith this interpretation, Marcus et al. (1992) found that
children begin to overregularizeat the age at which they first start using
regular forms consistentlyin the past tense; that, presumably, is the point
at which the regular rule has been acquired. As mentioned, the fact that
the regular rule is applied even to high-frequency irregular stems, which
remain high in frequency in children's input throughout development,
showsthat children treat the regular rule as having an unlimited range.
Proponentsof connectionistmodels of languagehave offered two kinds
of counterarguments, but both are inadequate. One is that the RM model
was a two-layer perceptron, and that three-layer models, whose hidden
layer's weights are trained by error-back-propagation, perform much
better (see, e.g., Plunkett and Marchman 1991, 1993; MacWhinney and
Leinbach 1991). However, Sproat (1992), Prasadaand Pinker (1993), and
Marcus (1995) have shown that hidden-layer models have the same
problems as the original RM model. The other is that the effectsof regu-
larity in English come from the fact that regular verbs are in the majority
in English, fostering the broadest generalization. German presents the
crucial comparison. Marcus et al. (1995) reviewed the grammar and
vocabulary statistics of German in detail and documentedthat the parti-
ciple -t and plural -s are found in a minority of words in the language,
compared to irregular alternatives, but nonethelessapply in exactly the
" default" circumstanceswhere accessto memorizedverbs or their sounds
Nature of Human Concepts 241

fails, including novel, unusual-sounding, and derived words (i.e., thefiied

out exampleshave exact analoguesin German). The findings were verified
in two experimentseliciting ratings of novel German words from German
adults. The crosslinguistic comparison suggeststhat default suffixation
arises not becausenumerous regular words reinforce a pattern in asso-
ciative memory but from a memory-independent, symbol-concatenating
mental operation.
In sum, pattern associatorshandle irregular subclassesreasonablywell
but handle the regular class poorly, both in terms of computational
ability and in terms of psychological fidelity . We suggestthat this is a
symptom of the relative suitability of this architecture to handle family
resemblanceand classicalcategoriesin general. The reasons, we suggest,
are straightforward:
. Classicalcategoriesare the product of formal rules.
. Formal rules apply to objects regardlessof their content- that is what
" formal rule" means.
. Pattern associators soak up patterns of correlation among objects'
contents- that is what they are designedto do.
. Therefore, pattern associatorsare not suited to handling classicalcate-
.
gones.
We conclude that the brain contains some kind of nonassociative
architecture, usedin languageand presumably elsewhere.

10.3.5 EpistemologicalCategoriesversusOntological Categories

Rey (1983) has pointed out that even if people can be shown to use pro-
totype (or classical) categories, it doesn't mean that the world contains
prototype (or classical) categories- that is, that the lawful generalizations
of how the world works, as captured by the best scientific description,
make referenceto one kind of category or the other. That raisesa ques-
tion : if there is a psychological distinction between the representations
of prototype and classical categories, is it becausethese representations
accurately reflect different kinds of categoriesin the world? Or does the
human system of categorization arise from some limitation or quirk of
our neurological apparatus that does not necessarilycorrespond to the
lawful groupings in the world?
The questions of what kinds of categoriesare in the mind and what
kinds of categoriesare in the world are clearly related. If the mind evolved
to allow us to grasp and make predictions about the world , the mental
242 Pinker and Prince

systemthat forms conceptual categoriesshould be built around implicit

assumptionsabout the kinds of categoriesthat the world contains, in the
same way that a visual algorithm for recovering structure from motion
might presupposea world with rigid objects and might work best in sit-
uations where the assumptionis satisfied.
Because the English past tense system shows classical and family
resemblancecategoriesbut must have a very different ontology from that
underlying conceptsof tools, vegetables, animals, and other entities that
ordinarily compose conceptual categories, an analysis of the source of
classicaland family resemblancecategoriesin the past tense systemmay
help us to identify the distinctive conditions in which thesetwo kinds of
categoriesarise.

10.3.6 Where Do the Propertiesof the Regular and Irregular Classes

ComeFrom?
The properties of the regular classare simply products of the regular rule.
From any speaker's perspective, the class exists " in the world " in the
sensethat other speakersof the language possessthe rule and use it in
speaking and understanding. This in turn comes from the basic require-
ment for parity in any communicative system. Language can only func-
tion if the rule systemthat generatesfoffi1s is shared by a community of
speakers. Thus, one person's use of a past tenserule (or any rule) in pro-
duction presupposesthat that samerule is in the head of the listener and
will be used to interpret the produced foffi1. Similarly, use of a rule in
comprehension presupposesthat the speakerused it in programming his
or her speech. So the answerto the question " What classof entities in the
world is picked out by a rule-generatedclass, such as the regular verbs?"
is " The class of entities that can be generatedby a replica of that rule in
other speakers' minds."
For the irregulars, the issueis more complex. Of course, irregulars, like
regulars, are usable only becausethey are shared by other speakers. But
unlike the caseof regulars, where the rule is so simple and efficient that it
naturally fits into a grammar sharedby all membersof a community, the
composition of the irregular classis so seeminglyillogical that one must
ask how other speakerscame to possessit to begin with .
Earlier we suggestedthat the family resemblancestructure of the irreg-
ular past tense subclassesis related to the fact that irregulars must be
memorized and human memory has an easier time with family resem-
blance categories (Rosch and Mervis 1975). Interestingly, the obvious
Nature of Human Concepts 243

Darwinian metaphor, in which the most easily memorized verbs survive,

does not apply to the psychology of the child doing the learning. Note
that family resemblancestructure is not a property that some individual
verbs have and others lack, but a property of an entire classof verbs. But
unlike the subjects of Rosch and Mervis's experiment, children are not
given two classesto learn, one with a random organization, the other with
a family resemblancestructure, with the latter being better retained in
memory. One might supposethat the similarity of a verb to other verbs
affectshow easy it is for the child to memorize that verb, and that in the
aggregate, a family resemblancestructure arises. But this, too, does not
properly characterizethe acquisition of irregular forms. There is relatively
little changein the composition of the subclassesbetweenone generation
and the next; children end up pretty much learning the same irregulars
that their parents learned. Moreover, if the children's memory really
shapedtheir irregular classes , we would expect them to arrive at classical
categories, not family resemblancecategories. For example, a rule that
said " All verbs ending in ing go to ang" would have much higher inter-
item similarity than the current English ing class, so verbs like bring
would be even easier to memorize. In fact, given children's ability .to
regularize the irregular verbs by assimilating them to the regular rule
(bringed), they would be in a position to obliterate irregularity altogether
if their memory were all that fragile.
A more accurate version of the Darwinian metaphor would point to
effectsof memory not in the child doing the learning in a given generation
but in the children (and adults) of previous generationswhose learning
shapedthe input to the current generation. Even though each generation
reproduces the previous generation's irregulars with high accuracy,
changesoccasionally creep in. These can be characterized as a kind of
convergent evolution toward certain forms. For example, some lower-
frequency irregular verbs may be consistently regularized in a given gen-
eration, and this might be more likely for verbs that were most dissimilar
from other irregulars and hencemost weakly protected from forgetting.
(Marcus et al. (1992) documentedthat irregulars that are more dissimilar
from other irregulars are more prone to being overregularized by chil-
dren.) In the other direction, some regulars might be attracted into an
irregular class becauseof their high similarity to existing irregulars, as
is happening with sneak/snuck (cf. stick/stuck, string/strung, etc.). If some
of these occasional forgettings and analogies get fixed in a language
community in a contagion-like process(seeCavalli-Sforza and Feldman
244 PinkerandPrince

1981) and accumulate acrossgenerations, classesof verbs with a family

resemblancestructure can arise. The past tenseforms quit and knelt, for
example, are fairly recent additions to the language and are presumably
irregular because of their similarity to verbs like hit and feel. This
phenomenoncan be seeneven more clearly in the more rapid processof
dialect formation in smaller communities, where forms such as bring-
brang, slide-slud, and drag-drug are common (seeMencken 1936).
Though this convergent evolution process surely occurs, it cannot
explain the entire structure of the irregulars in English. First, it does not
capture the historical facts completely. The language never contained
arbitrary irregular classeswhose memberswere attracted into or drifted
out of prototype classesbecauseof fussy learners, leaving the next gener-
ation with a slightly more orderly classthan they had found. Rather, as
we will illustrate, the irregular subclassesare in evidencefrom the earliest
sources. Second, the account posits a kind of harmony betweenproperties
of the memory of one generation and properties of the memory of suc-
ceeding generations: the errors of forgetting and assimilation of genera-
tion n result in a stimulus set that is easierfor generationn + 1 to acquire
without error, becausethe memory of both generationsis biased toward
remembering items that are similar along multiple dimensions to other
items. But in doing so it begsthe question of why memorization of cate-
gories in any generation should be biased toward partial similarities to
begin with . Why does memory work that way? Why not do away with
rememberingpatterns of irregularity altogether and give the next genera-
tion a nice regular class?
In the history of English, divergencehas been the more prominent
trend. That confronted learnersin each generationwith the task of learn-
ing classeswhose family resemblancestructure was not simply causedby
the psychology of previous generationof learners. The Old English strong
classes , ancestorsof most of today's irregular verbs, evolved out of classes
that can be traced back to Proto-Indo-European. Many scholarsbelieve
that the Proto-Indo-European classeswere defined by regular rules: the
number and type of segmentsfollowing the vowel within the stem deter-
mined the kind of changethe vowel underwent (Johnson 1986; Prokosch
1939; Campbell 1959). By the time of Old English, the patterns were more
complicated, but they were still more pervasive and productive and tol-
erated fewer arbitrary exceptions than the alternations in the modern
English irregular subclasses . That is, many stems that are now regular
but fit the characteristic pattern of an irregular subclassin fact used to
Nature of Human Concepts 245

undergo the irregular change: deem/dempt, leanfleant, chide/chid, seem/

sempt, believe/beleft, gl~eet/gret, heat/het, bite/bote, slide/slade, abide/abode,
fare/fore, help/holp, and many others. Furthermore, there was a moderate
degreeof productivity within the classes(Johnson 1986).
Beginning in the Middle English period, there was an even greater
decline in the productivity and systematicityof the strong past tensesub-
classes . The main causeswere the huge influx of new words from Latin
and French that needed a general, condition-free past tense operation,
and the widespread shifts in vowel pronunciation that obscured regu-
larities in the vowel change operations. The " weak" suffixing operation
was already being used for verbs derived from nouns in Old English,
which did not fit the sound patterns defining the strong classesof verbs, so
their extensionto borrowed words was natural (seeMarcus et al. 1995for
further discussion).
In sum, there has been a consistenttrend in the history of English since
the Proto-Indo-European period for the strong classes , originally defined
by phonological properties of their stems, to becomelists of items to be
learned individually . This had an interestingconsequence . Originally , lists
would have been relatively homogeneous, owing to their once having
been generatedby rule-like operations. But then, a variety of unrelated
processes , operating on individual items, destroyed the homogeneity of
the classes . Here are three examples:
Phonologicalchange The verbs blow, grow, throwJ know, draw, fly , slay all
begin with a ( consonant-sonorant) cluster except for know. The reason
that know is exceptional leaps from the page in the way it is spelled. As it
was originally pronounced, with an initial k, it did fit the pattern; when
syllable-initial kn mutated to n within the sound pattern of the language
as a whole, know was left strandedas an exception within its subclass.
Morphological category collapse In Old English, past tenseswere dis-
tinguished by person and number. For example, sing had a paradigm that
we can simplify as follows:
(5) Singular Plural
1st sang sung
2nd sung sung
3rd sang sung
When the number distinctions collapsed, eachverb had to pick a form for
its past tenseas if playing musical chairs. Different verbs made different
choices; hence, we have sing/sang/sung alongside sling/slung/slung. The
contrast betweenfreeze/ froze and cleave/cleft has a similar cause.
246 Pinker and Prince

Attrition Earlier, the class in which I dl changedto It I had the following

members: bend, lend, send, spend, blend, wend, rend, shend, build, geld,
gild, gird (Bybee and Slobin 1982b). The classis succinctly characterized
as containing a vowel followed by a sonorant followed by d. In modem
American English, the verbs geld, gird, gild, wend, rend, and shendare
now obsoleteor obscure. The residue of the classhas five members, four
rhyming with end and one with ild. Although logically it still can be
characterizedas ending in a <vowel-sonorant-d) cluster, the presenceof
regular verbs ending in eld and ird, and the highly specificnature of the
rhyme with end, makesit more natural to representthe classas containing
verbs that rhyme with end but with one exception.
We conclude that a class of items that originally is homogeneous
becauseit is generatedby a rule can acquire a family resemblancestructure
by divergent evolution once the rule ceasesto operate and the effects of
unrelated processesacting on individual members accumulate through
history. Superimposedon thesepatterns is a convergentprocessin which
the accumulated effects of the analogizing and forgetting tendenciesof
previous generationsof learnerscausepartly similar forms to accreteonto
an existing class. Thus, a learner in a single generationis confronted with
family resemblancestructures as products of thesedivergent and conver-
gent historical processes , and these structures can be said to exist in the
world independentof his or her psychology.

10.4 Implicationsfor Conceptual

Categories

Do the discoveriesabout classical and family resemblancecategoriesin

past tensefonns offer insight into the role of classical and family resem-
blance categoriesin the domain of conceptual categorieslike birds and
mothers? To answer this question, we must begin by considering what
conceptualcategoriesare for .

10.4.1 The Function of ConceptualCategories: Inferenceof Unobserved

Properties
No two objectsare exactly alike. So why do we useconceptualcategories?
Why don't we treat every objects as the unique individual that it is? And
why do we form the categorieswe do? Why lump together salmon, min-
now, and sharks, as opposedto sharks, leaves, and spaghetti? Sometimes
it is suggestedthat peopleneedcategoriesto reducememory or processing
load. But given that the cortex has on the order of a trillion synapsesand
Natureof HumanConcepts 247

that long-term memory is often characterizedas " infinite," the suggestion

carries little force. Furthermore, for many categories(e.g., months, base-
ball teams, one's friends) both the category and every individual member
of it are stored in memory.
Bobick (1987), Shepard(1987), and Anderson (1990) have attempted to
reverse-engineerhuman conceptualcategoriesin terms of their function in
people's dealingswith the world . They have independentlyproposedthat
categoriesare useful becausethey allow us to infer objects' unobserved
properties from their observed properties (see also Rosch 1978; Quine
1969). Though we cannot know everything about an object, we can
observesomethings; the observedproperties allow us to assignthe object
to a category, and the structure of the category then allows us to infer the
values of the object's unobservedproperties. Categoriesat different levels
of a hierarchy (e.g., cocker spaniels, dogs, mammals, vertebrates, animals,
living things) are useful becausethey allow a variety of trade-offs between
the easeof categorization and the power of the licensed inference. For
low-level, specific categories, one must know a lot about the object to
know that it belongs in the category, but one can then infer many
unobservedaspectsof the nature of the object. For high-level, generalcate-
gories, one need know only a few properties of an object to know it
belongs to the category, but one can infer only a few of its unobserved
properties once it is thus categorized.
To be concrete: Knowing that Peter is a cottontail, we can predict that
he grows, breathes, moves, was suckled, inhabits open country or wood-
land clearings, spreadstularemia, and can contract myxomatosis. If we
knew only that that he was a mammal, the list would include only grow-
ing, breathing, moving, and being suckled. If we knew only that he was an
animal, it would shrink to growing, breathing, and moving. On the other
hand, it 's much harder to tag Peter as a cottontail than as a mammal or
an animal. To tag him as a mammal, we need only notice that he is furry
and moving, but to tag him as a cottontail , we have to notice that he is
long-eared, short-tailed, and long-hind-legged, and that he has white on
the underside of his tail . To identify ,-'ery specific categories, we have to
examine so many properties that there would be few left to predict. Most
of our everyday categoriesare somewherein the middle: " rabbit," not
mammal or cottontail; " car," not vehicle or Ford Tempo; " chair," not
furniture or Barcalounger. They represent a compromise between how
hard it is to identify the category and how much good the category does.
These compromises correspond to Rosch's (1978) notion of the " basic
level" of a category.
248 PinkerandPrince

We can get away with inductive leapsbasedon categoriesonly because

the world works in certain ways. Objects are not randomly distributed
through the multidimensional spaceof properties that humans are inter-
ested in; they cluster in regions of co-occurring properties that Bobick
calls " natural modes" and Shepardcalls " consequentialregions." These
modes are the result of the laws of form and function that govern
the processesthat create and preserveobjects. For example, the laws of
physics dictate that objects denser than water will be found on lake
bottoms rather than lake surfaces. Laws of physics and biology dictate
that objects that move quickly through fluid media have streamlined
shapes, and bigger objects tend to have thicker legs. If we know some of
the coordinates of an object in property space, the existenceof natural
modes allows us to infer (at least probabilistically) some of its unknown
coordinates.

10.4.2 Classical Categories: Inferenceswithin Idealized Lawful Systems

All this raisesthe question of what kinds of regularities in the world gen-
erate natural modes that humans can exploit by forming concepts. In the
most generalsense, regularities in the world are the result of scientific and
mathematical laws (e.g., of physics, geometry, physiology). Laws can be
captured in formal systems, given a suitable idealization of the world . By
" formal system" we mean a symbol manipulation scheme, consisting of
a set of propositions and a set of inferencerules that apply to the propo-
sition by virtue of their form alone, so that any knowledge not explicitly
stated in the proposition cannot effect the inferencesmade within it . For-
mal systems, we suggest, are the contexts in which classicalcategoriesare
defined. Therefore, under whatever idealization of the world a set of
scientificor mathematicallaws applies, the world containsclassicalcatego-
ries. For example, when the texture, material, thickness, and microscopi-
cally ragged edgesof real-world objects are provisionally ignored, some
can be idealized as plane geometryfigures. Under this idealization, objects
with two equal sidescan be assignedto the category " isoscelestriangle."
Once the object is assignedto the category, one can make the inference
that it also has two equal angles, among other things. Frictionless planes,
ideal gases, randomly interbreeding local populations, and uniform com-
munities of undistractable speaker-hearers are other idealizations under
which regularities in the behavior of objects can be captured in formal
systems. A smart organism could use formal systemsas idealizations of
the world to infer unknown properties from known ones. In the psychol-
ogy of categorization, no lessthan in the history of science, idealization or
Nature of Human Concepts 249

selective ignoring of salient correlational structure is crucial to appre-

hending causal laws.
We suggest, then, that wherever classicalcategoriesare to be found in
human cognition, they will be part of a mentally representedformal
systemallowing deductions to be made. Given the function of concepts,
why else would one bother to assign an object to a classical category?
What is unnatural, then, about traditional experiments in concept for-
mation, such as those of Bruner, Goodnow, and Austin (1956), in which
subjectslearn categorieslike " red square with two borders," is not that
the categorieshave sharp boundaries or necessaryand sufficient condi-
tions, but that the categoriesare not part of a systemallowing interesting
inferences to be drawn- they are unnatural becausethey are literally
useless.
Though one tends to think of formal systemsas the province of sys-
tematic education in modern societies, there are a variety of kinds of
formal systemscapturing inference-supporting regularities that could be
accessibleto people, including those in preindustrial and preagricultural
societies. For example, bodies of folk scienceneed not resemble their
counterpartsin modern scientific systems, but they can reproducesomeof
their visible predictions with alternative means. Mathematical intuitions
too are incorporated into many other systemsof common knowledge.
Here are someexamples:
. Arithmetic , with classical categorieslike " a set of three objects," sup-
porting inferenceslike " cannot be divided into two equal parts," inde-
pendent of the properties of objects that can be grouped into threes.
. Geometry, with classicalcategorieslike " circle," supporting inferences
like " all points equidistant from the center" or " circumferenceis a con-
stant multiple of diameter," regardlessof whether previously encountered
circles are sectionsof tree trunks or drawings in sand.
. Logic, with classicalcategorieslike " disjunctive proposition," support-
ing inferenceslike " is true if its second part is true" or " is false if the
negationsof both its parts are true."
. Folk biology, with classical categories like " toad of kind x," which
support inferenceslike " extract of mouth gland when boiled and dried is
poisonous," regardlessof its similarities to nonpoisonoustoads or its dis-
similarities to other poisonoustoads.
. Folk physiology, with the famous all-or-none category " pregnant,"
supporting the inferences" female," " nonvirgin," and " future mother,"
regardlessof weight or body shape.
250 PinkerandPrince

In addition, the world of humans contains other humans, and there is

reasonto expectmentally representedformal systemsto arise that govern
the conduct of humans with one another. Given the fuzziness and
experience-dependentindividual variation inherent to family resemblance
categories, it is not surprising that conflicts of interest betweenindividuals
will often be resolved by reasoning within systemsthat have a classical
structure, allowing all-or-none decisionswhose basis can be agreedto by
all parties. There is a rationale to assigningdrinking privileges to people
after their twenty-first birthday, arbitrary though that is, rather than
attempting to ascertain the emotional maturity of each individual when
he or she asks for a drink . Furthermore, Freyd (1983) and Smolensky
(1988) have suggestedthat certain kinds of socially transmitted knowl-
edge are likely to assumethe form of discretesymbol systemsbecauseof
constraints on the channels of communication with which they must be
communicatedbetweenindividuals and transmitted betweengenerations.
It is not hard to identify formal systemsinvolved in social interactions
that define classicalcategories:
. Kinship, with classicalcategorieslike " grandmother of X ," supporting
inferences like " may be the mother of X 's uncle or aunt" or " is the
daughter of one of X 's great-grandparents," regardlessof hair color or
propensity to bake muffins.
. Sociopolitical structure, with classical categories like " president" or
" chief," supporting inferences like " carries out decisions on entering
wars," regardlessof physical strength, height, sex, and so on.
. Law, with classical categorieslike " felon," supporting inferenceslike
" cannot hold public office," regardlessof presenceor absenceof a sinister
appearance, social class, and so on.
. Language, with the category " verb," supporting the inference " has a
past tenseform suffixedwith [d] unlessit has an irregular root," regardless
of its phonological properties.
It is unlikely to be a coincidencethat humans uniquely and nearly uni-
versally have language, counting systems, folk science, kinship systems,
music, and law. As we have discussed , classicalcategoriesderiving from
formal systemsrequire a neural architecture that is capable of ignoring
the statistical microstructure of the properties of the exemplarsof a cate-
gory that an individual has encountered. One can speculate that the
development of a nonassociativeneural architecture suitable to formal
systemswas a critical event in the evolution of human intelligence.
Nature of Human Concepts 251

10.4.3 Family ResemblanceCategories: Inferenceswithin Historically

Related Similarity Clusters
In a previous section, we showed that learners of English are presented
with a family resemblancestructure and must cope with it if they are to
speakthe samelanguageas their parents. Are there caseswhere learners
of conceptualcategoriesare similarly forced to cope with a family resem-
blance structure in nature if they are to be able to make inferences
about it? Many people have noted similarities betweenlinguistic and bio-
logical evolution (see, e.g., Cavalli-Sforza and Feldman 1981), and there
is a particularly compelling analogy in the foffi1ation of family resem-
blance categoriesin the evolution of biological taxa.
It is generally believed that a novel speciesevolvesfrom a small inter-
breeding population occupying a local, hence relatively homogeneous
stable environment. Through natural selection, the organisms become
adapted to the local environment, with the adaptive traits spreading
through the population via sexual reproduction. As a result, the popula-
tion assumesa morphology that is relatively uniform- sinceselectionacts
to reduce variation (Sober 1984; Ridley 1986)- and predictable in part
from engineeringconsiderationsto the extent that the organism's niche
and selection pressurescan be identified (Hutchinson 1959; Williams,
1966; Dawkins 1986).
Subsequentgeographic dispersal can causethe members of the ances-
tral population to form reproductively isolated subgroups. They are no
longer homogenizedby interbreeding, and no longer subject to the same
set of selection pressuresimposed by a local environment. In the first
generation following dispersal, the speciesis still homogeneous. Then, a
set of distinct processesdestroys the homogeneity of class: genetic drift ,
local geographic and climatic changesimposing new selection pressures ,
adaptive radiations following entry into empty environments, and local
extinctions. As a result, the descendantsof the ancestral speciesform
a family resemblancecategory- the category of " birds," for example.
Robins, penguins, and ostriches share many features (e.g., feathers)
becauseof their common ancestry from a single population adapted to
flying, while differing becauseof independentprocessesapplying to dif-
ferent membersof that population through history.
This suggeststhat as in the casesof irregular past tensesubclasses, the
family resemblancestructure of many biological taxa comes from the
world, not just the minds of those learning about them. Note that such
family resemblancestructures are not always identical with classically
 252 Pinker
andPrince

defined categories, and they may be indispensableeven in the best scien-

tific theories. Many traditional biological taxa are somewhat arbitrary ,
serving as useful summariesof similar kinds of organisms. There are, to
be sure, somebiological categoriesthat are well defined, including species
(a population of interbreeding organisms sharing a common gene pool)
and monophyletic groups or clades (all the descendantsof a common
ancestor also belonging to the category). But many important biologi-
cal taxa are neither. For example, fish comprise thousands of species ,
including coelocanthsand trout . But the most recent common ancestor
of coelocanthsand trout is also an ancestor of mammals. Therefore, no
branch of the genealogicaltree of organismscorrespondsto all and only
fish; trout and coelocanth are grouped together and distinguished from
mammals by virtue of their many shared properties. To some biologists
this is reasonto deny the scientific significanceof the category altogether,
but most probably agreewith the sentiment captured by Gould when he
writes, " A coelocanth looks like a fish, tasteslike a fish, acts like a fish,
and therefore- in some legitimate sensebeyond hidebound tradition - is
a fish" (1983, 363). In other words, biologists often recognizea category
that is characterizedas a cluster of co-occurring properties. Indeed, some
taxonomists have tried to characterizetaxa with the help of clustering al-
gorithms that usecriteria similar to thosethought to lead to the formation of
prototype conceptualcategoriesin humans(seeRidley 1986; Bobick 1987).
Thus, we have noted two examples of family resemblancecategories
that exist in the world and that have the same genesis: a law-governed
processcreating a relatively homogeneousclass, followed by a cessation
of the influence of the processand the operation of independenthistorical
causesthat heterogenizethe class, though not to such an extent that the
intermember similarities are obliterated entirely. Sinceobjects can escape
the direct influence of laws while retaining some of their effects, a smart
organism cannot count on always being able to capture the world 's regu-
larities in formal systems. For example, no observer knowing only the
United StatesConstitution would be able to explain why presidentsare
always wealthy white Christian males. Similarly, presumablyno observer,
not even a scientist equipped with a knowledge of physiology and ecol-
ogy, would be able to explain why penguins have feathers, like robins,
rather than fur , like seals. Instead, it will often be best simply to record
the interpredictive contingencies among objects' properties to infer
unknown properties from known ones. Thus, a smart observercan record
the contingenciesamong feathers, wings, egg-laying, beaks, and so on, to
Natureof HumanConcepts 253

note that the world contains a set of objects in which these properties
cluster and to use the presenceof one subset of properties to infer the
likely presenceof others.
Just as irregular subclasseswere shapedby both divergent and conver-
gent historical processes , so in the domain of conceptual categoriesthere
is a convergent processthat can causeobjects to cluster around natural
modes even if the objects are not linked as descendantsof a more homo-
geneous ancestral population. For example, there is no genealogical
account of chairs that parallels the oneswe give for languagesor species .
The similarities among chairs are causedsolely by a convergentprocess,
in which a set of properties repeatedly arises becauseit is particularly
stable and adaptive in a given kind of environment, and severalhistori-
cally unrelated entities evolve to attain that set. Examples from biology
include nonhomologous organs such as the eyes of mammals and of
cephalopods, the wings of bats and of birds, and polyphyletic groups
such as cactuslikeplants (which have evolved succulentleaves, spines, and
corrugated stemsas adaptations to desert climates in severalparts of the
world). As in the caseof divergent evolution discussedabove, there is a
mixture of shared and distinct properties that are respectively caused
by law-governed adaptation and historical accident, though here the
influencesare temporally reversed. For example, although vertebrate and
cephalopod eyesare strikingly similar, in vertebratesthe photoreceptors
point away from the light sourceand incoming light has to passthrough
the optic nerve fibers, whereasin cephalopodsthe photoreceptors point
toward the light in a more sensiblearrangement. The differenceis thought
to have arisen from the different evolutionary starting points defined by
the ancestorsto the two groups, presumably relating to differencesin the
embryological processesthat lay down optic and neural tissue. Artifacts
such as chairs develop via a similar process; for a chair to be useful, it
must have a shapeand material that is suited to the function of holding
people up, but it is also influenced by myriad historical factors such as
style, available materials, and ease of manufacture with contemporary
technology. Social stereotypes, arising from the many historical accidents
that cause certain kinds of people to assumecertain roles, are another
example.
We might expect family resemblancecategoriesto be formed whenever
there is a correlational structure in the properties that people attend to
among sets of objects they care about, and that the world will contain
opportunities for such clustersto form whereverthere are laws that cause
254 PinkerandPrince

properties to be visibly correlated and historical contingencies that cause

the correlation to be lessthan perfect - which is to say, almost everywhere .

10.4.4 InteractionsbetweenClassical and Family Resemblance

Categories
The referents of many words, such as bird and grandmother, appear to
have properties of both classicaland family resemblancecategories. How
are thesetwo systemsto be reconciled? The distinction betweencoresused
for reasoningand stereotypesusedfor identification was of no help in the
case of English past tense forms, and the distinction does not do much
better when applied to conceptual categories. Many classical categories
have no family resemblance identification procedure associated with
them, for example, the number " - 3." Many family resemblancecatego-
ries have no classicalcategory serving as a core that they identify, such as
" seafood" or Wittgenstein's famous example, " game." Furthermore,
some classicalcategoriescan be identified by simple, easily computable,
all-or-none tests. For example, odd numbers can be quickly identified by
testssuch as " Divide by 2 and check for remainder" or " Seeif last digit is
1, 3, 5, 7, or 9" ; in fact, the featuresof the associatedfamily resemblance
class, such as " has many odd digits" (which Armstrong, Gleitman, and
Gleitman (1983) found to be a feature that led subjectsto judge that a
given number was a better example of the " odd" class), are not even
probabilistically diagnostic. On the other side, family resemblanceclasses
can support nonperceptual reasoning, sometimesquite reliably, such as
" Presidents are well-off," " Vegetables are not served for dessert," or
" Tools have metal in them." We are not denying that categoriesmay
have " cores" in the sensethat somekinds of knowledge are given priority
over others when they conflict, but it does not seemthat this distinction
can be equated either with quick-identification versus reasoning or with
classicalversusfamily resemblancecategories(Armstrong, Gleitman, and
Gleitman (1983) and Rey (1983) mention someof theseproblems).
A more likely reconciliation is that peoplehave parallel mental systems,
one that records the correlational structure among setsof similar objects,
and another that setsup systemsof idealizedlaws. Often a categorywithin
one systemwill be linked to a counterpart within the other. In general, we
might expect family resemblancecategories to be more accessibleto
observersthan classicalcategories. Most objectsin the world are cluttered
by the effectsof the myriad historical processesthat led to their creation
and preservation, obscuring underlying laws. In the lucky caseswhen
Natureof HumanConcepts 255

people are able to seethese laws peeking through the clutter and try to
capture them in idealized systems, the elementsof thesesystemsmay be
seento apply to many of the objects belonging to the family resemblance
clusters that were independently formed through simple observation of
the correlational structure displayed by frequently encountered exem-
plars. In such cases, languagesappear to assignthe sameverbal label to
both. This is what leads to the ambiguity of A penguin is a perfectly good
bird, one of whosereadingsis true, the other false. It also is what leadsto
such paradoxesas the behavior of Armstrong, Gleitman, and Gleitman's
subjects, who could assert both that odd numbers form an all-or-none
category tolerating no intermediate degreesof membershipand that 13 is
a better example of it than 23.
The fact that these systemsare distinct is at the heart of Putnam's
(1975) and Kripke 's (1972) well-known argument that natural kind terms
are not defined by a set of conditions that pick out the members of the
category in the world . Thus, even though we think of " animal" as a nec-
essarypart of the definition of cat, if we were to discoverthat cats were in
fact robots controlled from Mars, we would not conclude that cat no
longer referred to the entities formerly called cats, or that it did refer to
catlike entities on someother planet that really were animals. Rather, the
label cat is rigidly assignedto a set of objects in the world . According to
Putnam, people have a " stereotype" of such objects that helps them with
tentative identification, but they will defer to an expert in establishing
category membershipmore definitively.
Schwartz (1979) points out that such intuitions about natural kind
terms are driven by a belief that their membershave an " underlying trait "
in common. People act as if they believe in the existenceof such a trait
even if they are prepared to accept that their current belief about the
nature of the trait is incorrect, even if they have no idea of the nature of
the trait , indeed even if no one knows the nature of the trait . Schwartz's
analysissuggeststhat people's intuitions are influencedby a metatheory, a
kind of essentialism,that assertsthat the varying forms that an object can
assumeare causally related in terms of their relation to a hidden trait
or essence . These essencesare clearly not family resemblancecategories
since it is possible, indeed typical, that they are not associatedwith any
properties of the relevant objects at all, let alone a cluster of frequently
co-occurring properties (that is, if Putnam and Kripke are correct, there is
no property associatedwith the concept underlying " cat" or " gold" that
cannot be relinquished by a person while he or she still believesthe con-
256 PinkerandPrince

cept to apply to that category of objects). Rather, the hidden essences

must be representedas abstract symbols within internally represented
formal systems, defining slots for particular traits provided by folk or
formal science, and allowing inferences to be made about heredity,
growth, physical structure, change, and behavior. Subsequentresearchby
Keil (1989) and Gelman (Gelman, Coley, and Gottfried 1994) has gath-
ered evidencefor essentialistthinking in preschool children and adults in
nonliterate cultures. More generally, Medin (1989), Murphy (1993), Rips
(1989), and Smith, Langston, and Nisbett (1992) have emphasizedthe
importance of intuitive rule-like theories in the organization of people's
conceptual categories.
In sum, natural kind terms like cat or gold are linked both to stereo-
types, or family resemblancecategories acquired by observing the cor-
relational structure in sets of similar familiar objects, and to abstract
essences, or hidden traits within an intuitive theory that unite an object's
varying appearancesand provide the infrastructure for bits of folk science
and institutionalized science. (The Putnam-Kripke puzzles arise from
thought experimentsin which thesesystemsare separated.)
The human tendency to induce categories from clusters of similar
objectsone has encountered, to construct formal systemsof rules applying
to ideal objects, and to link entities of the two kinds with each other is
probably the root of many apparent paradoxesin the study of concepts
and often within the conceptual systemsthemselves.For example, exactly
this duality can be found in the legal system in the distinction between
reasoning by constitutionality and reasoning by precedent. Legal ques-
tions are commonly resolvedby appealing to precedents, with more simi-
lar prior decisions carrying more weight. However, when the constitu-
tionality of a current decisionis at issue, only a restricted set of principles
is relevant, and similarity to earlier casesmust be ignored.

10.5 Conclusion

It may be surprising to see so many parallels drawn between two phe-

nomena that seemto be in such different domains. We are not claiming
that past tenseforms and conceptual categoriesare alike in all essential
respectsor that they are generatedby a singlecognitive system. But often
widespreadsimilarities in remote domains make the casefor somecom-
mon underlying principles compelling. English past tenseforms come in
two versionsthat are identical in function and at first glancediffer only in
Natureof HumanConcepts 257

size and degreeof uniformity . On closer examination, they turn out to

representtwo distinct systemsthat correspond point for point with clas-
sical and family resemblancecategories, respectively. Moreover, the two
systemsare linked with distinct psychological faculties, developmental
courses, real-world causes, and computational architectures. A funda-
mental distinction must lie at the heart of this duality. Specifically, we
suggest, human concepts can correspond to classical categories or to
family resemblancecategories. Classicalcategoriesare defined by formal
rules and allow us to make inferences within idealized law-governed
systems. Family resemblance categories are defined by correlations
among features in sets of similar memorized exemplars and allow us to
make inferencesabout the observableproducts of history.

Note
1. There are also problems with the model ' s treatment of these phenomena (see
Pinker and Prince 1988; Lachter and Bever 1988; Sproat 1992) .

References

Anderson , J. R . ( 1990) . The adaptive character of thought . Hillsdale , NJ : Erlbaum .

Armstrong , S. L ., Gleitman , L . R ., and Gleitman , H . ( 1983). What some concepts
might not be. Cognition , i3 , 263- 308.
Aronoff , M . ( 1976). Word formation in generative grammar . Cambridge , MA :
MIT Press .

Bobick , A . ( 1987) . Natural object categorization . Unpublished doctoral disserta -

tion , Department of Brain and Cognitive Sciences, MIT .
Bruner , J. S., Goodnow , J., and Austin , G . ( 1956) . A study of thinking . New York :
Wiley .
Bybee, J. L . ( 1985). Morphology . Philadelphia : John Benjamins .
Bybee, J. L . ( 1991) . Natural morphology : The organization of paradigms and
language acquisition . In T . Huebner and C . Ferguson ( Eds .), Crosscurrents in
second language acquisition and linguistic theories. Amsterdam : John Benjamins .
Bybee, J. L ., and Moder , C . L . ( 1983) . Morphological classes as natural catego-
ries . Language , 59 , 251 - 270 .

Bybee, J. L ., and Slobin , D . I . ( 1982a). Rules and schemes in the development and
use of the English past tense. Language , 58, 265- 289.
Bybee, J. L ., and Slobin , D . I . ( 1982b) . Why small children cannot change
language on their own : Suggestions from the English past tense. In A . Ahlqvist
( Ed .), Papers from the 5th International Conference on Historical Linguistics
(Current Issues in Linguistic Theory Vol . 21; Amsterdam Studies in the Theory
and History of Linguistic Science IV ) . Philadelphia : John Benjamins .
 258 PinkerandPrince

Campbell , A . (1959). Old English grammar . Oxford : Oxford University Press.

Cavalli -Sforza , L . L ., and Feldman , M . W . ( 1981) . Cultural transmission and
evolution : A quantitative approach . Princeton , NJ : Princeton University Press.
Dawkins , R . ( 1986). The blind watchmaker . New York : Norton .

Fodor , J. A . ( 1981) . The present status of the innateness controversy . In J. A .

Fodor , Representations . Cambridge , MA : MIT Press.
Fodor , J. A ., Garrett , M . F ., Walker , E . C . T ., and Parkes, C . H . ( 1980) . Against
definitions . Cognition , 8, 263- 367.

Freyd, J. J. (1983). Shareability: The social psychology of epistemology. Cognitive

Science , 7, 191 - 210 .

Gelman , S. A ., Coley , J. D ., and Gottfried , G . M . ( 1994) . Essentialist beliefs in

children : The acquisition of concepts and theories . In L . A . Hirschfeld and S. A .
Gelman ( Eds .), Mapping the mind : Domain specificity in cognition and culture .
New York : Cambridge University Press.
Gordon , P. ( 1985) . Level -ordering in lexical development . Cognition , 21, 73- 93.
Gould , S. J. ( 1983) . What , if anything , is a zebra? In S. J. Gould , Hens ' teeth and
horses ' toes . New York : Norton .

Hooper , J. B . ( 1976). Introduction to natural generative phonology . New York :

Academic Press .

Hutchinson , G . E. ( 1959) . Homage to Santa Rosalia , or why are there so many

kinds of animals . American Naturalist , 93 , 145 - 159 .

Johnson , K . ( 1986) . Fragmentation of strong verb ablaut in Old English . Ohio

State University Working Papers in Linguistics , 34, 108- 122.
Keil , F . C . ( 1989) . Concepts, kinds, and cognitive development. Cambridge , MA :
MIT Press .

Kelly , M . H . ( 1992) . Darwin and psychological theories of classification . Evolu -

tion and Cognition , 2, 79- 97.
Kim , J. J., Marcus , G . F ., Pinker , S., Hollander , M ., and Coppola , M . ( 1994).
Sensitivity of children 's inflection to morphological structure . Journal of Child
Language , 21, 173- 209.
Kim , l . l ., Pinker , S., Prince , A ., and Prasada , S. ( 1991) . Why no mere mortal has
even flown out to center field . Cognitive Science, 15, 173- 218.
Kiparsky , P. ( 1982a). From cyclical to lexical phonology . In H . van der Hulst
and N . Smith (Eds.), The structure of phonological representations
. Dordrecht:
Foris .

Kiparsky, P. (1982b). Lexical phonology and morphology. In I . S. Yang (Ed.),

Linguistics in the morning calm . Seoul: Hanshin .
Konner , M . ( 1982) . The tangled wing . New York : Harper and Row .
Kripke , S. ( 1972). Naming and necessity. In D . Davidson and G . Harman ( Eds.),
Semantics of natural language. Dordrecht : Reidel .
Natureof HumanConcepts 259

Lachter , J., and Bever , T . G . ( 1988) . The relation between linguistic structure and
associative theories of language learning : A constructive critique of some con -
nectionist learning models . Cognition , 28, 195- 247.
Lakoff , G . ( 1987) . Women, fire , and dangerous things: What categories reveal
about the mind . Chicago : University of Chicago Press.
MacWhinney , B ., and Leinbach , J. ( 1991). Implementations are not conceptuali -
zations : Revising the verb learning model . Cognition , 40, 121- 157.
Marcus , G . F . ( 1995) . The acquisition of inflection in children and multilayered
connectionist networks . Cognition , 56, 271- 279.
Marcus , G . F ., Brinkmann , U ., Clahsen , H ., Wiese, R ., and Pinker , S. ( 1995) .
German inflection : The exception that proves the rule . Cognitive Psychology , 29,
189 - 256 .

Marcus , G . , Pinker , S . , Ullman , M . , Hollander , M . , Rosen , T . J ., and Xu , F .

( 1992) . Overregularization in language acquisition . Monographs of the Society for
Research in Child Development , 57(4), Serial No . 228.
McClelland , J. L ., and Rumelhart , D . E . ( 1985). Distributed memory and the
representation of general and specific information . Journal of Experimental Psy-
chology : General, 114, 159- 188.
McClelland , J. L ., Rumelhart , D . E ., and The PDP Research Group . ( 1986) .
Parallel Distributed Processing: Explorations in the microstructure of cognition :
Vol. 2 . Psychological and biological models. Cambridge , MA : MIT Press.
Medin , D . L . ( 1989). Concepts and conceptual structure . American Psychologist ,
44 , 1469 - 1481 .

Mencken , H . ( 1936) . The American language. New York : Knopf .

Murphy , G . L . ( 1993). A rational theory of concepts . In G . H . Bower ( Ed .), The
psychology of learning and motivation : Vol. 29 . New Yark : Academic Press.
Osherson , D . N ., and Smith , E . E . ( 1981) . On the adequacy of prototype theory as
a theory of concepts . Cognition , 15, 35- 58.
Pinker , S. ( 1984) . Language learnability and language development. Cambridge ,
MA : Harvard University Press.
Pinker , S. ( 1991). Rules of language . Science} 253} 530- 535.
Pinker , S., and Mehler , J. ( Eds.). ( 1988) . Connections and symbols. Cambridge ,
MA : MIT Press .

Pinker , S., and Prince , A . ( 1988) . On language and connectionism : Analysis of a

Parallel Distributed Processing model of language acquisition . Cognition , 28,
73 - 193 .

Pinker , S., and Prince , A . ( 1994). Regular and irregular morphology and the psy-
chological status of rules of grammar . In S. D . Lima , R . L . Corrigan , and G . K .
Iverson ( Eds.), The reality of linguistic rules. Philadelphia : John Benjamins .
Plunkett , K ., and Marchman , V . ( 1991) . U -shaped learning and frequency effects
in a multi -layered perceptron : Implications for child language acquisition . Cogni -
tion , 38 , 43 - 102 .
260 Pinker and Prince

Plunkett, K ., and Marchman, V. (1993

). From rote learningto systembuilding.
Cognition, 48, 21- 69.
Prasada
, S., and Pinker, S. (1993
). Generalizationsof regular and irregular
morphology. Languageand CognitiveProcesses
, 8, 1- 56.
Prokosch, E. (1939
). A comparative
Germanic
grammar.Philadelphia
: Linguistic
Societyof America.
Putnam,H. (1975 ). Themeaningof 'meaning ' . In K. Gunderson(Ed.), Language ,
mind, andknowledge . Minneapolis, MN: Universityof MinnesotaPress .
Quine, W. V. O. (1969 ). Natural kinds. In W. V. O. Quine, Natural kindsand
otheressays . New Yark: ColumbiaUniversityPress .
Rey, G. (1983 ). Conceptsandstereotypes . Cognition , 15, 237- 262.
Ridley, M. (1986 ). Theproblemsof evolution . Oxford: OxfordUniversityPress .
Rips, L. J. (1989 ). Similarity, typicality, andcategorization . In S. Vosniadouand
A. Ortony (Eds.), Similarity and analogicalreasoning . New York: Cambridge
UniversityPress .
Rosch, E. (1973
). On the internalstructureof perceptualand semanticcategories
.
In T. E. Moore(Ed.), Cognitivedevelopment andtheacquisitionof language. New
York: AcademicPress .
Rosch, E. (1978 ). Principlesof categorization . In E. RoschandB. B. Lloyd (Eds.),
Cognitionandcategorization . Hinsdale, NJ: Erlbaum.
Rosch, E. (1988 ). Coherences and categorization : A historicalview. In F. Kessel
(Ed.), Thedevelopment of language andaflanguageresearchers : Paperspresented
to RogerBrown. Hillsdale, NJ: Erlbaum.
Rosch, E., and Mervis, C. B. (1975 ). Familyresemblances : Studiesin the internal
representation of categories . CognitivePsychology , 7, 573- 605.
Rumelhart, D. E., and McClelland, J. L. (1986 ). On learningthe pasttensesof
Englishverbs. In J. L. McClelland, D. E. Rumelhart, and the PDP Research
Group, ParallelDistributedProcessing : Explorationsin themicrostructure of cog-
nition.. Vol. 2. Psychological andbiologicalmodels . Cambridge , MA : MIT Press .
Schwartz , S. P. (1979 ). Natural kind terms. Cognition , 7, 301- 315.
Shepard , R. N. (1987 ). Towarda universallaw of generalization for psychological
science . Science , 237, 1317 - 1323.
Slobin, D. I. (1971
). On the learningof morphologicalrules: A reply to Palermo
and Eberhart. In D. I. Slobin(Ed.), Theontogenesisof grammar : A theoretical
symposium . New York: AcademicPress .
Smith, E. E., Langston , C., andNisbett, R. (1992
). Thecasefor rulesin reasoning
.
CognitiveScience , 16, 1- 40.
Smith, E. E., and Medin, D. L. (1981
). Categories
andconcepts
. Cambridge
, MA:
HarvardUniversityPress .
Smith, E. E., Medin, D. L., and Rips, L. J. (1984
). A psychological approachto
concepts: Comments onRey's" Concepts andStereotypes ." Cognition
, 17, 265- 274.
Nature of Human Concepts 261

Smolensky , P . ( 1988 ) . On the proper treatment of connectionism . Behavioral and

Brain Sciences , 11 , 1 - 74 .

Sober , E . ( 1984 ) . The nature of selection . Cambridge , MA : MIT Press .

Sproat , R . ( 1992 ) . Morphology and computation . Cambridge , MA : MIT Press .

Stromswold , K . J . ( 1990 ) . Learnability and the acquisition of auxiliaries . Un -

published doctoral dissertation , Department of Brain and Cognitive Sciences ,

MIT .

Ullman , M . ( 1999 ) . Judgments of English past tense forms : Implications for the

mental computation of English morphology . Language and Cognitive Processes .

Whittlesea , B . W . A . ( 1989 ) . Selective attention , variable processing , and dis -

tributed representation : Preserving particular experiences of general structures . In

R . G . Morris ( Ed . ) , Parallel distributed processing : Implications for psychology

and neuroscience . New York : Oxford University Press .

Williams , G . C . ( 1966 ) . Adaptation and natural selection : A critique of some cur -

rent evolutionary thought . Princeton , NJ : Princeton University Press .

Wittgenstein , L . ( 1953 ) . Philosophical investigations . New York : Macmillan .

Xu , F . , and Pinker , S . ( 1995 ) . Weird past tense forms . Journal of Child Language ,

22 , 531 - 556 .
Chapter 11
Some Evidence for Impaired Myrna Gopnik
Grammars

Over the last few years, specific language impairment (SLI ) has become a
hot topic because it may have the potential to tell us something about the
biological basis of language . I am afraid that comments about this
research, on both sides, have often generated more heat than light . If we
are really interested in the science of it all , then it is important to get the
issues out on the table and seewhich ones we can agree about , which ones
are still outstanding , and how we could resolve them . This chapter is
intended to make a stab at clarifying some of the issues.
The fact that we would even be discussing language and biology in the
same breath would have been unimaginable when I was an undergraduate
in 1955 (though Darwin , himself , did suggest that learning language might
be an instinct ). Over the last 40 years the picture has changed radically .
Research has shown that all languages, despite their seeming differences,
are built on the same general plan and that most children acquire their
native language as easily as they learn to walk upright , without explicit
teaching and with no apparent effort . Newborns can't do it , but it appears
in the first year of life , and the necessary precursors for this achievement
seem to be there at birth . Experiments with very young babies have shown
that humans come equipped with special abilities to selectively pay
attention to and process language ( KuhI1991 ; Kuhl and Meltzoff 1997) .
These data suggest that there is a biological basis to language . It follows
that if language is part of the biological endowment of humans , then
humans must have some genetic properties that build the particular kinds
of brain circuitry that are specialized for human language . If this is true ,
then it would not be surprising to find that some change in this genetic
endowment can interfere with the way brain circuitry is built and thereby
impair the ability to acquire or use language in the normal way . And in
fact this does happen . Though most children do acquire language without
264 Gopnik

any consciouseffort, there are some children who have real trouble with
language, and their linguistic problems persist into adulthood. What
makes this population particularly interesting is that this disorder seems
to be associatedwith some geneticfactors and neurological anomalies. If
this is true, then this situation provides a ready-made natural experiment
to investigatevarious hypothesesabout the biology of language.
This " ready-made" looks easier than it is, however. There are lots of
piecesto the argument that have to be put together, linguistic, psycholin-
guistic, genetic, and neurological, before we can be sure about what we
are seeing. The first problem is to show that at least some casesof this
disorder are really genetic.

11.1 Genetics

Several epidemiological studies have shown that an individual with a

developmentallanguagedisorder is significantly more likely to have a rela-
tive who is also affectedthan is an unaffectedindividual (Tomblin 1989,
1996). Data from the ongoing Genetic Language Impairment Project
conducted at McGill University confirms this pattern of familial cluster-
ing of SLI (Palmour 1997). In this study, we looked at subjectswho had a
clinical history of language impairment, but no history of any other
exclusionarycriteria, suchasimpaired auditory acuity, mental retardation,
or autism. Then we took detailed family histories. Of the 95 subjectsin this
initial study, 53 (55.8%) had at least one affected first- or second-degree
relative. In 21 of the families there was a clear pattern of multigenera-
tional impairment. Our researchteam has also found evidenceof familial
aggregation in language-impaired subjects in England (Gopnik 1990),1
Greece(Dalalakis 1994), and Japan (Fukuda and Fukuda 1994).
So we have convergingevidencefrom a number of independentstudies
that show that languageimpairment clustersin families. The question is,
" Is this pattern the result of somegeneticfactors or is it the result of social
or linguistic factors that prevail in this family?" The latter seemsextremely
unlikely becausesome membersof the family have perfectly normal lan-
guageand other individuals in the samefamily appear to have grammars
that violate the universal properties that are found in language. Within
genetics, the way to distinguish betweenthe influence of the environment
and the influence of genesis to compare the pattern of impairment in
monozygotic (identical) with that in dizygotic (fraternal) twins. If the
crucial factors for clustering are social, then both kinds of twins should
Evidencefor Impaired Grammars 265

look similar. If , on the other hand, the pattern is due to genetic factors,
then the twins with the samegenesshould be more alike than those who
are merely siblings genetically but have shared a similar social context.
Severalindependentstudieshave shown that this disorder is significantly
more concordant in monozygotic twins than in dizygotic twins (Bishop,
North , and Donlan 1995; Tomblin 1997; Tomblin and Buckwalter 1998).
Therefore, both the epidemiological data and the twin studiesstrongly
suggestthat genetic factors are associatedwith this disorder. But genes
do not code for behavior such as language. They code for proteins that
control developmentthat, in turn, may have consequences for the way in
which languagedevelops. To really understandwhat is going on, then, we
have to find somelikely consequences that the geneticfactors could have
for development; and sincelanguagehappensin the brain, the most likely
place to look is neurology.

11.2 Neurology

Previous studieshave documentedneurological anomaliesin the brains of

subjectswith familial languageimpairment (Plante et al. 1991). Our team
has examinedmagneticresonanceimages(MRIs ) from five of the affected
adult membersof the families we have been studying and has found that
" comparedto controls, the CSF (cerebro-spinal fluid)/ grey and CSF/ grey
+ white ratios were significantly high in FLI -adults [familial language
impaiffilent] . . . whereas the grey/white ratio was significantly low. . . .
Thesefindings suggestthat FLI in adults may be associatedwith cortical
atrophy" (Kabani et al. 1997). Though thesedata are all very interesting
and go in the right direction, they are far from the end of the story. The
global patterns documentedin these studies do not provide the kind of
fine-grained picture that we needif we are to really understandthe details
of how the neurological anomalies interact with the linguistic problems.
Our imaging team, headedby Alan Evans, is now developing procedures
that can look at the MRI data in much finer detail, and this has the
potential for clarifying someof thesequestions.
One interesting question is how these neuroanatomical anomalies
arise. We are just beginning to get somehints of how this might happen.
Current evidenceseemsto suggestthat this geneticimpairment interferes
with normal brain developmentin the fetus. Gallagher and Watkin (1997)
used 3D ultrasonic neuroimaging to study in utero brain development
from 24 to 32 weeks of gestational age. Three of the fetuseswere from
266 Gopnik

families that had no history of languageimpairment (- FLI ). The fourth

fetus was from a family that did have a clear history of languageimpair-
ment (+ FLI ). Gallagher and Watkin found that though all the fetuses
had overall brain volumes within the normal range, their patterns of
growth were significantly different. In particular, the + FLI fetus showed
limited growth in those regions of the brain that are associatedwith lan-
guage performance. This evidencefrom fetal development is consistent
with the evidencefrom older subjects.

11.3 Neuropsychology

The data given above suggestthat the geneticfactors associatedwith this

disorder have consequencesfor the development of the brain and that
theseneuroanatomical anomalieshave direct consequences for the acqui-
sition of language. Since we know that genescan have many different
effects, it is to be expectedthat somegenesthat affect languagemay have
pleiotropic effects and so not be " specific" to language. Moreover, the
extensive anomalies that have been observed in brain structure would
make it surprising if these subjects had no other deficits. And , indeed,
clinicians report that some of these language-impaired subjectsdo have
other problems including low performance IQ, dyslexia, spatial rotation
difficulties, depression, and apraxias; however, none of theseother specific
deficits reliably occurs with the language disorder, and there are many
individuals who have one of these other problems without having any
languagedisorder. The question is whether the languageimpairment that
we seein theseindividuals comesfrom a separateand special " language
faculty" that is out of order, or whether somemore general cognitive or
perceptual processingsystemis not functioning and the purported " lan-
guage" problems are merely a result of a breakdown in a much more
general system.
In our own project, Dr . H . Chertkow, a neurologist, has examined 30
of our Canadian subjectsand finds that some, but not all, of the individ -
uals with this languagedisorder do have other nonlinguistic problems and
that some of the membersof thesefamilies who are not impaired on our
diagnostic languagetest also have other problems.
F or example, as shown in table 11.1, in a sample of 12 subjectsfrom
these families, 6 were clearly impaired by our diagnostic language test
and 6 were not. Neurological testing revealedthat 3 of the 6 language-
impaired subjectshad mild impairment in left and right hemispheresen-
268 Gopnik

sorimotor function, but then so did 1 subject who showed no language

impairment by our tests. Two of the language-impaired subjectsexhibited
some impairment in mental calculation, as did 1 non-language-impaired
subject. In general, the language-impaired subjects appear to be more
vulnerable to having other disorders in addition to their language dis-
order, but there seemsto be no necessarypattern to theseother problems.
And , as we have shown, it is by no means the casethat theseother dis-
orders always co-occur with languageproblems. The fact that this disor-
der affects neuroanatomy makes it not surprising that other areasmight
also be implicated in somecases. This doesnot mean, however, that these
other disorders causethe languageimpairment.
The evidence indicates that this linguistic disorder has some genetic
foundation and that the genetic factors produce anomalous neuro-
anatomical structures. If we want to know how this affects language, we
must have a detailed and extensivepicture of preciselywhat goeswrong
with thesesubjects' language.

11.4 Linguistics

Our research goals in linguistics over the years have been clear: to
describethe grammatical systemthat producesthe pattern of errors that
are observed. After 10 years and almost 100 language-impaired subjects
and well more than 100 controls representingfour different native lan-
guages(English, French, Greek, and Japanese ), the data convergeto tell
us that the language-impaired subjects cannot construct normal repre-
sentations for grammatically complex words and they therefore cannot
use rules that depend on the content of theserepresentations. For exam-
ple, though they appear to have no problems with the larger grammatical
categoriesof languagelike noun and verb, they are unable to recognize
that words can be composedof roots and affixes. They therefore do not
construct abstract linguistic features such as tense, number, aspect, case,
or gender, and they do not recognizethat there are linguistic rules that
operate on these features. But this does not mean that they cannot use
words that, from the point of view of the nonnal grammar, have these
features. For example, in spontaneousspeechthey usually use the correct
forms such as " books" or " walked." They can do this becausethey
memorize theseforms as unanalyzed chunks that in their lexicon simply
encodethe semanticmeaning of " more than one" or " in the past" rather
than the sublexicalfeatures" plural" or " past." They are more accuratein
Evidencefor Impaired Grammars 269

using the " more than one" forms than the " in the past" forms. This is
probably becausein English the relationship between the form with the
plural marker and the semanticcontext of " more than one" is very regular
and direct and therefore is easy to memorize. However, the relationship
between the past-marked forms and temporal past context is extremely
complicated and not at all regular or direct (Shaer 1996), and therefore it
is more difficult to assign a consistentmeaning relationship betweenthe
inflected foffi1 and the meaning of pastnessby any rote means.
How do we know that theseindividuals do not construct complex rep-
resentations? We have two sourcesof evidence, linguistic and psycholin-
guistic. The linguistic evidencecomesboth from spontaneousspeechand
from a wide seriesof linguistic tests. One of the things that makesparents
notice that their children have difficulties with languageis that they make
mistakes in spontaneousspeechin foffi1s like past tensesand pronouns
long after their playmates and siblings have got the systemfigured out.
And these problems persist into adulthood. But as these individuals get
older, they do not make theseerrors very often. Many of them get tense
right over 80% of the time. This is in contrast to the control subjects, who
get tense right virtually all of the time. Marking tense in your native
language is an automatic, unconsciousprocess, not something that you
get wrong. And this leads us to a very interesting problem. Is someone
who gets tenseright 80% of the time impaired, or simply absent-minded?
Let's put it another way. Generally speaking, people get their own name
right all of the time. If someonegot his or her name right 80% of the time,
we might begin to wonder. What our model suggestsis that the language-
impaired subjects, in fact, never get tense right . They do not have the
category " tense" in their grammar. What they do 80% of the time is
produce a word that has the samesurfaceform as the marked foffi1 in the
normal grammar. It looks like they get tensemarking right most of the
time, but they get there by a very different route- by memorization or by
explicit rule (Paradis and Gopnik 1997).
The only way to figure out what these individuals really know about
language is to give them linguistically significant tests. The data from
thesediagnostic testsconvergeto tell us that the pattern of what they can
do and what they cannot do is strikingly similar in all of the five pop-
ulations- speakersof English from Canada and England, and speakersof
French, Greek, and Japanese - that we have looked at (seefigure 11.1).
These tests show that the language-impaired speakershave significant
problems with grammaticality judgments, tense production, and deriva-
Evidencefor Impaired Grammars 271

Table 11.2
Ability to produce tense marking (% correct). Subjects were given items like
" Every day I walk to school. Just like every day, yesterday I ." This task
requiresthe subject to recognizethat the temporal context specifiedin the second
sentencerequires a particular verb form .
English English French Japanese Greek
(England) (Canada) (Canada)
Language-impaired
subjects 38.3 52.3 46.7 48.1 20.0
Controls 91.7 93.5 96.4 97.9 87.1

tional morphology.2 They have fewer problems with a pointing task that
requires them to auditorily distinguish between words like " book" and
" books" or " books" and " cooks." These general diagnostic tests have
beensupplementedby much more detailed testsdesignedto test particular
hypotheses. For example, we have data about tense from spontaneous
speech, grammaticality judgment tasks, grammaticality rating tasks,
storytelling, and tenseproduction; and theseresults are consistentover a
wide range of types of stimuli and responses - oral, aural, written, point-
ing, and so on (Gopnik 1994). We also have results that show that these
speakershave problems with plurals and comparatives.
One question was whether theseproblems with tensein English could
be accountedfor by the fact that regular past tensein English is encoded
by a form that has been described as " phonologically vulnerable"
(Fletcher 1990) or by somespecialand particular problems with our first
set of English-speaking subjects, who all came from one particularly
interesting family . In order to resolvethis question, we decidedto look at
more English-speaking subjects and at subjects whose native language
was not English. These data show that the problem with tenseis unique
neither to our first subjectsnor to English. It showsup in every popula-
tion we have looked at (seetable 11.2; seealso Fukuda 1994; Dalalakis
1996; Gopnik et al. 1996; Royle 1996; Gopnik 1998; Ullman and Gopnik ,
to appear) . All of thesepopulations have similar problems with produc-
ing correct tense forms, no matter how tense is encoded in their native
language- in a final stressedsyllable as in French or in a three-syllable
element like " mashita" in Japanese . Interestingly, both the Greek con-
trol and language-impaired subjects have lower scores than the other
subjects; this happens becauseeach verb in Greek has more than 60
272 Gopnik
Table 11 .3

Forms of Greek " lik -os" (" wolf " )

Singular Plural

Nominative Iik os Iik i

Genitive Iik u Iik on

Accusative Iik 0 Iik us

Vocative Iik e (!) Iik i

different inflected forms, as opposed to 4 in English, so there are sim-

ply many more ways to go wrong and it is harder to be right just by
chance.
The next question that we had to addresswas whether the subjects'
problem with language was confined to tense, as has been suggested
(Rice, Wexler, and Cleave 1995), or whether they made similar errors in
other parts of the inflectional system. Tests show that across the four
languagesthey make errors in number, case, and gender.3
All of this led us to think that these individuals had a particular
problem with inflectional rules, but Jenny Dalalakis's innovative work on
compounds and diminutives in Greek made it clear that the language-
impaired subjectshave just as much difficulty with finding the root of a
complex word as they have with adding an inflection (Dalalakis 1996). It
is not merely that they cannot add an inflection to a root; they do not
even recognize that there are roots and inflections. There is no way to
check this in English becausethere are words in English like ' 'walk" and
" book" that do not have any overt inflectional marking. In Greek, how-
ever, all nouns and verbs are inflected. For example, the word for " wolf "
has eight different forms (seetable 11.3). It would seemto be child's play
to figure out that " lik " is the root and all of the rest are inflectional forms.
But not for these children. They make serious errors in tasks like com-
pounding and diminutivization that depend upon the manipulation of
roots. They seemto know that they are supposedto take part of the word,
but they are not sensitiveto the boundariesbetweenroots and inflections
so sometimesthey take a part that is shorter than the root and sometimes
they take a part that is longer than the root . The rule for compounding in
Greek requires that the two roots be conjoined (la ), in some caseswith
the insertion of an " 0" that marks the morphemeboundary (lb ); then the
inflection is added to the end of the conjoined roots.
Evidencefor Impaired Grammars 273

(1) a. lik -anthrop -os - t

titan thropos
wolf man masc.sing. werewolf
nom .

b . lik -0 - fol -a -+ likofola

wolf compound nest fem .sing. wolfnest

morpheme nom .

(A similar rule holds for forming diminutives . For the details , seeDalalakis
1996.) The subjects were shown pictures and told (in Greek ), for exam-
ple, " This is a mouse (pondikos ) and this is a man (anthropos ) that be-
comes a mouse. We would call him a 'mouseman ' (pondikanthropos )."
They were asked to form similar compounds in response to other pictures .
The language -impaired subjects had significantly more difficulty with this
task than did the controls (see table 11.4) . Sometimes the language -
impaired subjects produced fonDs in which the first root was shorter
than required (2a), and sometimes they produced forms in which it was
longer (2b).
(2) a. anthrop -fagh -os -+ * anthrofaghos
man eater masc.sing. man -eater
nom .

b. lot -fagh -os -+ * loto ~faghos

. . .

perSImmon eater masc .smg . persImmon - eater

nom .

It is clear then that these subjects' problem with inflectional rules is an

epiphenomenon of their problem with knowing that there are roots and
inflections at all .

N ow we are back to one of the problems that we discussed at the

beginning : where do the " inflected " forms that these speakers do produce
come from ? One of the compensatory mechanisms that we have suggested
is memorization - they simply learn inflected words one at a time as

Table 11 .4

Greek compounding and diminutivization (% correct )

Language -
impaired Young Age-matched
sub,jects controls controls

Compounding 16 95 99
Diminutivization 38 100 100
274 Gopnik

everyonehas to do for irregulars. If this is true, then they should be more

successfulwith regularly inflected words that they have heard frequently
than with those that are rare. And that is precisely what we find: fre-
quency is a more robust predictor of problems than regular versusirreg-
ular inflection (Fukuda and Fukuda 1994; Gopnik 1994; Gopnik and
Goad 1997; Ullman and Gopnik , to appear) .
This is precisely the opposite from what we seewith normal controls,
for whom regularly inflected forms do not show any frequency effects.
This strongly suggeststhat memory and not rules are producing the
inflected forms used by language-impaired subjects. To clinch this argu-
ment, we looked at the absolute endpoint of the frequency scale, novel
words that these individuals could never have seenbefore. If they were
using a rule, they should have had no problem with this task. If they were
producing actual forms from memory, then it should have been impossi-
ble for them. We testedsuch forms in all four languagesand over a wide
range of different kinds of rules (for results, seetable 11.5). These data
clearly and convincingly show that the language-impaired subjectsper-
form significantly worse than non-language-impaired subjects on every
grammatical task and for every languagepopulation that we have tested.
One might then ask, " If thesespeakersdo not have the rule, how do they
ever get it right at all?" The answerseemsto be that they do not ever get
the unconscious, automatic rule for inflection right . What they can do is
use an explicit rule- in the caseof the English speakers, for example, a
rule that tells them to add an " -s" for plurals or an " -ed" for past. And
there is clear evidencethat this is what they are doing. In our original
scoring of thesetests, we were very generousand gave the subjectscredit
for producing an inflected form even if it did not sound preciselylike the
normal form . In a very detailed phonological analysis of these subjects'
attempts at producing plurals, Goad and Rebellati (1995) found that even
when the English-speakinglanguage-impaired subjectsdo produce a kind
of sibilant at the end of the word given in the stimulus sentence , they do
not obey the normal rules in English that govern the incorporation of the
affix into the word. For example, they do not have voicing assimilation
and say " wug-s" instead of " wug-z." The plural does not behavelike an
affix in the grammar of the language-impaired subjects. It is as if they
were producing a kind of compound by adding a wordlike " s" that means
" more than one." This careful analysis showsthat only 9% of the forms
that they produce phonologically resemblethe " correct" form . A similar
analysis of the novel verbs shows the same pattern. Apparently, then,
Evidence
forImpaired
Grammars 275
Table 11.5
Ability to mark novel words grammatically (% correct). In each of these tests
the subjects were given a context, usually in pictures, which required that a
grammatical rule be applied to a novel word: " This pencil is weff. This pencil is
even ."
Language-impaired
Controls subjects
Past tense
English (in England) 95.4 38.0
English (in Canada) 93.5 52.3
French (in Canada) 92.6 33.3
Greek 87.1 20.0
Japanese 89.1 37.0
Plurals
English (in England) 95.7 57.0
English (in Canada) 99.2 58~3
Greek 79.8 42.1
Comparatives
English (in England) 74.0 21.0
Compounds
Greek 93.6 12.8
Japanese 80.5 20.2
Diminutives
Greek 83.9 40.2

these speakers produce forms that look " inflected " by very different
means than the normal grammar .
So far , all of our hypotheses about the way that the language-impaired
subjects must be representing words in their grammar have been inferred
from their performance on linguistic tests. But there is another , more direct
way to look at their mental lexicon . Eva Kehayia conducted a series of
on-line psycholinguistic tests to probe the way in which language-
impaired subjects were processing complex words . It might be the case
that these subjects have normal representations for the inflected forms ,
but some processing difficulties make these forms unavailable to them in
spontaneous speech or elicited production tasks. Psycholinguistic research
on aphasics shows that though they make errors in spontaneous speech
and in linguistic tests, their mental representations appear to be intact
( Kehayia and Jarema 1994; Kehayia , in press) . The language-impaired
 276 Gopnik

subjectsin thesetestswere 12 speakersof Canadian English, 7 speakersof

French, and 4 speakersof Greek (in Greece). They were compared with
24 controls in each language. The general form of the tests required the
subjectsto look at a single word that appearedon a computer screenand
decide whether it was a real word in their language or not. The word
stayed on the screen until the subject pushed the button to register a
decision. Half of the testswere simple lexical decisiontasks and half were
primed lexical decision tasks. The experimental model assumesthat the
amount of time that it takes a subject to make a decision is a measureof
the way in which the subject is processingthe item. Though the subjects
thought that we were interestedin their ability to judge whether a word
was a real word or a novel word, the real experimental variables inves-
tigated inflection, compounding, and derivation. The results are consis-
tent acrossall of the tests and all of the languages(seetable 11.6). Both
groups are above 90% accurate in judging whether a sequenceis or is
not a word; however, they are very different in the way that they process
the words and nonwords. In the processingexperiments, the English-,
French-, and Greek-speaking control subjectsall show a significant dif-
ferencein reaction times, more than 30 milliseconds, for inflected versus
uninflected words and nonwords. Affected individuals show no such dif-
ference. This indicates that the language-impaired subjects, crosslinguisti-
cally, are building mental representationsthat differ in nature from those
of non-language-impaired individuals.
These results suggestthat these subjectsare insensitive to the internal
structure of words. They treat all words as if they were simple unana-
lyzed chunks. They are good at knowing if a word is a real word or not:
they simply look it up in their mental lexicon. In some cases, this takes
them longer than controls becausethey have more items to search, but in
other cases(e.g., Greek compounds), where the rules for forming the
complex word are intricate, their whole-word strategy is actually more
efficient. This result is illustrated in table 11.7. Notice that the language-
impaired individuals process all three forms in the same way, whereas
the controls processthem significantly differently. The controls perform
fasteston the existing compoundsand sloweston the compoundsthat are
made up of real words but that do not exist in the language. What is
particularly interesting is that the language-impaired subjects' whole-
word lookup strategy makes them perform significantly faster than
the controls on nonword compounds and especially faster on novel
compounds.
Evidencefor Impaired Grammars

I
278 Gopnik

Table 11 .7

Reaction time (in ms)

Language-impaired
subjects(Greek) Controls (Greek)
N onword compounds
kigofaghos 817 870
(" kigoeater " ; like
English " wugeater " )
Navel compounds
migofaghos
(' 'flyeater " )
Existing compounds
hortofaghos
(" vegetable eater" ==
" vegetarian " )

The results of the experimentsin this study clearly show that the per-
formance pattern of the language-impaired subjects is significantly dif-
ferent from that of the control subjects on word and nonword targets
for inflection, derivation, and compounding. The controls in this study
behavejust as reported elsewherein the literature (Taft and Forster 1975;
Laudanna, Badecker, and Caramazza 1989; Kehayia and Jarema 1994).
They processproductively formed complex words by meansof decompo-
sition, and they processsimple and idiosyncratic complex words by means
of whole-word access . The language-impaired subjectsappear to process
111111111I1
1111111111

all of the items by means of whole-word access(Kehayia 1994, in press;

Dalalakis 1996). The results of these on-line studies are consistent with
what the off-line linguistic tests have shown: the language-impaired sub-
jects do not build complex representationsfor words; instead, they treat
both complex and simple words in the same way- as if they were
unanalyzedchunks.
The question then is, " What do their representationslook like?" There
are two possibilities for representingwords with no internal structure.
(3) a. V [+ past]
walked
" move on foot" (+ past)
b. V
walked
" move on foot in the past"
Evidence
for ImpairedGrammars 279

The difference betweentheserepresentationsis that in (3a) the informa-

tion that this word refers to something that happened in the past is
expressedby a morphological feature (+ past) and in (3b) the concept of
pastnessis expressedas part of the meaning of the word.
As I have argued, the data from a wide range of grammaticality judg-
ment tasks, production tasks, novel-form tasks, and reaction time experi-
ments indicate that these subjects show no sensitivity to the internal
structure of words. These data strongly suggestthere is no evidencethat
the lexical entries of language-impaired speakers contain sublexical
(morphological) featuresfor tense, number, and perhapsother inflectional
properties. Sincethere is no evidencefor such featuresfrom the linguistic
behavior of thesespeakers, we have no grounds for introducing such fea-
tures into their representations. Given that regular pastsend in a coronal
stop in English, and given that language-impaired individuals can some-
times add a pastlike ending to novel words, we assumethat these indi-
viduals store such forms in family resemblanceclassesand that they can
thus perform " wug" -type tasksthrough analogy (Gopnik and Goad 1997).
We therefore have a broad range of crosslinguisticevidence, both on-line
and off-line, that all tells the samestory: that the language-impaired sub-
jects, though they get the same linguistic input as their siblings, do not
build the samesort of grammar. The grammars that they do build appear
to rely on memory and analogy and not productive rules for handling the
morphological aspectsof language. The spontaneousspeechof the older
subjects, and even their performance on some linguistic tests, demon-
strates that a reasonablesimulacrum of morphology can be constructed
by using memory and analogy instead of productive rules. They do not
perform perfectly, but they are not terrible either. It is only when the
subjectsare given teststhat dependon productivity or when their internal
processesare monitored that the full extent of their inabilities can be
detected.
All of this would be interesting evenif it only provided a description of
the difficulties that a small group of language-impaired subjectshave. But
if it is all true, and if this disorder is related to somegeneticfactors that in
turn have consequencesfor neurological development, then apparently
there are some genetic and neurological precursorsthat are required for
the normal courseof theseaspectsof languagedevelopmentto take place.
This does not mean that we think that there is a gene for morphology!
From the point of view of linguistics, there is good evidencethat though
morphology is one of the striking aspectsof languagethat seemsto stump
280 Gopnik

these individuals , it is by no means the only one they have trouble with .
The question that is still open is whether their prosodic and syntactic dif -
ficulties are directly related to the same out -of -order linguistic system that
causes their problems in morphology or whether these other language
problems are an independent result of their general pattern of neurologi -
cal damage . The answer to this question would shed light on the relative
independence of the three levels of grammar . Though there is abundant
evidence that this disorder is associated with some genetic factors , no one
yet knows what these factors are . And even if the genes associated with
this disorder were found , it would not follow necessarily that these same
genes would be the ones that guide the normal linguistic system to develop ;
the factors that make a system function in the normal case might not be
the same as those that can cause the system to break down . Of course ,
there is much more to find out . Weare only at the beginning of this
intellectual journey , but it does appear that it will turn out that Darwin
was right that language is part of the biological nature of humans .

Notes

1 . This family has been the subject of much debate ( Fletcher 1990; Vargha -
Khadem et al . 1995) . For a d'etailed discussion of some of these disputes , see
Gopnik and Goad 1997. From the data , it is clear that the language impair -
ment seen in this family is not caused by oral apraxia or low performance IQ .
Moreover , as I will show , the same pattern of impairment is found in English -
speaking subjects in Canada as well as in French -, Greek -, and Japanese-speaking
subjects.
2. Since Japanese does not have a rich , productive system of derivational mor -
phology , we did not test the Japanese-speaking subjects on this task .
3. They also make errors in syntax and in some phonological processes such as
prosody , but discussing these in detail would take us too far afield here (van der
Lely and Harris 1990; van der Lely 1997; Clahsen 1992; Piggott and Kessler 1994;
Rice , Wexler , and Cleave 1995; Goad , in press) .

References

Bishop , D . V ., North , T ., and Donlan , C . ( 1995) . Genetic basis of specific

language impairment : Evidence from a twin study . Developmental Medicine and
Child Neurology , 37, 56- 71.
Clahsen , H . ( 1992) . Linguistic perspectives on specific language impairment
( Theorie des Lexikons Arbeitspapier 37) . Dusseldorf : Heinrich Heine Universitat .
Dalalakis , J. ( 1994). Familial language impairment in Greek . McGill Working
Papers in Linguistics , 10, 216- 227.
Evidencefor Impaired Grammars 281

Dalalakis , J. ( 1996) . Developmental language impairment : Evidence from Greek

and its implications for morphological representation . Unpublished doctoral dis-
sertation , McGill University .
Fletcher , P. ( 1990) . Untitled scientific correspondence . Nature , 346, 226.
Fukuda , S. ( 1994) . Lexical representation of Japanese complex verbs : A theoreti -
cal model and implication from dysphasic children . McGill Working Papers in
Linguistics , 10, 194- 205.
Fukuda , S. E ., and Fukuda , S. ( 1994) . Developmental language impairment in
Japanese: A linguistic investigation . Mc ~ ill Working Papers in Linguistics , 10,
150 - 177 .

Gallagher , T ., and Watkin , K . ( 1997) . 3D ultrasonic fetal neuroimaging and

familial language disorders : In utero brain development . Journal of Neuro -
linguistics , 10, 187- 201.
Goad , H . (in press) . Plurals in SLL . Prosodic deficit or morphological deficit ?
Language Acquisition .
Goad , H ., and Rebellati , C . (1995) . Pluralization is compounding in SLI . In
P. Koshinen ( Ed .), Proceedings of the Annual Meeting of the Canadian Linguistic
Association . Toronto : University of Toronto , Department of Linguistics .
Gopnik , M . ( 1990). Feature -blindness : A case study . Language Acquisition , 1,
139 - 164 .

Gopnik , M . ( 1994) . Impairments of tense in a familial language disorder . Journal

of Neurolinguistics , 8, 109- 133.
Gopnik , M ., Dalalakis , J., Fukuda , S. E ., Fukuda , S., and Kehayia , E . ( 1996) .
Genetic language impairment : Unruly grammars . In W . G . Runciman , J. Maynard
Smith , and R . I . M . Dunbar ( Eds .), Evolution of social behaviour patterns in pri -
mates and man. Oxford : Oxford University Press.
Gopnik , M ., and Goad , H . ( 1997) . What underlies inflectional error patterns in
genetic dysphasia ? Journal of Neurolinguistics , 10, 109- 137.
Gopnik , M . ( 1998). Familial language impairment : More English evidence. Folia
Phoniatrica et Logopaedica .
Kabani , N ., MacDonald , D ., Math , M ., Evans , A ., and Gopnik , M . ( 1997) .
Neuroanatomical correlates of familial language impairment : A preliminary
report . Journal of Neurolinguistics , 10, 203- 214.
Kehayia , E . (1994) . Whole -word access or decomposition in word recognition in
familial language impairment : A psycholinguistic study . McGill Working Papers
in Linguistics , 10, 123- 128.
Kehayia , E . (in press) . Morphological representation and processing of inflected
and derived words in Greek -speaking individuals . In Proceedings of the 2nd
International Conference on Greek Linguistics .
Kehayia , E., and Jarema , G . ( 1994) . Morphological priming (or Prim # ing ) on
inflected verb forms : A comparative study . Journal ofNeurolinguistics , 2, 83- 94.
 282 Gopnik

Kuhl , P . K . ( 1991 ) . Human adults and human infants show a " perceptual magnet

effect " for the prototypes of speech categories , monkeys do not . Perception &

Psychophysics , 50 , 90 - 107 .

Kuhl , P . K . , and Meltzoff , A . N . ( 1997 ) . Evolution , nativism , and learning in the

development of language and speech . In M . Gopnik ( Ed . ) , The inheritance and

innateness of grammars . New York : Oxford University Press .

Laudanna , A . , Badecker , W . , and Caramazza , A . ( 1989 ) . Priming homographic

stems . Journal ofM ~ mory and Language , 28 , 531 - 546 .

Palmour , R . ( 1997 ) . Genetic studies in specific linguistic impairment . Journal of

Neuro lingu is tics , 10 , 215 - 230 .

Paradis , M . , and Gopnik , M . ( 1997 ) . Compensatory strategies in genetic dyspha -

sia : Declarative memory . Journal of Neurolinguistics , 10 , 173 - 185 .

Piggott , G . L . , and Kessler Robb , M . ( 1994 ) . Prosodic constraints in familial

language impairment : A preliminary report . McGill Working Papers in Linguis -

tics } lO } 16 - 23 .

Plante , E . , Swisher , L . , Vance , R . , and Rapcsak , S . ( 1991 ) . MRI findings in boys

with specific language impairment . Brain and Language , 41 , 52 - 66 .

Rice , M . L . , Wexler , K . , and Cleave , P . ( 1995 ) . Specific language impairment as a

period of extended optional infinitive . Journal of Speech and Hearing Research ,

38 , 850 - 863 .

Royle , P . ( 1996 ) . Verb production in French DLI subjects . Unpublished Master ' s

thesis , McGill U ni v ersi ty .

Shaer , B . ( 1996 ) . Making sense of tense : Tense , time refel ' " ence and linking theory .

Unpublished doctoral dissertation , McGill University .

Taft , M . , and Forster , K . I . ( 1975 ) . Lexical storage and retrieval of prefixed

words . Journal of Verbal Learning and Verbal Behavior , 14 , 638 - 647 .

Tomblin , J . B . ( 1989 ) . Familial concentration of developmental language impair -

ment . Journal of Speech and Hearing Disorders , 54 , 287 - 295 .

Tomblin , J . B . ( 1996 ) . Genetic and environmental contributions to the risk for

specific language impairment . In M . L . Rice ( Ed . ) , Toward a genetics of language .

Mahwah , NJ : Erlbaum .

Tomblin , J . B . ( 1997 ) . Epidemiology of specific language impairment . In M .

Gopnik ( Ed . ) , The inheritance and innateness of grammars . New York : Oxford

University Press .

Tomblin , J . B . , and Buckwalter , P . R . ( 1998 ) . Heritability of poor language

achievement among twins . Journal of Speech , Language & Hearing Research , 41 ,

188 - 199 .

Ullman , M . , and Gopnik , M . ( to appear ) . The production of inflectional mor -

phology in hereditary specific language impairment . Applied Ps } ' cholinguistics , 20 .

van der Lely , H . ( 1997 ) . Language and cognitive development in a grammatical

SLI boy : Modularity and innateness . Journal oj Neurolinguistics , 10 , 75 - 107 .

Evidencefor Impaired Grammars 283

van der Lely, H ., and Harris, M . (1990). Comprehensionof reversiblesentencesin

specificallylanguageimpaired children. Journal of Speechand Hearing Disorders,
55 , 101 - 117 .

Vargha -Khadem , F ., Watkins , K ., Alcock , K ., Fletcher , P., and Passingham , R .

( 1995) . Praxic and nonverbal cognitive deficits in a large family with a genetically
transmitted speech and language disorder . Proceedings of the National Academy of
Sciences , 92 , 930 - 933 .
Chapter 12
The Role of Semantics in Paul Bloom
Solving the Bootstrapping
Problem

Some knowledge of languageis expressedin terms of grammatical cate-

goriessuchasnouns, verbs, and adjectives. For instance, English-speaking
adults know that adjectivesprecedenouns within NPs- we say " the big
dog," not " the dog big." This chapter is about a puzzle that arisesfrom
this fact, first raised by Fodor (1966). Children must learn the linguistic
categories that words and phrases belong to; they must learn that the
English word " dog" is a noun and the English word " big" is an adjective.
But syntactic categoriesdo not reduceto nonsyntactic categories. Nouns,
for instance, do not universally sound the same, or occur in the same
position, or share the sametype of meaning. Pinker (1984) has described
this as " the bootstrapping problem" in language acquisition. How do
children use nonsyntactic information (such as phonology, position, or
meaning) to arrive at syntactic knowledge?
The proposal defendedhere is that children use semantic information
to solve this problem. This is not a new proposal. Macnamara (1982)
proposed one important semantic theory, in which children's first word
combinations are the result of rules that map over semanticcategories-
rules such as " modifier + object word" that generate phrases such as

I have been working on different drafts of this chapter for over five years; the
version presentedhere is a very modified version of what I presentedin my doc-
toral dissertation (Bloom 1990a, chap. 3). I thank SusanCarey, LouAnn Gerken,
Lila Gleitman, Jane Grimshaw, Lisa Menn, and StevenPinker for discussionof
theseissues, and Ellen Courtney, Ray Jackendoff, and Karen Wynn for extensive
comments on a previous draft. I am particularly grateful to John Macnamara,
who first introduced me to the study of mind and language, and encouragedand
guided me in the years that followed.
Preparation of this chapter was supportedby grants from the SpencerFounda-
tion and from the Sloan Foundation.
 286 Bloom

" good boy " and " nice hat . " Since these rules cannot account for much

of what children hear ( e . g . , " good story " and " nice smile " ) , they are

forced to gradually expand and transform categories such as " object

word " until they correspond to syntactic categories such as " noun ." As

Macnamara ( 1992 , 134 ) put it , " The child climbs to grammar on a

semantic ladder and then kicks the ladder away ."

A different approach is the theory of " semantic bootstrapping "

( Grimshaw 1981 ; Pinker 1984 ) , which posits that children possess a lan -

guage acquisition device ( LAD ) that leads them to categorize names for

things as nouns , names for actions as verbs , and so on . Through applica -

tion of these mappings , along with innate knowledge of structural uni -

versals of grammar , children parse some of the utterances they hear and

thereby learn some language - particular rules of grammar . After they have

acquired such rules , they can use them to categorize new words that do

not fall under the syntax - semantics mappings , such as nouns that do not

name things and verbs that do not name actions .

I will begin by defending the premise that children learn the syntactic

categories that words belong to through use of semantic information . But

I ' ll also argue that neither of the two approaches summarized above is

adequate . I will then discuss an alternative that differs from previous

proposals in certain significant ways , explore implications of this alterna -

tive for the study of word learning , and conclude with some more general
remarks .

12 . 1 Why Semantics ?

One alternative to the view that children use semantics to solve the boot -

strapping problem is that they group words and phrases together accord -

ing to their distributional properties , such as their absolute location in the

sentence , what inflections they appear with , and what words they precede

and follow . Under this view , a child comes to understand that " dog " and

" door " belong to the same linguistic class because they can each follow

" the , " can precede the suffix " - s" and the word " is , " and so on . In the

strongest version of this account , semantic properties of the words and

phrases are irrelevant ( e .g . , Maratsos and Chalkley 1981 ; see also Harris
1954 ) .

This proposal is partially motivated by the widely perceived failure to

find semantic definitions for syntactic categories . For instance , there are

count nouns that do not name physical things ( e .g . , " joke " ) and verbs
Solving the Bootstrapping Problem 287

that do not name actions (e.g., " know " ) . Although child language does
have a semantically transparent flavor , with a high proportion of nouns
describing things , and verbs describing actions (Macnamara 1982), even
1- and 2-year-olds know words that do not fit this pattern (Nelson ,
Hampson , and Shaw 1993; Pinker 1984) . The possibility that children
might use prosodic information to carve up sentences into grammatically
relevant units (Gleitman and Wanner 1982; Hirsh -Pasek et al . 1987) as
well as phonological information to distinguish function morphemes
from members of open-class categories (Gerken , Landau , and Remez
1989) introduces further sources of evidence that a nonsemantic learning
proced ure could use.
Nevertheless , it has long been known that an unconstrained distribu -
tional learning model has serious limitations as a solution to the boot -
strapping problem (Gordon 1982, 1985; Pinker 1979, 1984, 1987). There
are an enormous number of surface correlations that children could en -

code. A child exposed to " Look at the dog " could note that " look " is the
first word of the sentence , that it appears to the immediate left of " at ,"
two words to the left of " the," three words to the left of " dog," and so on
for each of the other words in the sentence . But most of these facts are

irrelevant - it does not matter from the standpoint of grammar that

" look " appears in the same sentence as " the." Focusing on a specific lin -
guistic domain , Gordon ( 1982) has calculated that a child who tried to
distinguish count nouns from mass nouns through purely distributional
differences would have to sift through about 8 billion possible contexts in
order to converge on the distinction . Yet children command count / mass
syntax by the age of 2 1/ 2 (Gordon 1982, 1988; Soja 1992).
Recent theories have responded to these concerns by positing that
children are constrained in the sorts of correlations they look for . Under
one proposal , they consider only immediate adjacency relations ( Finch
and Chater 1993); under another , they focus only on the locations of
words relative to a small set of highly frequent function morphemes ,
such as " the" and " was" (Gerken and McIntosh 1993). Connectionist
implementations of these more constrained distributional theories (e.g.,
Redington , Chater , and Finch 1994) have done surprisingly well when
tested on their capacity to syntactically categorize caregiver speech from
the CHILDES corpora (MacWhinney 1991), although the full capacities
of these models are as yet unclear . 1
There are problems , however , with the claim that this is how children
actually solve the bootstrapping problem . For one thing , the proposal
288 Bloom

that children categorizewords through the exerciseof general inductive

capacitiesentails that there should not exist developmentaldissociations
betweensyntax acquisition and inductive learning in other domains. But
such dissociationsexist. One example is the population of children and
adults suffering form specific language impairment (SLI ) studied by
Gopnik (this volume). They are normal with regard to their nonlinguistic
learning abilities, but suffer form a severedeficit in grammar and mor-
phology. For instance, they have great difficulty with the count/massdis-
tinction (producing errors such as " I love bicycle" and " I play musics" ),
number agreement (" I find a cops" ), and noun-verb agreement (" The
ambulancearrive" ). A similar phenomenonshowsup with critical period
effectsin first and secondlanguage learning (Newport 1990); adults are
better learnersthan children in virtually all domains other than language.
These casesare problematic for the distributional account. If the count/
mass distinction is acquired, for instance, by simply noting correlations
betweendeterminersand nouns, it is a puzzle that SLI children and non-
language-impaired 30-year-olds are so bad at this task, while succeeding
at other tasks that clearly do involve the ability to attend to and learn
correlations.
Related to this, distributional mechanismshave no a priori commit-
ments regarding the number and type of categoriesthey can acquire. As a
result, they are necessarilyincompatible with the proposal that there
exists innate knowledge particular to specific linguistic categories. Even
if such a mechanismcould succeedat categorizing the input the child is
presentedwith , all it is going to produce is a seriesof distinct groups of
words- one with " dog" and " door" in it , another with " kick " and
" know," and so on. It has no way of telling which of theseclassescorre-
sponds to which syntactic categories and so it does not fully solve the
bootstrapping problem; it would tell you which words go together, but it
would not tell you how to find the nouns (seealso Pinker 1984).
The reason why this matters is that there are linguistic universals that
apply to specific syntactic categories. For instance, if a language has
Subject-Object-Verb order, it is likely to have postpositions; if it has
Subject-Verb-Object order, it is likely to have prepositions (Greenberg
1966). At a more abstractlevel, conditions on grammar suchas constraints
on question formation or restrictions on coreferenceappear to hold uni-
versally, subjectto constrainedparametric variation (e.g., Chomsky 1986;
Manzini and Wexler 1987). Most relevant to the point of this chapter, it
has long been known that words with the semantic role of referring to
Solving the Bootstrapping Problem 289

objects universally fall into the syntactic class of nominals ( Bloomfield

1933; Macnamara 1986) .
To the extent that these generalizations exist because of how the human
mind is structured , they pose a problem for a purely distributional theory
of children 's learning . A similar problem arises in cases in which children
are exposed to impoverished input (input in which distributional evidence
for linguistic categories is diminished or absent) but nevertheless come to
possesssyntactic rules and categories. Such instances show up in creolization
(Bickerton 1981), first language acquisition of American Sign Language
based on input from nonproficient users (Singleton and Newport 1993), and
" home -sign," in which children create a signed communication system
with no linguistic model at all (Goldin -Meadow and Mylander 1990) . In
all of these instances- and most likely in normal acquisition as well - the
child is imposing structure on the input , not merely abstracting it away .
Finally , the premise behind distributional theories is that syntactic
categorization is fundamentally holistic : the linguistic status of a word is
determined relative to how all other words behave ; there is no sense in
which it can be classified in isolation , through the use of semantic or any
other information . But there is considerable evidence that very young

children possess a rich understanding of syntax -semantics mappings . For

instance , in a classic acquisition experiment ( Katz , Baker , and Macna -
mara 1974), 17-month -olds were taught a new word as either a count
noun (" This is a wug " ) or a lexical NP (" This is wug " ). When the new
word was used to describe a doll , children were sensitive to this difference ;

they tended to interpret the count noun as naming the kind of doll , but
treated the NP as a proper name for that specific doll . Two - and 3-year-
aIds are also sensitive to contrasts between subclasses of nouns . In par -

ticular , they can focus on the difference between count nouns and mass
nouns (e.g., between " a wug " and " some wug " ) when inferring properties
of word meaning . They will tend to construe a novel count noun as
referring to a kind of individual and a novel mass noun as referring to
a kind of nonindividuated entity ( Bloom 1994; Brown 1957; Soja 1992).
Finally , certain early categorization errors , such as treating object mass
nouns like " furniture " and " toast " as count nouns ( Bloom 1994), are best
explained in terms of childrens using semantics as a clue to category
membership .
Other linguistic domains yield similar findings . For instance , N aigles
( 1990) found that 24-month -olds attend to the distinction between NP -V
and NP -V -NP contexts when inferring aspects of the meanings of verbs.
290 Bloom

Although the precise nature of children 's competence is a matter of con -

siderable debate ( Fisher et al . 1994; Pinker 1994), there is little doubt that
they can use mappings from syntactic categories to semantic structure in
the course of word learning and syntax acquisition , which militates against
any theory that children are initially restricted to attending to nonsemantic
properties of sentences.

12 .2 Two Semantic Theories

Under the theory presented by Macnamara (1982), children begin with

rules that map over semantic categories like " object word ." Since such
rules are inadequate to capture the patterns of adult language , children
come to expand such categories and make them more abstract , gradually
transforming them so that they are truly syntactic , such as " noun ." In this
way , the child moves from a semantic understanding of language to a
syntactic one (for similar approaches , see Braine 1988, 1992; Schlesinger
1988) .
One problem with this proposal was raised above . As Macnamara
explicitly states, his theory is incompatible with the notion of innate
knowledge about specific grammatical categories . Under his proposal ,
categories such as " noun " emerge only through a domain -general proce-
dure of abstraction . The child would be just as happy to stick with " object
word " since this is the natural category she forms - the shift to " noun "
occurs only because the input is incompatible with this initial interpreta -
tion . Put differently , " noun " is just the name for the category that used to
be " object word " ; there is nothing else to be said about it . As a result , any
evidence for innate syntactic knowledge about nouns poses a problem for
Macnama14a's theory .
Another problem is empirical . Some studies suggest that there is no
stage of language development in which the child 's rules map onto
semantic categories such as " object word ." For instance , an analysis
of children ' s very first word combinations shows that - contrary to Mac -
namara 's proposal - utterances like " big dog " and " nice hat " are not
generated by a rule such as " modifier + object word ." This is because
children 's early utterances also include expressions in which modifiers
are followed by substance names (" hot water ," " good milk ," etc.), names
for actions (" little fall ," " big hug " ), and names for nonmaterial entities
(" big hug," " big noise" ) . From the very start , their knowledge seems
best captured in terms of the syntactic categories " adjective " and " noun "
Solving
theBootstrapping
Problem 291

(Bloom 199Gb ) . More generally, there appearsto be no stageof language

development in which children's rules utilize different categories than
adults' rules do (Pinker 1984). It looks like syntax all the way down.
This finding favors the proposal that both semanticcategoriesand syn-
tactic categoriesare presentat the very start. This brings us to " semantic
bootstrapping." Grimshaw (1981) and Pinker (1984) have proposed that
the form-meaning correspondencesthat children exploit in the boots-
trapping procedure are not part of knowledge of languageitself, but are
instead the result of special mechanismsof language acquisition. Thus,
Grimshaw (1981, 171) posits as a principle of Universal Grammar that
" form and function are independent." Given this, whatever correspond-
encesexist, do so because" the LAD . . . gives priority to a grammar with
a one-to-one correspondencebetween form and function." Similarly,
Pinker (1984, 39) states, " Although grammatical entities- do not have
semantic definitions in adult grammars, it is possible that such entities
refer to identifiable semantic classesin parent-child discourse" ; and he
notes later that these correspondencesare only sure to hold in " basic
sentences ," not in passives, contextually dependentsentences , and so on
(but seePinker 1989for a somewhatdifferent proposal) .
One problem with this view is that by stipulating that the syntax-
semanticsmappings are part of a separateLAD , independent from the
rest of language, it cannot explain why these particular ones occur and
not others. There is no principled explanation for why the mapping is
from object namesto nouns instead of (for instance) from object namesto
modals. But thesemappings are unlikely to be arbitrary from the stand-
point of how languageworks. Would it be a possiblelanguageif the LAD
instead led children to categorize " dog" as a modal and " kick" as a
quantifier? Probably not; other universal properties of language, such as
structural constraints on coreferenceor conditions on quantifier inter-
pretation, could not naturally occur. So long as the syntax-semantics
mappings exploited in language acquisition are viewed as distinct from
the rest of linguistic knowledge, systematicrelationshipsbetweenthe role
of semanticsin language acquisition and the role of semanticsas part of
linguistic knowledgeitself are left unexplained.
This relatesto a further problem, which is that the proposedmappings
associatesyntactic categorieswith referenceto concrete entities; in the
case of nouns, the proposed semantic correlate is " name of person or
thing" (Pinker 1984, 41). There are advantagesto this. For one thing, it
provides the child with a tractable entry (via perception) into the linguistic
292 Bloom

system. For another , generalizations such as " Names for kinds of discrete
physical objects are nouns " are likely candidates for descriptive universals
of language ( Bloomfield 1933; Macnamara 1986). However , this theory
leaves certain semantic aspects of syntactic categories unexplained . It is
hardly an accident , for instance, that words such as " day ," " nightmare ,"
and " conference " are nouns and not prepositions or modals . Further -
more , even 3 -year - olds appear to be sensitive to the semantic correlates

of nominal syntax in nonmaterial domains , such as when learning names

of sounds or events ( Bloom 1994). At least within the nominal system,
then , there exist syntax -semantics mappings that appear to go beyond the
domain of names for people or things , and it would thus be preferable to
view generalizations like " Names for people or things are nouns " as by -
products of more abstract semantic properties of grammatical categories,
not as mental principles in their own right .
A final issue has to do with how semantic bootstrapping applies . Under
Pinker 's ( 1984) theory , children learn the meanings of some words prior
to learning their syntactic category . They learn the meanings of " the" and
" dog ," for instance, and use the mappings that are part of the boot -
strapping procedure to categorize them as a detenniner and a noun :
detenniner because " the " refers to definiteness in discourse ; noun because
" dog " is the name of a thing . Upon hearing a string of words like " the
dog ," the child parses it as detenniner -noun , and then uses principles of
X -bar theory to flesh out the phrase structure , so that both words are
dominated by an NP and " dog " is immediately dominated by an N ' . A
further mechanism analyzes the resulting tree and posits a phrase struc-
ture rule of the fonn NP --t Det - N ' , Once this rule is in the child ' s

grammar , he can use it to determine the syntactic category of words that

do not fall under the initial bootstrapping procedure . If he hears " the
joke ," he can use this rule to infer that "joke " is a noun , even though it
is not an object name . ( There are syntax -semantics mappings that apply
to phrases under Pinker 's theory ; these are used to annotate NPs with
different grammatical functions , such as Subject or Topic . But they apply
only once the NPs have been categorized through the procedure sketched
above.)
Pinker is explicit that semantic bootstrapping is not a theory of the
acquisition of word meanings ; it assumes that children have learned the
meaning of some words prior to syntactic development , but makes no
commitments as to how . Nevertheless , the specific procedure discussed
above does make one nontrivial assumption about word learning : it
Solving the Bootstrapping Problem 293

assumesthat children can learn the meaningsof nounsprior to categorizing

phrasesas NPs, that they can learn verbs prior to categorizing VPs, and
so on. The sameassumptionis also made by many theoriesthat posit that
children move from semantic categoriesto syntactic categories, such as
the theoriesproposed by Braine (1992) and Macnamara (1982).
But the claim that children learn lexical categoriesbefore phrasal ones
may be wrong. After all, nouns and verbs do not themselvesrefer to
objects or actions, they are not subjects or predicates, and they do not
have thematic roles. It is phrasal categoriesthat are initially " visible" to
the child through a semanticanalysis of sentencemeaning. Furthermore,
phrases- and not words- might also be identifiable on the basis of a
prosodic parse (Gleitman and Wanner 1982) and frequently appear as
isolated units (Peters 1983).
Consider some specificexamples. How does a child learn the meaning
of the word " dog" ? Presumably part of the solution rests on the child's
somehow figuring out that adults use the word " dog" with the intent
to refer to the category (or kind) of dogs. But the word " dog" does not
itself describe or refer to the category (or kind) of dogs. An adult who
wishesto describea dog is likely to point and say somethinglike, " Look
at the dog" or " That's a dog" or " There's that big dog again." The lin-
guistic unit that refers is not " dog" ; it is the NP " the dog" or " a dog" or
" that big dog." This suggeststhat the only way children can gain access
to the meaning of " dog" is through its contribution to the semanticsof
the NP. They must determinethat a phraselike " the dog" refersto a dog,
and they must figure out which of the words in this phrase refers to the
kind. Determiners are a similar case. Words like " the" and " a" appear
only in the context of phrasesand gain their meaning through their con-
tribution to the semanticsof the NP . Such words cannot be acquired in
isolation, suggestingagain that children learn them after analyzing utter-
anceslike " the dog" as NPs, not before.2 Note also that for lexical NPs-
like " Fred" and " he" - the single word is itself a referring element, and
there is no noun that the child must learn in order to gain accessto
the NP.
A related concern is that Jelinek (in press) and Kinkade (1983) argue
that Straits Salish lacks a noun/verb distinction; there are no lexical ele-
ments that are uniquely associatedwith the maximal projections NP and
VP. Put differently, those roots that would correspond to English nouns
are syntactically and morphologically identical to those corresponding
to English verbs, and each can appear within both NPs and VPs. Salish
 294 Bloom

doesmark a distinction betweennominals and predicatesbut only at the

phrasal level- it has NPs and VPs, just not nouns and verbs. Quechua
poses a related puzzle, becauseverb roots cannot stand as independent
morphemes and, more generally, there are many contexts in which the
syntactic category of a root can only be determined through its semantic
and morphological contribution to the larger syntactic constituent that it
belongsto (Courtney 1994; Lefebver and Muysken 1988). Thesephenom-
ena are difficult to explain in a theory that assumesa lexical category- to-
phrasal category analysisin the course of syntactic development.
Theseconsiderationsmotivate a somewhatdifferent conception of the
role of semanticsin solving the bootstrapping problem, one that I sketch
below.

12.3 A MappingTheory

The scope of the alternative is modest . It will focus mostly on the acqui -

sition of nominals , though it will also touch on the acquisition of deter -

miners and verbs . It has two aspects : a representational claim , concerning

the nature of the syntax - semantics mappings that are used to learn the

syntactic categories that words belong to , and an acquisition claim , con -

cerning the nature of the learning process .

12 . 3 . 1 The Representational Claim

In Pinker ' s and Grimshaw ' s theory , the mappings children use are part

of a separate LAD , independent from knowledge of language . The alter -

native proposed here is that the mappings children exploit in language

learning are instead derived from syntax - semantics mappings that exist as

part of knowledge of language itself . This alternative is based to a large

extent on proposals made by Jackendoff ( 1990 , 1996 ) and Macnamara

( 1986 ) , though the specifics differ in certain regards .

In a language such as English , there is a lexical / phrasal contrast within

the nominal system . Nouns are words that refer to kinds , whereas NPs are

words or phrases that can refer to individuals . Thus , " dog " itself does not

refer to dogs , but it can be quantified and modified to form an NP , which

can refer . For instance , " that dog " picks out a particular individual that is

a dog , " those big dogs " picks out a set of dogs that are big , and so on .

Pronouns and proper names are lexical NPs because , unlike words such as

" dog , " they inherently pick out individuals , not kinds .
Solving the Bootstrapping Problem 295

The notion that NPs, and not nouns, refer to individuals is familiar
from work within formal semantics, where nouns are sometimesviewed
as predicates, which must combine with determinersto establishreference
(e.g., Barwise and Cooper 1981). More generally, the notion is that nouns
are semantically incomplete (or " unsaturated" ; seeHigginbotham 1983);
only NPs are semanticallycomplete. As such, only NPs can participate in
certain forms of semanticinteraction, such as having thematic roles, par-
ticipating in coreferencerelations, and being able to refer to entities in the
world .
Although the lexical/phrasal contrast within nominals may not hold
across all languages, one could posit as a universal of language the
following :
NPs refer to individuals .

This is a unidirectional mapping ; it does not preclude the possibility that

some NPs might serve other semantic roles. This qualification is essential:
some NPs do not refer at all . An expletive such as " it " as in " It is raining "
exists only to satisfy a grammatical condition in English stating that
tensed clauses must have subjects (Chomsky 1981; Rothstein 1983). There
are also NPs that have semantic substance , but do not refer to individuals .
An NP such as " a dog" has a different semantic role in a sentence such
as " Fred is not a dog " than in a sentence such as " A dog walked into
the room ." Only the second sentence actually refers to a dog ; in the first ,
the NP serves as a predicate and has much the same semantic role as the
Adjective Phrase " happy " in " Fred is not happy " (see Williams 1983 for
discussion). Finally , some NPs , such as " some water " or " much sand,"
refer to portions of stuff , not to individuals . In general, NPs headed by
mass nouns do not refer to individuals , since mass nouns refer to kinds
of nonindividuated entities , not kinds of individuals (Jackendoff 1990;
Macnamara 1986) .3
The contrast between " object " and " individual " is crucial here. This
shift to " individual " follows naturally if one assumes that the mapping
children apply is part of more general properties of language and cogni -
tion , and not a separate LAD . NPs in natural language are not limited to
referring to objects . They can refer to parts (" a hand " ), portions of sub-
stance (" a puddle " ), bounded events (" a race" ), abstract notions like
jokes and ideas, collections of objects like flocks and families , and so on .
Nominals behave identically from the standpoint of linguistic semantics
regardless of what sorts of individuals or kinds of individuals they refer
296 Bloom

to. For instance, the numerical contrast between" two cookies" and " five
cookies" is identical to the contrast between" two jokes" and " five jokes,"
and the same properties of binding theory apply regardlessof whether
the referring expressionsrefer to material (" the cookie" ) or nonmaterial
(" the joke" ) entities. In general, the semanticstructuresof languagedon't
care whether or not the individuals denoted by nominals are physical
objects.
In light of this, it might be most parsimonious to view the child's ten-
dency to construe phrasesthat describeobjectsas NPs as a specialcaseof
the mapping that statesthat NPs can refer to individuals. In particular,
children might possessthe following bias concerning the relationship
betweenlinguistic semanticsand nonlinguistic cognition:
Physical whole objects are canonical individuals.
There is considerablesupport for the claim that children are biased to
construe whole objects as individuals, support that is quite independent
from considerationsof syntactic development. When children hear a word
(especially a count noun) used to describe a novel object, they show a
strong bias to take the word as referring to the kind of object, not to a
part, or property, or anything else (seeMarkman 1989). Related to this,
Shipley and Shepperson(1990) askedpreschoolersto count different sorts
of entities. For instance, they showedchildren a picture of five forks, one
of them broken into two pieces, and asked, " Can you count the forks?"
Most of the children said, " Six," presumably becausethere were six dis-
cretephysicalobjectsin front of them- although only five forks. Shipleyand
Sheppersoncall this the " discretephysical object" bias, which is equivalent
to the above claim that, when it comesto quantifying over individuals,
objectsare highly salient. Note also that a considerableamount of knowl-
edgeabout objectsappearsto be innate in humans (Spelkeet al. 1992).
Although objects are salient individuals, even infants and young
children can quantify over individuals that are not objects. In one study,
infants were exposedto either two sounds or three soundsand then pre-
sentedwith two pictures: one with two objectsand one with three objects.
Even 6-month-olds tended to look longer at the picture that showed the
same number of objects as there were sounds, providing some evidence
that infants possessnotions of " two individuals" and " three individuals,"
where " individual " encompasses both discretesoundsand discreteobjects
(Starkey, Spelke, and Gelman 1990). When there are no salient objects
present, 2-year-olds have no problem counting sounds or actions; this
Solving the Bootstrapping Problem 297

capacity emerges at the same time that they are able to count objects

( Wynn 1990 ) .

Returning to language , even 20 - month - olds know many nouns that

describe nonobject kinds , such as parts ( " hand " ) , events ( " race " ) , and

periods of time ( " hour " ) ( Nelson , Hampson , and Shaw 1993 ) . Finally ,

children can use syntactic cues to learn count nouns that fOffi1 NPs

referring to discrete sounds ( Bloom 1994 ) , bounded portions of stuff like

puddles ( Soja 1992 ) , and collections of objects ( Bloom and Kelemen

1995 ) .

These considerations support the view that the relationship between

NPs and objects is not direct . Instead , even young children possess an

abstract semantic notion of " individual " that maps onto NPs . Objects are

salient individuals , but other entities can be individuals as well . It is likely ,

for instance , that children can categorize a word or phrase as an NP by

hearing it used to refer to a collection , or to a bounded substance , or to a

discrete sound . One factor that determines what can count as a possible

individual is boundedness , either in space or in time , which is why NPs

can refer to entities like puddles ( Soja 1992 ) and discrete sounds ( Bloom

1994 ) . A more fundamental consideration might have to do with having a

single causal role - if adults see a group of objects moving together and

acting as a single agent ( e . g . , a flock or an affi1Y ) , they will construe

the objects as a single individual , but if the objects are stationary , the

individual interpretation is much less salient ( Bloom 1996 ) . For differ -

ent perspectives on the category of " individual " and constraints on pos -

sible kinds of individuals ( i . e . , possible count nouns ) , see Bloom 1996 ,

Langacker 1987 , and Macnamara 1986 .

12 . 3 . 2 The Acquisition Mechanism

The discussion above has focused on a putative semantic correlate of

NPs - individuals - and its relationship to the notion of discrete physical

object . But children must somehow use this knowledge to categorize the

NPs they are exposed to . Further , they must categorize other parts of

speech , such as determiners .

As discussed above , the usual proposal is that children first categorize

words as lexical categories and then use grammatical information to infer

further aspects of phrase structure . The problems with this view motivate

an alternative , which is that children start off parsing sentences they hear

into phrases , perhaps with the aid of prosodic information , and use

structural principles of grammar along with further semantic information

298 Bloom

to determine which words belong to which lexical categories . In the case

of nominals , the relevant procedure can be summarized as follows :
Primacy -of- NPs procedure
Categorize a word (or combination of words ) that refers to an individual
as an NP . Recategorize it as a noun (or as a noun and other parts of
speech) only if there is semantic evidence that this initial NP -construal is
mistaken .

This proposal leads to the following acquisition scenarios.

Children can categorize their first pronouns and proper names as lexical
NPs by noting that they refer to discrete physical objects, categorizing
them as referring to individuals (through the relationship between non -
linguistic cognition and this semantic notion ) and thus as NPs (through
the mapping between semantics and grammar ) . Prior to the acquisition
of other parts of speech, this is the only way in which new words can be
encoded as NPs , and it might serve as the foundation for children 's grasp
of syntax . In other words , children 's very first syntactic categories might
be NPs . In support of this , pronouns and proper names are among the
very first words learned by children across a variety of cultures (Gentner
1982) .
The only way to acquire determiners and common nouns is through
decomposition , as these words rarely appear in isolation . Through the
procedure above, children initially categorize phrases such as " the dog "
or " some cups" as NPs and must do further analysis in order to catego-
rize " dog " and " cups" as nouns and " the" and " some" as determiners .
In particular , only after hearing the different words used in distinct
phrases (e.g., " the cat," " a cup " ) will they decompose the phrase into two
components (the determiner " the" and the noun " cup " ), by noting the
different contributions that the two words make toward the meanings of
the phrases in which they appear . (This process would be facilitated if
children are capable of distinguishing closed-class forms like determiners
from open-class forms like nouns on the basis of phonological properties ,
such as weak versus strong stress; Gerken , Landau , and Remez, 1989.)
Note that identifying the parts of a given NP as a determiner and a noun
is not sufficient to decompose it . There must be semantic evidence for
compositionality as well - the meaning of the determiner and the meaning
of the noun must combine in the expected manner to form the meaning of
the referential NP . Otherwise , the phrase is stored in the lexicon as a
whole , as is the case with NP idioms such as " the big cheese" (see below ).
Solving the Bootstrapping Problem 299

As Pinker ( 1984) notes, this decomposition process is similar to the one

that takes place in the course of morphological acquisition . Children ac-
quiring English must learn that the words " cups," " dogs," and " houses"
are not semantic monads , but instead decompose into the aspect of
meaning expressed by the noun and the notion of plurality expressed by
the plural morpheme . The suggestion here is that an analogous procedure
takes place when children hear a phrase like " the dog ." They initially
encode it as a semantically simple NP and store it in their lexicon ; only
later do they decompose it into two semantic units , the determiner " the"
and the noun " dog ." In support of this , morphemes that are semantically
equivalent to words such as " the" and " some" are often affixes in other
languages, just as the plural morpheme is in English .
This proposal seems to make the wrong prediction about children 's
spontaneous speech, however , as it suggests that they should produce
phrasal NPs prior to count nouns - " a cup" before " cup ," for instance .
This does not occur , of course . But the reason for this may be that chil -
dren are capable of extracting the noun from a determiner -noun NP early
in language development , perhaps even prior to the onset of spontaneous
speech. In addition , determiners are not phonologically salient ; sometimes
young children might simply not perceive them (Gleitman and Wanner
1982), leading them to parse a phrasal NP like " a cup " as just " cup " and
then to categorize " cup" as a noun through the semantic decomposition
procedure discussed below . Slightly older children do appear to have
identified certain English words as determiners , distinct from nouns , and
this would allow them to parse the NP in an adultlike way (Gerken ,
Landau , and Remez 1989; Katz , Baker , and Macnamara 1974). Note also
that even if a 2-year-old did store phrases such as " a cup " or " the dog " in
his lexicon as nondecomposed NPs , such phrases would likely be pro -
duced as just " cup " and " dog ," since 2-year-olds have problems pro -
ducing determiners in their spontaneous speech, because of articulatory
limitations (e.g., Gerken , Landau , and Remez 1989).
To see how semantic evidence might lead to decomposition when no
syntactic motivation is present, imagine being shown an inanimate entity ,
like a cup , and being told , " This is wug ." From the standpoint of gram -
mar alone , " wug " could be a lexical NP that refers to that specific indi -
vidual (i .e., a proper name) or it could be a noun that refers to that kind of
individual . Even young children assume that such words are nouns ( Katz ,
Baker , and Macnamara 1974), possibly because they are unwilling to
construe things like cups as having their own names. This semantic
300 Bloom

consideration is sufficient to motivate a decomposition of the single word

into a noun preceded by a null determiner , even without any syntactic
motivation for this analysis .
Idioms provide the opposite situation . Although an NP idiom like " the
big cheese" appears to be syntactically decomposable (into determiner -
adjective -noun ), it is not semantically decomposable and thus must be
acquired by the child as a lexical NP . Even once the child knows the
meanings of the constituent words (" the," " big ," and " cheese" ), the phrase
will only be syntactically decomposed to the extent that it is semantically
transparen t .
This claim is supported by psycholinguistic evidence. Cutler ( 1982)
discusses the role of certain factors in determining how frozen different
idioms are (i .e., how accessible they are to syntactic movement , such as
passivization ) . She noted a strong trend that she sums up as " the colder
the older ." In other words , older idioms are more resistant to syntactic
transformation than newer ones. One explanation is that older idioms
may be harder to make sense of (especially to children ) and thus may
be acquired as unanalyzed elements; hence, they would be " colder ."
Supporting this , there is experimental evidence suggesting that degree of
semantic " analyzability " is correlated with how syntactically decom-
posable a given idiom is (Gibbs and Nayak 1989).
Expletives belong to the class of NPs that do not refer to individuals .
There are two potential explanations of how such words could be cate-
gorized , both of which are consistent with the theory proposed above.
The first proposal is that the acquisition of expletives is simply inde-
pendent of the proposed mapping . Since expletives do not refer , they will
not be categorized through the mapping from individuals to NPs ; instead ,
they will be learned through syntactic inference . In particular , once chil -
dren have acquired the meaning of a verb like " raining " - that it has
no semantic arguments - and given that they know overt subjects are
obligatory in English (knowledge present early in development ; see
Bloom 1993; Valian 1991), this will force them to categorize " it " as an
expletive NP on the basis of a sentence such as " It is raining ."
A second proposal makes a stronger assumption about the syntax -
semantics mappings . It is possible that not any lexical NP can be an
expletive , but only those lexical NPs that are initially acquired through the
syntax -semantics mapping . To put it differently , there may be a sufficiently
strong correspondence between the syntactic fact of being an NP and the
semantic fact of referring to an individual , such that although lexical NPs
Solving the Bootstrapping Problem 301

can be co-opted to serve other semantic roles (as in the case of expletives),
they can only enter the lexicon via the syntax -semantics mapping . If this is
so, children would not be capable of learning the expletive " it " unless
they had already acquired " it " as a referential prnoun .
This leads to two predictions : ( 1) children should use an NP as an
expletive only after they have acquired its referential meaning , and (2) no
language should have NPs that are solely expletive and have no referen-
tial double life , because children would never be able to acquire such
words . Nishigauchi and Roeper ( 1987), who present a similar theory of
the acquisition of expletives , argue that both predictions hold , but admit -
tedly the evidence is relatively scanty .
What about verbs ? The discussion above has focused on NPs and parts
of NPs , but children must also acquire other parts of speech. No doubt
syntax -semantics mappings can apply in these domains as well . Just as
there is a mapping between individuals and NPs , so there may be similar
mappings between events and VPs, propositions and sentences, and so on .
The semantic bootstrapping theory also posits such mappings , although ,
as with nominals , these are presumed to be independent of more general
linguistic knowledge .
It is also likely that once children have categorized some words or
phrases as NPs , principles of grammar could lead them to categorize
words belonging to other syntactic categories, even in the absence of any
prior knowledge of what these words mean . Consider a child who knows
(through the procedure discussed above) that " Mary " and " Joe" are NPs
that refer to individuals , and then hears these NPs used with the novel
word " kissed" (e.g., " Mary kissed Joe" ). Assuming that children can de-
teffi1ine whether a string of words is a sentence as opposed to a fragment ,
and that they expect semtences to express propositions , the utterance
would be semantically encoded as follows , using a notation based on
Jackendoff 's ( 1990) :
PR 0 POSITION

([IND IVID U AL ] ??? [IND IVID U AL ])

Mary kissed Joe

One likely candidate for a universal of language is some constraint to

the effect that all NPs with semantic content must get their semantic role
from some predicate , as with the Theta Criterion of Chomsky ( 1981),
though the proper formulation of such a constraint is a matter of some
debate (see Jackendoff 1987) . If children possessthis constraint , then they
 302 Bloom

can categorize " kissed " as a predicate ( since there must be something
licensing these NPs and this is the only candidate ) and give it the follow -
ing representation :

EVENT

" KISSED " : ( [INDIVIDUAL ] , [INDIVIDUAL ])

Finally , if children possess some mapping from EVENTs to verbs , this

could allow children who have only acquired NPs to use syntax - semantics
mappings to categorize novel words as verbs .

12 .4 Mechanisms of Word Learning

In the discussion of how children learn pronouns and proper names , I

assumed that children have the capacity to categorize a word or string of
words as an NP by noting that it is used to refer to an individual ( in many
cases , a whole object ) - and that they can do this even if they have no
prior knowledge of the syntax of the particular language they are exposed
to . Gleitman ( 1990 ) , however , discusses certain issues that arise from the
assumption that children can determine the reference of a word without

already knowing its linguistic category , and these are worth considering
here .

At first blush , it is not obvious where the problems lie . All children
must do is hear a word or phrase ( e .g ., " Freddie " ) , follow an adult ' s gaze
or point , and note that the adult is intending to refer to an object (e .g .,
to Freddie ) . Since Freddie is an object , the child will construe it as an
individual , and since NPs refer to individuals , " Freddie " must be an NP .
Along these lines , note that even 12 - and 13 -month - olds are drawn to
focus on the individual when exposed to a new word that describes an
object - and will do so regardless of whether the word is a noun or an

adjective , suggesting that overt syntactic cues are not essential at this early
stage of word learning , at least not for the acquisition of names for object
kinds ( Waxman 1994 ) .
However , there are situations in which children cannot be assured
of exposure to ostensive labeling . For instance , Kaluli mothers do not
engage in labeling behavior with their children . Instead , they provide their
children with extensive training in conversational interaction , often mod -
eling sentences for them ( Schieffelin 1985 ; see also Heath 1983 ) . Despite
this difference in input , Kaluli children acquire words at much the same
rate as children acquiring languages such as English . To see the problem
Solving the Bootstrapping Problem 303

this raises, imagine such a child seeinga bird in the sky and hearing, for
example, " Zav goop wicket mep." By hypothesis, the child knows no
words at this point, no inflections, and nothing about the syntax. (And
in fact, it is optimistic even to assumethat she is capable of parsing
the utterance into distinct words and phrases; if not, she will hear
" Zavgoopwicketmep." ) Which word or string of words (if any) does the
child construe as referring to the bird? How do Kaluli children learn any
word meanings?
There are two possibletypes of solutions. The first is that children per-
form a constrained distributional analysis; they store sentence -situation
pairs in memory and then look for correlations between specific words
or phrases and discrete aspectsof meaning. For instance, if the same
child hears" Blub mendle wicket mep" to refer to another bird, shemight
assumethat " wicket mep" (which was present in both sentences ) is used
to refer to birds and is thus an NP . The analysis might be simplified if
children only store strings of stressedphrases(instead of whole sentences )
and representationsof relevant individuals (instead of entire situations).
But such an analysis might conceivably lead to problems. Imagine a
child who seesone bird, hears the equivalent of " Isn't that pretty?" , sees
another bird, and hears the equivalent of " That's also pretty." Such a
child, using the procedure above, would infer that " pretty" is an NP re-
ferring to the bird. Such errors hardly ever occur with real children.
Nevertheless, if children only hear NPs in the context of sentencesthat
include other phrases, some such proceduremust be applying.
The alternative is to deny the premise above. Perhapsall children are
exposedto somewords in isolation, and thesecan be categorizedas NPs.
One candidate class of words is proper names, as there is evidencethat
these are privileged in lexical acquisition. They appear among the very
first words of children learning a range of different languages(Gentner
1982)- including Kaluli children. Further, proper namesmight be the one
classof words that all children are guaranteedof having been exposedto
in isolation, or at least in some special stressedcontext. Regardlessof a
society's beliefsconcerninghow children learn namesfor kinds of entities,
it is unlikely that any culture assumeschildren are born knowing the
proper namesof the peoplesurrounding them. Thesemust be taught, even
to other adults.
This predicts that even in cultures in which adults do not standardly
label objects for children, proper nameswill be taught to children. Look -
ing at Kaluli children, we find this to be the case. Schieffelin (1985, 534)
notes,
304 Bloom

There are no labeling games to facilitate or encourage the learning of object

names . This is primarily due to the linguistic ideology of the culture . It is only

in families who are acquiring literacy that one sees any attention paid to saying

the names of objects , and this activity is initiated by the child when the mother

is looking at books . When extended by the child to other contexts , the mother ' s

response is disinterest .

In contrast , because of the cultural importance placed on learning the proper

names and kinterms of the individuals with whom they interact , Kaluli children

are consistently encouraged to master a large number of proper names , kinterms ,

and other relationship terms . . .

Perhaps every cu ~ture has some class of nominals that are special with

regard to interaction with children . In Western societies , this is a very

broad class , including virtually all object and substance names ; within the

Kaluli society , it is much more narrow , restricted to proper names and

relationship terms . Another distinct candidate for privileged nominals

is the class of deictic pronouns , like " this " and " that ." These are also

universal , show up early in child language , and serve to draw children ' s

attention to objects and other individuals in the environment .

In sum , it might be that all cultures use some NPs that refer to

individuals in isolation , allowing children to botstrap into the language

system . If not , then children must be capable of somehow learning the

meanings of words that are embedded in sentences , by extracting the

referential NPs from such sentences through a constrained distributional

analysis . Thus , an adequate theory of lexical acquisition must assume

either strong extrinsic constraints ( all cultures use some nominals in iso -

lation ) or a powerful learning mechanism ( one that can learn words that

are not presented in isolation ) . Either possibility would be of considerable

in teres t .

12 .5 Conclusion

Higginbotham ( 1989 , 119 ) has described the question of the autonomy of

syntax within linguistic theory as having " perished or been transformed "

in that the initial considerations in favor of autonomy " seem to have been

not refuted or vindicated but merely superseded by later developments . "

The same sort of transfoffi1ation has not yet occurred in the field of

language development , where it is often assumed that " semantic " theories

of acquisition are an alternative to " syntactic " theories , and experimental

data are interpreted as favoring one sort of theory over the other . Child -

ren ' s early understanding of semantic structure is presented as evidence

Solving the Bootstrapping Problem 305

against innate syntactic knowledge ( Braine 1992 ; Lieberman 1991 ) ;

conversely , early command of syntax and morphology is interpreted as

showing that semantics plays little or no role in grammatical development

( Levy 1988 ) .

Some recent work rejects this dichotomy and attempts to develop a

theory of language acquisition consistent with the notion of a distinct

level of syntactic representation that relates systematically to nonlinguistic

cognition ( e . g . , Gleitman 1990 ; Pinker 1989 ) . The proposal here falls

into this research program , as it posits that innate mappings from syntax

to semantics - as well as properties of the interface between semantic

categories like " individual " and cognitive notions like " discrete physical

object " - are used by children to determine the syntactic categories that

new words and phrases belong to . These mappings are presumed to be

present prior to the onset of language acquisition , in contrast to the view

that syntax is in some sense abstracted or derived from prior semantic

structure , as maintained by Braine ( 1988 ) , Macnamara ( 1982 ) , Schlesinger

( 1988 ) , and others .

The proposal here was originally developed as an alternative to seman -

tic bootstrapping , but it accepts two fundamental insights behind that

proposal : children use word meaning to determine the syntactic catego -

rization of new words , and syntactic categories are available at the outset ,

distinct from nonlinguistic cognition . It differs from semantic boot -

strapping by claiming that children do not need to utilize mappings from

a special language acquisition device and by adopting the assumption that

syntactic development runs from phrase to word , not vice versa . This

provides a plausible explanation of how children solve the bootstrapping

problem , one that can apply across different languages and that is con -

sistent with how children learn the meanings of words .

Notes

1 . One specific problem involves homophony ( Pinker 1979 ) . Many of the words

children learn can appear as both nouns and verbs . Full grammatical knowledge

of English requires knowing that " drink " is a verb when used to describe an action

and a noun when used to describe a substance . To this extent at least , the syntactic

categorization of some words must be deternlined relative to their meanings .

2 . This is related to Pinker ' s ( 1984 ) own proposal , which treats the acquisition of

determiners in the same way as the acquisition of nominal morphology , such as

the plural morpheme . The difference is that Pinker proposes that the segmentation

of " the " and " dog " is the result of initially analyzing " the - dog " as a single noun ,

not , as argued here an NP .

306 Bloom

3 . Mass nouns can be used to form NPs that establish reference to individuals

when they appear with constructions that supply a principle of individuation over
the kind that the mass noun refers to , as in " a grain afsand " or " a glass afwater "
(see Macnamara 1986).

References

Barwise , J., and Cooper , R . ( 1981) . Generalized quantifiers and natural language .
Linguistics and Philosophy , 4, 159- 219 .
Bickerton , D . ( 1981) . Roots o/ language. Ann Arbor , MI : Karoma .
Bloom , P. ( 1990a). Semantic structure and language development. Unpublished
doctoral dissertation , MIT .

Bloom , P. ( 199Gb). Syntactic distinctions in child language . Journal of Child

Language , 17, 343- 355.
Bloom , P. ( 1993) . Grammatical continuity in language development : The case of
subjectless sentences. Linguistic Inquiry , 24, 721- 734.
Bloom , P. ( 1994) . Semantic competence as an explanation for some transitions in
language development . In Y . Levy (Ed .), Other children , other languages: Theo-
retical issues in language development. Hillsdale , NJ : Erlbaum .
Bloom , P. ( 1996) . Possible individuals in language and cognition . Current Direc -
tions in Psychological Science, 5, 90- 94.
Bloom , P., and Kelemen , D . ( 1995). Syntactic cues in the acquisition of collective
nouns . Cognition , 56, 1- 30.
Bloomfield , L . ( 1933). Language . New York : Holt .
Braine , M . D . S. ( 1988) . Modeling the acquisition of linguistic structure . In
Y . Levy , I . M . Schlesinger , and M . D . S. Braine ( Eds.), Categories and processes
in language acquisition theory . Hillsdale , NJ : Erlbaum .
Braine , M . D . S. ( 1992) . What sort of innate structure is needed to " bootstrap "
into syntax ? Cognition , 45, 77- 100.
Brown , R . ( 1957) . Linguistic deternlinism and the part of speech. Journal of
Abnormal and Social Psychology , 55, 1- 5.
Chomsky , N . ( 1981) . Lectures on government and binding . Dordrecht : Foris .
Chomsky , N . ( 1986). Knowledge of language: Its nature , origin , and use. New
York : Praeger .
Courtney , E . ( 1994). Children 's first parses and syntactic rules: A proposal . Un -
published manuscript , University of Arizona .
Cutler , A . ( 1982) . Idioms : The colder the older . Linguistic Inquiry ", 13", 317- 320.
Finch , S. P., and Chater , N . ( 1993). Learning syntactic categories : A statistical
approach . In M . Oaksford and G . D . A . Brown ( Eds.), N eurodynamics and psy -
cholo gy . London : Academic Press .

Fisher , C ., Hall , D . G ., Rakowitz , S., and Gleitman , L . ( 1994) . When it is better

to give than receive: Syntactic and conceptual constraints on vocabulary growth .
Lingua , 92, 333- 375.
Solving the Bootstrapping Problem 307

Fodor , J. ( 1966) . How to learn to talk : Some simple ways . In F . Smith and
G . Miller ( Eds.), The genesis of language. Cambridge , MA : MIT Press.
Gentner , D . ( 1982). Why nouns are learned before verbs : Linguistic relativity
versus natural partitioning . In S. A . Kuczaj ( Ed .), Language development: Vol. II
Language } thought } and culture . Hillsdale , NJ : Erlbaum .
Gerken, L . A ., Landau, B., and Remez, R. E. ( 1989). Function morphemesin
young children 's speech perception and production . Developmental Psychology ,
27 , 204 - 216 .

Gerken , L . A ., and McIntosh , B . J. (1993) . Interplay of function morphemes and

prosody in early language . Developmental Psychology , 29, 448- 457.
Gibbs , R ., and Nayak , N . ( 1989) . Psycholinguistic studies on the syntactic
behavior of idioms . Cognitive Psychology , 23, 100- 138.
Gleitman , L . R ., ( 1990) . The structural sources of word meaning . Language
Acquisition , 1, 3- 55.
Gleitman , L . R ., and Wanner , E . ( 1982). Language acquisition : The state of the
state of the art . In E. Wanner and L . Gleitman ( Eds .), Language acquisition : The
state of the art . Cambridge : Cambridge University Press.
Goldin -Meadow , S., and Mylander , C . (1990). Beyond the input given : The
child ' s role in the acquisition of language . Language , 66, 323- 355.
Gordon , P. ( 1982) . The acquisition of syntactic categories: The case of the count/
mass distinction . Unpublished doctoral dissertation , MIT .
Gordon , P. ( 1985) . Evaluating the semantic categories hypothesis : The case of the
count / mass distinction . Cognition , 20, 209- 242.
Gordon , P. ( 1988). Count / mass category acquisition : Distributional distinctions in
children ' s speech. Journal of Child Language , 15, 109- 128.
Greenberg , J. H . ( 1966) . Some universals of grammar with particular reference to
the order of meaningful elements. In J. H . Greenberg ( Ed .), Universals of lan-
guage. Cambridge , MA : MIT Press.
Grimshaw , J. ( 1981) . Form , function , and the language acquisition device . In
C . L . Baker and J. McCarthy (Eds .), The logical problem of language acquisition .
Cambridge , MA : MIT Press.
Harris , Z . S. ( 1954). Distributional structure . Word, 10, 140- 162.
Heath , S. B . ( 1983) . Ways with words. Cambridge : Cambridge University Press.
Higginbotham , J. ( 1983) . Logical form , binding , and nominals . Linguistic Inquiry ,
14 , 395 - 420 .

Higginbotham , J. ( 1989) . The autonomy of syntax and semantics . In J. L . Garfield

( Ed .), Modularity in kno }vledge representation and natural -language understanding .
Cambridge , MA : MIT Press.
Hirsh -Pasek, K ., Kemler -Nelson , D . G ., Jusczyk , P. K ., Wright , K ., and Druss ,
B . (1987) . Clauses are perceptual units for prelinguistic infants . Cognition , 26,
269 - 286 .
308 Bloom

Jackendoff, R. (1987). The status of thematic relations in linguistic theory.

Linguistic Inquiry, 18, 369- 411.
Jackendoff, R. (1990). Semanticstructures. Cambridge, MA : MIT Press.
Jackendoff , R . ( 1996). The architecture of the language faculty . Cambridge , MA :
MIT Press .

Jelinek , E . (in press) . Quantification in Straits Salish. In E . Bach , E . Jelinek ,

A . Kratzer , and B . Partee ( Eds .), Quantification in natural language. Dordrecht :
Kluwer .

Katz , N ., Baker , E ., and Macnamara , J. ( 1974). What ' s in a name? A study of

how children learn common and proper names. Child Development , 45, 469- 473.
Kinkade , M . D . ( 1983) . Salish evidence against the universality of " noun " and
" verb " . Lingua , 60, 25- 40.
Langacker , R . W . ( 1987) . Nouns and verbs . Language , 63, 53- 94.
Lefebvre , C ., and Muysken , P. ( 1988). Mixed categories: Nominalizations in
Quechua. Dordrecht : Kluwer .
Levy , Y . ( 1988) . On the early learning of formal grammatical systems: Evidence
from studies of the acquisition of gender and countability . Journal of Child
Language , 15, 179- 186.
Lieberman , P. ( 1991). Uniquely human: The evolution of speech, thought, and
selfless behavior . Cambridge , MA : Harvard University Press.
Macnamara , J. ( 1982) . Namesfor things: A study of human learning . Cambridge ,
MA : MIT Press .

Macnamara , J. ( 1986) . A border dispute: The place of logic in psychology .

Cambridge , MA : MIT Press.
MacWmnney, B. (1991). The CHILDES project: Tools for analyzing talk.
Hillsdale , NJ : Erlbaum .

Manzini , R ., and Wexler , K . ( 1987) . Parameters , binding theory , and learnability .

Linguistic Inquiry , 18, 413- 444 .
Maratsos , M . P. and Chalkley , M . ( 1981) . The internal language of children 's
syntax : The ontogenesis and representation of syntactic categories . In K . Nelson
( Ed .), Children 's language, vol. 2 . New York : Gardner Press.
Markman , E . M . ( 1989) . Categorization and naming in children : Problems of
induction . Cambridge , MA : MIT Press.
Naigles , L . ( 1990). Children use syntax to learn verb meanings . Journal of Child
Language , 17, 357- 374.
Nelson , K ., Hampson , J., and Shaw, L . K . ( 1993). Nouns in early lexicons : Evi -
dence, explanations , and extensions . Journal of Child Language } 20} 61- 84.
Newport , E . ( 1990). Maturational constraints on language learning . Cognitive
Science , 14 , 11 - 28 .

Nishigauchi , T ., and Roeper , T . ( 1987). Deductive parameters and the growth of

empty categories , In T . Roeper and E . Williams ( Eds .), Parameter setting and
language-acquisition . Dordrecht : D . Reidel .
Solving the Bootstrapping
Problem 309

Peters , A . M . ( 1983 ) . The units of language acquisition . New York : Cambridge

University Press .

Pinker , S . ( 1979 ) . Formal models of language learning . Cognition , 1 , 217 - 283 .

Pinker , S . ( 1984 ) . Language learnability and language development . Cambridge ,

MA : Harvard University Press .

Pinker , S . ( 1987 ) . The bootstrapping problem in language acquisition . In B .

MacWhinney ( Ed .) , Mechanisms of language acquisition . Hillsdale , NJ : Erlbaum .

Pinker , S . ( 1989 ) . Learnability and cognition . Cambridge , MA : MIT Press .

Pinker , S . ( 1994 ) . How could a child use verb syntax to learn verb semantics ?

Lingua , 92 , 377 - 410 .

Redington , M ., Chater , N ., and Finch , S . ( 1994 ) . The potential contribution of

distributional information to early syntactic category acquisition . Unpublished

manuscript , University of Edinburgh .

Rothstein , S . ( 1983 ) . The syntactic forms of predication . Unpublished doctoral

dissertation , MIT .

Schieffelin , B . ( 1985 ) . The acquisition of Kaluli . In D . I . Slobin ( Ed .) , The cross -

linguistic study of language acquisition : Vol . 1 . The data . Hillsdale , NJ : Erlbaum .

Schlesinger , I . M . ( 1988 ) . The origin of relational categories . In Y . Levy , I . M .

Schlesinger , and M . D . S . Braine ( Eds .) , Categories and processes in language

acquisition . Hillsdale , NJ : Erlbaum .

Shipley , E . F ., and Shepperson , B . ( 1990 ) . Countable entities : Developmental

changes . Cognition , 34 , 109 - 136 .

Singleton , J ., and Newport , E . ( 1993 ) . When learners surpass their models : The

acquisition of sign language from impoverished input . Unpublished manuscript ,

University of Rochester .

Soja , N . N . ( 1992 ) . Inferences about the meanings of nouns : The relationship

between perception and syntax . Cognitive Development , 7, 29 - 45 .

Spelke , E . S . , Breinlinger , K . , Macomber , J . , and Jacobson , K . ( 1992 ) . Origins of

knowledge . Psychological Review , 99 , 605 - 632 .

Starkey , P ., Spelke , E . S ., and Gelman , R . ( 1990 ) . Numerical abstraction by

human infants . Cognition , 36 , 97 - 127 .

Valian , V . ( 1991 ) . Syntactic subjects in the early speech of American and Italian

children . Cognition , 40 , 21 - 81 .

Waxman , S . R . ( 1994 ) . The development of an appreciation of specific linkages

between linguistic and conceptual organization . Lingua , 92 , 229 - 257 .

Williams , E . ( 1983 ) . Semantic vs . syntactic categories . Linguistics and Philosophy ,

6 , 423 - 446 .

Wynn , K . ( 1990 ) . Children 's understanding of counting . Cognition , 36 , 155 - 193 .

 ~
Chapter 13

Sortals and Kinds: SusanCarey and F ei Xu

An Appreciation of
John Macnamara

Rereading A Border Dispute (Macnamara 1986), we were struck by the

degreeto which this work , published in 1986 and thus in progress the
years before that, anticipates our researchprogram in 1997, and by how
illuminated we still can be by John's words. Each time we return to this
rich work , we encounter ideas missed before. The similarity between
John's concernsin that book and those that currently engageus is not an
accident, of course, for we have studied the book, read parts of it in
graduateseminarson severaloccasions, and assignedit to and discussedit
with our students.
What John did, first in Namesfor Things (Macnamara 1982) and then
in A Border Dispute, was spell out in detail what it is we learn when we
learn a name like " Freddie" or a noun like " dog." ! As he insisted, we can
only begin to provide an account of learning after we have an adequate
account of what is learned. And what he saw more clearly than any psy-
chologist before him was the extent of the knowledge required to master
such simple words, words that appear in children's vocabulary before the
age of 2.

13.1 What Is LearnedWhena ChildLearns a Proper N ODD

In his analysisof nameslike " Freddie" (the name of the new family dog),
John focusedon severalproperties of proper nouns (PN s): They designate
individuals, which they pick out across all times, places, and counter-
factual worlds. The identities of the individuals are traced by means of
sortals, which denote kinds. Finally , the referencesof PNs are not medi-
ated by a Fregeansense(Kripke 1972).
A PN is often learnedthrough ostension; the mother points out the new
puppy and says, " This is Freddie." In order to learn the PN from such
312 Carey and Xu

input, toddlers must have the ability to refer and to understand when
others are referring. They must have someother meansof identifying the
bearer of the name- a demonstrativesuchas " this" or " that," a point, or
a gaze could all do the trick . They must understand the is of identity,
which specifiesthe relation between this and Freddie. And they must
know a sortal conceptthat tracesthe identity of Freddie through time and
counterfactual situations. Note that identity plays two roles in this
account: (1) the identity of Freddie and the individual picked out by the
demonstrative, point, or gaze, and (2) the identity of Freddie over time,
determinedby the criteria of identity supplied by the sortal dog.

13.2 Sortals

Sortal conceptssupply principles of identity (and as we show later, they

do other work as well). Freddie the puppy turns into an adult dog,
changessize, coloring, location, owner, but is still the same dog. The
sortal dog designatesa kind of animal, so to representdog the child must
also represent the sortal kind. F or the child to learn from ostension
" That's a dog" the sortal dog, he or she must have been directed by
somethingto assignto " dog" the correct referent- not an individual dog,
or a group of dogs, or someproperty of dogs, but a kind.
As John characterized sortals, they are concepts that do the logical
work of count nouns: they provide principles of individuation , numerical
identity, and application. In John's terminology, " principles" are meta-
physical, not psychological. John acceptedthe Kripke /Putnam view that
meaningsare not in the head. What makes a dog a dog is a property of
the world , not of the mind, as is what makes one dog different from
another dog, and that puppy the sameas this dog. By force of representing
sortal concepts, minds deploy criteria of individuation , numerical identity,
and application, but theseare best shots and revisable. As psychologists
studying concepts, the best we can do is to study the criteria our subjects
deploy.
Most psychological work on conceptsconcernscriteria of application:
the basesof people's categorization of entities such as dogs, tables, peo-
ple, and so forth . All conceptsprovide criteria of application- exemplars
of properties such as red and predicatessuch as jumping must be identi-
fied in order for these concepts to be used. What differentiates sorta!
conceptsfrom all others is that only they provide criteria for individua-
tion and numerical identity.
Sortals and Kinds 313

To see the role sortals play in our conceptual life , consider these two

questions : ( 1) How many are there in this room ? ( 2 ) Is that the same as

what was here before ? We cannot begin to answer either question without

specifying individuals - how many what ? We can count tables , people ,

legs , fingers , or shirts , but we can ' t count the red , the sleeping , or the

wood . Only sortals provide criteria of individuation . Similarly , " same , " in

the sense of numerical identity , means same one , and a sortal is needed to

specify the individual being traced through time . The body we see on the

floor is the same one that existed just before Freddie died , but Freddie ,

whose identity is traced by the sortal dog , not the sortal body , has ceased

to exist . Individuals are established by sortals and different sortals provide

very different criteria of identity .

The criteria of individuation that sortals provide are logically distinct

from the criteria of numerical identity . To see this , consider two dogs ,

Freddie and George . As you conceptualize them playing on the lawn , the

criteria of individuation provided by the sortal dog establish two indi -

viduals there , and if they go behind a wall and reemerge , the sortal dog

again establishes two individuals . But additional criteria are needed to

establish which one is Freddie and which is George , or which one

reemerging from behind the wall was the first dog to disappear . Although
the criteria for individuation establish two individuals each time , criteria

for numerical identity are needed in order to determine which dog was
which .

13 . 3 The Continuity Thesis

We agree with John ' s characterization of what is learned when one learns
a PN or a count noun . But John assumed that as soon as a child produced

or comprehended a word like " Freddie " or " dog , " the child had mastered

a PN or sortal term , respectively . Indeed , he thought that the fact that

young infants are sensitive to similarity among exemplars of toy dogs and

differentiate these from exemplars of toy trucks suggests that even pre -

linguistic infants have a sortal concept dog or stuffed dog ( Cohen and

Younger 1983 ) . But as Quine ( 1960 , 1969 , 1974 ) forcefully argued , as

Millikan ( 1998 ) has recently endorsed , and as psychologists studying the

earliest stages of language acquisition ( e . g . , Vygotsky 1962 ; Dromi 1987 )

have repeatedly stressed , young children could establish representations of

words that differ markedly , even qualitatively , from those adults establish
for the same words .
314 CareyandXu

Psychologists have offered many different proposals for how young

children's word meanings may differ qualitatively from those of adults.
Vygotsky (1962), for example, suggestedthat toddlers' words refer to
complexes of properties, such as Dromi 's (1987) son's use of " paper"
to mean pens, paper, pencils, crayons, the act of writing or drawing or
cutting, and so on. Words in no natural language unite such disparate
entities in a single category. Landau, Smith, Jones (1988), for another
example, propose that early count nouns refer to entities specified by
shapes. Although there are words in natural languagesthat refer to shape
(e.g., " round," " square," " dog-shaped" ), these are not count nouns like
" dog." If proposals such as theseare correct, then early count nouns are
not sortals and their meanings are indeed qualitatively different from
those that adults assignthem.
Perhapsthe most articulated proposal for how the earliest meaningsof
early count nouns might differ from those of adults is Quine's (1960, 1969,
1974; seealso Millikan 1998for a nearly identical proposal) . Quine also
explicitly denied that infants' earliest nouns function as sortals; he
claimed that infants lack the capacity for what he called " divided refer-
ence," which is the capacity to refer to distinct individuals of a given kind .
On Quine's proposals, the first meaningsof words like " Mama," " red,"
" water," and " table" do not differ qualitatively or systematically from
each other. They all refer to " scatteredportions of what goes on" when
adults use the words, that is, to featuresof the perceptual experiencesthe
child is having when he or shehears the words. If this is so, then " table"
does not pick out a sortal concept (seeCarey 1994for a detailed exposi-
tion of Quine's speculationsand a review of the empirical evidenceboth
consistentand inconsistentwith them).
We will argue below that these proposals are wrong; infants do have
the capacity for divided reference, and terms like " dog" serve as sortals
from the very earliest stagesof languageacquisition. However, proposals
such as thosereviewedabove are seriousempirical possibilities. Our point
in raising them is to question a methodological tenet John articulated
repeatedly during his career- a methodological commitment to strong
continuity of logical and conceptual resourcesthroughout development.
For example, in A Border Dispute, he wrote, " I believe it is wise psycho-
logical strategy to assumethat a child's mind resemblesan adult's, unless
there is evidenceto the contrary. That is, I take seriouslythe obvious null
hypothesis- that the two are similar. This placeson researchthe onus of
discoveringdifferencesbetweenchild and adult" (p. 56). John recognized,
Sortals and Kinds 315

of course, that ultimately it is an empirical question; strong continuity

could be false. WhereasJohn believedthat there is no convincing evidence
against continuity, we disagree.

13.4 Where EvidenceFavors Continuity

John assumed, and we agree, that much of what is necessaryto support

the acquisition of sortals and PN s is innate, or at least in place before the
first words are learned (the capacity for reference, some demonstratives,
the is of identity). We will not belabor this point here, merely noting that
research both before and since A Border Dispute supports John's
assumption. As just one example of the kind of work we have in mind,
consider demonstratives and reference. To understand reference and
demonstratives, the infant must understand that other people can pay
attention to entitiesin the world, that words (and other mental entitiessuch
as someemotional states) refer to things in the world , and that people can
indicate in various ways what they are referring to. In a justly famous
paper, Scaifeand Bruner (1975) showedthat infants as young as 7 months
of age follow eye gaze; if an adult turns to look at something, infants will
turn their head and eyesin the samedirection. Hood, Willen , and Driver
(in press) have recently shown that under restricted circumstances,2 even
3-month-olds will do the same. Johnson, Slaughter, and Carey (in press)
recently showedthat this behavior is elicited, among 12-month-olds, by a
facelessnonsenseobject, so long as that object reacts contingently to the
child. They argue that the baby takes contingent reaction as evidencethat
the object is capable of intentional states, and takes a sudden turn from
" facing" the baby to " facing" 45 degreesto the left or the right as a shift
in attention. The babies turned to seewhat the object was attending to.
Finally , in a seriesof elegantexperiments, Baldwin and her collaborators
(Baldwin and Moses 1994) showed that infants under a year monitor
what the mother is looking at when she says, " Yuck " in a disgustedtone
of voice, or " Mmm , how nice." Even if they are examining an unfamiliar
object at the time of the emotional utterance, they do not assumethat the
mother's " Yuck" refers to it . And by 15 months of age, infants similarly
monitor the gazeof the speakerwhen the speakerusesa novel word (e.g.,
" That's a blicket" ), evenif they themselvesare attending to a novel object
that is a perfectly good candidate for the referent of that word at the time
of the utterance. In sum, infants and toddlers attribute to other people
both intentionality and the capacity to refer, and they have at least some
316 Carey and Xu

heuristics for figuring out what entities a person is referring to , just as

John 's continuity assumptions dictate .

13.5 Prelinguistic Representation of Sortal Concepts

In order for a child to learn the meaning of a sortal term like " dog " or a
PN like " Freddie ," more is required than an understanding of reference,
intentionality , and demonstratives . The child must represent sortal con-
cepts, specifically the sortal dog. When do sortal concepts become avail -
able to the 'child ? Our focus here is the state of evidence that children
possess a sortal such as dog that provides criteria of individuation and
identity of Freddie . Even John , strong continuity theorist as he was, did
not think that the sortal concepts themselves, such as dog, table, and book ,
are innate . So how and when are they acquired ?
John explicitly denied that there is an unlearned expression synony -
mous with dog in the language of thought . Rather , he thought the origin
of the sortal dog is a gestalt supplied by vision , a prototype abstracted
from experience with dogs. Indeed , prototype abstraction is a powerful
psychological process (Posner and Keele 1968) available to infants
(Cohen and Younger 1983; Quinn 1987). John 's idea was that the psy-
chological representation of basic-level sortals such as dog is equivalent to
same kind as [ gestalt of dog J .

13.5.1 Prima Facie Evidence for Prelinguistic Representations of Sortals

In support of this proposal , John pointed to the experimental evidence
that children are able to classify objects into categories for which they
have no natural language symbols . Since A Border Dispute , there is now
even more evidence of this type . For example , Quinn and Eimas have
carried out an extensive series of studies showing that infants as young as
3 months of age perceive the similarity among photographs of exemplars
of a variety of natural kinds (e.g., dogs, cats, horses). In their procedure ,
infants are familiarized with pairs (typically six) of photographs of a
single kind of object (e.g., six pairs of cats, taken from different angles, of
different breeds, coloring , and so forth ). On test trials they are presented
with a pair of photographs consisting of a new cat and a member of a
contrasting category (e.g., a dog , a horse, a car) . A looking time prefer -
ence for the exemplar of the novel category is taken as evidence that
infants are sensitive to the categorical similarity among the original
Sortals and Kinds 317

familiarization set (e.g., Quinn, Eimas, and Rosenkrantz 1993; Eimas and
Quinn 1994). In thesestudies, infants as young as 3 months of age succeed
in categorization at the basic level (e.g., distinguishing dogs from cats).
Also, Mandler and her colleagueshave carried out an extensiveseriesof
studies with older babies, using the manual habituation procedure pio-
neered by Ruff (1986), in which infants examine a series of toys from
a single category (e.g., a bird, a turtle, a dog, a fish) and then are given
either a new animal (e.g., an elephant) or a new exemplar from a novel
category (e.g., a car) to play with . Again, novelty preferenceis a measure
of categorization. With this technique, infants show evidenceof catego-
rization at the more superordinatelevel (what Mandler calls the " global
level" ) by 7 months of age, and at the basic level by 9 months of age (see
also Mandler, Bauer, and McDonough 1991; Oakes, Madole, and Cohen
1991; Waxman and Markow 1995).
John was aware, of course, that these data do not prove that infants
representsortals such as dog, stuffeddog, animal, vehicle, or car. He asked
of a study using a manual hab'ituation procedure (Cohen and Caputo,
cited in Cohen and Younger 1983) whether one could be sure that infants
had formed a concept such as stuffed animal, animal, or toy animal, as
opposedto basing their responseon a perceptual feature common to all
11 toys in the study, such as texture or shape. But he clearly meant his
question to be rhetorical, for he answeredit , evidently to his own satis-
faction, with the observation that anyone familiar with 7-month-olds
has the distinct impression that they already know many categoriesof
objects- animals, bowls, cups, hands, bottles, and so forth . By that age,
they behavedifferently toward cups and toward their contents.
John's responsemissesthe mark in two ways. First of all, it is not clear
that there are good data on this point . What data there are suggestthat
knowing what to do with membersof kinds, at least newly encountered
kinds, is not robust until 12 months and later. Baldwin, Markman , and
Melartin (1993) asked when infants would project a distinctive action
relevant to a kind of object (e.g., turning it a certain way to produce a
sound) on the basis of kind membership, and found only shaky and weak
evidence for this ability at 9 months, in the face of definitive success
at ages 12 and 14 months. Second, supposewe did grant the point (that
babies of this age know that to do with bottles). Babiescould simply rec-
ognize examples of bottleness, bottlehood, or bottle-shape, and have
associatedthis property with other properties such as containing milk or
juice. As Quine so rightly pointed out, recognizing a shapeand associat-
318 Carey
andXu

ing that shape with other properties is not equivalent to categorizing

under a sortal.
Another way to put this second point is that John's proposal has a
logical problem: samekind as [ gestalt of dog] contains criteria neither of
individuation nor of numerical identity. For this expressionto serveas the
representation of the earliest sortals, it must be elliptical for same kind
of individual as [ gestalt of dog} , but then what specifiesthe individual?
Further, as indicated above, the habituation data surely show sensitivity
to similarity in shape, but shape(e.g., round, square, or dog-shaped) is a
property, not a sortal. At the very least, we would want to know whether
representationsof [ gestalt of dog] and [ gestalt of cup] provide criteria
of individuation and numerical identity for infants. We have carried
out three extensive series of studies with 10- and 12-month-old infants
addressingthis point and have clear and consistentevidencethat for 10-
month-olds the answeris probably not.

13.5.2 Sortals for Kinds of Objects Not Representeduntil Late in the

First Year of Life
Imagine the following scenario: a duck emergesfrom one side of a screen
and returns behind it , and then a ball emergesfrom the other side and
then returns. How many objects are behind the screen? For adults, the
answer is clear: at least two, a duck and a ball. We know there are two
becausewe code them under different sortals: a duck and a ball must be
two different individuals. We (Xu and Carey 1996) asked whether the
answer is similarly clear for babies. In a seriesof four experiments, we
showed infants scenariosas in figure 13.1. Sometimesthe objects were
membersof contrasting kinds at the global level (e.g., a duck and a car),
the level of contrast infants of this ageconsistentlyhave beenshown to be
sensitiveto in manual habituation studies. Sometimesthe contrasts were
of highly familiar functional kinds (e.g., a cup and a ball; a bottle and a
book), the level at which exemplarselicit contrasting actions according to
John's intuitions . Sometimesinfants were fully habituated to each object
as it emergedfrom each side; other times they were more briefly fami-
liarized with a fixed number of emergences .
We infer how many objects the infants posit in this situation by
removing the screenand measuring how long the infant looks at what is
revealed there- either the expectedoutcome (for adults) of two objects
or the unexpectedoutcome (for adults) of only one of the two objects. If
infants have the sameexpectationsas adults, they look longer at the un-
Sortals and Kinds 319

Property /Kind Condition

1. Screen introduced

"' cc ct
.\
_c/)"J!c
"<
2. Object 1
\c
--";~
_C brought out

3. f - Object 1
returned

4.
T
~:- Object

brought
2

out

5. Object 2

returned

Steps 2-5 repeated

Screen removed ,
revealing
,~il? "t ;~J
111111111111

6. Expected outcome

or

"',",-- II
'
,"
i
'\ 0-
- rw
: Y !

Unexpected outcome

Figure 13.1
Schematic representation of the property fkind condition in Xu and Carey 1996
320 CareyandXu

expected outcomes in experiments such as these that exploit the violation -

of -expectancy looking -time methodology (see Spelke 1990; Spelke et al.
1992; Baillargeon 1995 for reviews).
The results were the same under all the conditions : 10 -month -olds

failed to draw the inference that there should be two objects behind the
screen, whereas 12-month -olds succeeded in doing so. Further controls
established that the method was sensitive for 10 -month - olds as well as

older infants ; if the younger babies were given spatiotemporal evidence

that there were two numerically distinct objects in these events (e.g., if
they were shown an elephant and a car together, at the same time, next to
the screen for a few seconds before the successive emergences took place),
then they too succeeded by looking longer at the unexpected (one-object )
outcome than at the expected (two -object ) outcome .
We do not deny the lessons from the visual and manual habituation
studies . Ten - month - olds are sensitive to the differences between cups and
cars, ducks and balls (and they are so under the conditions of our ( 1996)
studies; they take longer to habituate to a duck and a car alternately ap-
pearing from each side of the screen than to a car repeatedly appearing
from behind the screen). The conclusion to be drawn from our (1996)
studies is that 10-month -olds do not yet represent sortals such as duck ,
car , animal , vehicle , cup , bottle , book . The differences that babies under 10
months of age are sensitive to are irrelevant to object individuation .
This conclusion may be premature , however . In order to arrive at a
representation of two objects in our ( 1996) task , infants must be able to
recall the representation of the first object , including its kind , and com -
pare this recalled representation with that of the second object . Perhaps
the failure of the 10-month -olds reflects an information -processing limi -
tation on short -term memory , rather than the unavailability of the rele-
vant sortal concepts. To address this possibility , we explored whether
kind contrasts would help babies individuate objects under conditions
of ambiguity in two very different paradigms , one of which makes no
demands on short -term memory and one of which makes even greater
information -processing demands than does our ( 1996) paradigm .
In the studies reported in Xu and Carey 1996, ambiguity regarding
object individuation was introduced by occlusion ; the object is not always
visible , so there is no unambiguous spatiotemporal evidence about how
many objects are present in a scene. Ambiguity can also be introduced by
shared boundaries , as in figure 13.2 (top panel). How many objects are
depicted in this figure ? Adults respond that there are two , a duck and a
 Sortals and Kinds 321

HABITU
ATI0N

"' ~

,'
-
r
oil
' ~ ~CC
)
tJf ,

APART
(EXPECTED
)

\I.,iCc
(1 ~c):~
('
,coccc
TOGETHER
(UNEXPECTED
)
C'/!f;\) 8
;~
)\Z1
g
Figure 13.2
Schematic representation of the duck -car experiments in Xu , Carey , and Welch
1998

car . However , it is also possible that there is only one , the top half of
which resembles a duck and the bottom half of which resembles a car .
When adults parse the object into two , they are using their knowledge of
the sortals duck and car to do so .
We asked when infants would segment an ambiguous object such as
that in figure 13.2 into two separate objects , the duck and the car . Here
the kind distinction is at the globallev ,el (an animal and a vehicle ) . We
also examined a display consisting of two familiar functional kinds (a cup
on top of a shoe ) . We again used a violation -of -expectancy looking -time
paradigm . Infants were habituated to the ambiguous duck / car display , or
to the ambiguous cup / shoe display , with a hand poised a few centimeters
above it . After the habituation , the hand grasped the top object and lifted ,
revealing either the expected outcome (for adults ) of just the top object
coming up (figure 13.2 , bottom left ) or the unexpected outcome (for
adults ) of the single duck / car or cup / shoe object being lifted as a single
piece (figure 13.2 , bottom right ) . Notice that in this experiment , no short -
term memory demands are placed on the child . Both the duck and the car ,
322 Carey
andXu

or the cup and the shoe, are fully visible at all times when the ambiguous
display is in view. We found exactly the sameage effectsas we reported
in Xu and Carey 1996. Ten-month-olds failed to use the kind contrasts
betweenthe duck and the car, or the cup and the shoe, to infer that there
were two separateindividual objects in this display; they did not look
longer at the unexpectedoutcome of the single object being raised as a
piece. Twelve-month-olds, in contrast, succeededat the task, confirming
both that 12-month-olds have constructed relevant sortals and that the
task sensitively reflects such representations. The sensitivity of the task
was further confirmed by the finding that if the 10-month-old infants were
given spatiotemporalevidenceconcerning the number of objects in this
display (if the objects were briefly moved, laterally, relative to each other
at the beginning of each habituation trial ), they succeeded ; they looked
longer if the object moved as a singlepiece (Xu , Carey, and Welch 1998).
Thus, data from two quite different paradigms, which place very dif-
ferent information-processing demands on the child, converge on the
sameconclusion. Even though infants can discriminate ducks from cars,
cups from shoes, contrasts such as these do not provide criteria for
individuation until sometimebetween10 and 12 months of age.
The two studiesreviewedabove dependupon a rather indirect measure
of how many objects infants take to be involved in these events- their
pattern of looking times at outcomes of one or two objects, compared
to baseline preferencesfor these same outcomes in the absenceof any
habituation of familiarization . Other, more direct measuresof how many
objectsinfants believeto be involved in someevent can be imagined. One
cannot ask the babies, for they obviously cannot talk . But we could have
them reach for hidden objects and measurehow often and persistently
they reach as a function of how many objects are in a box and as a func-
tion of what kind of evidencethey have had concerning this matter. This
is the approach taken in our third line of studies in this series. In these
experiments, on two object trials infants watched the experimenterpull
out an object (say, a toy telephone) from a box, return it to the box, then
pullout a secondobject (say, a duck), and return it to the box. On one-
object trials, the experimenterpulled out one object (say, the telephone),
twice. Infants were then allowed to reach into the box (the opening of the
box was well covered with fabric so the infants could not seewhat was
inside) to retrieve the objects. Infants at both 10 and 12 months always
reachedin and retrieved the first object, which they were allowed to play
with for some time before it was removed from them. Unbeknownst to
Sortals and Kinds 323

them , the second object had been removed from an opening at the rear of
the box , so they could not have felt it when they retrieved the first object .
At issue was what happened then . At 12 months , infants reached persis-
tently for the second object on two -object trials , exploring the inside of
the box , apparently trying to find it . On one-object trials they either did
not reach a second time at all , or did so only cursorily (after all , there was
nothing else to do ; the box was left in place and the first object had been
taken away). At 10 months of age, the pattern of reaching was identical
on one-object and two -object trials : cursory reaching in both cases. At
10 months , however , infants succeeded in a condition in which they were
given spatiotemporal evidence that there were two objects on two -object
trials . ( Twelve -month -olds succeeded in this condition as well , of course.)
If both objects had been shown together before being replaced in the box ,
babies of both ages reached persistently for the second object , differ -
entiating these trials from the one-object trials ( Prevor 1997; Van de
Walle , Prevor , and Carey 1998) .
Note that this is exactly the same pattern of results as in the first
two series of studies. Twelve -month -olds succeeded in establishing repre-
sentations of two objects on the basis of kind differences, whereas 10-
month -olds did not . At both ages babies succeeded when provided
spatiotemporal evidence that there were two objects in the box .
The first lesson we draw from these widely different studies of object
individuation is that John was wrong in the inference he drew from young
infants ' success at categorizing dogs or animals together and differ -
entiating them from cats or vehicles, or from young infants ' knowing
what to do with a cup , ball , bottle , or shoe . These achievements do not
mean that young infants represent the relevant sortal concepts . If we are
right in our interpretation of our results, we have provided evidence
against John 's continuity thesis: infants 10 months of age and younger are
clearly capable of seeing animals as similar to each other , but they do not
represent a sortal animal . Coming to represent one's first sortals for kinds
of objects would be a good candidate for the construction of a new rep-
resentational resource .

13.6 Success Based on Sortals or Properties at 12 Months ?

So far , we have drawn two conclusions from the above data : ( 1) 10-
month -olds do not represent sortals for kinds of objects, and (2) 12-
month -olds do represent sortais for kinds of objects
. The inferencefrom
324 Carey and Xu

the 10-month-olds' failures in these tasks to the first conclusion seems

warranted, but the inference from the 12-month-olds' successesto the
second conclusion is not, for the following reason. Although concepts
lexicalizedas properties do not provide criteria for individuation (we can't
count the red in this room, the big in this room, the striped in this room),
in the situations we have describedabove the property differencesamong
the objects certainly do provide relevant information about number. For
example, if A seesB draw a black, plastic thing from a box, return it , and
then draw out a red, striped, round, rubber thing, A might infer there are
two objects in the box, even if A did not identify the two objects as a toy
telephone and a ball. From what we have presentedso far, we do not
know whether the 12-month-olds are individuating the objects on the
basisof kind differencesor on the basisof property differences. We do not
know whether their representationalsystemmarks the distinction between
properties and kinds.
We have recently completeda seriesof studieswith 12-month-olds that
begins to addressthis question (Xu , Carey, and Quint 1997). The pro-
cedure described in Xu and Carey 1996 (seefigure 13.1) was followed,
except that the objects were, for the most part, objects of a single kind
that differed in properties such as color or size. For instance, a pink cup
emergedfrom the left, and an identical yellow cup then emergedfrom the
right; or a big cup emergedfrom the left, and a small cup then emerged
from the right; or a coffee mug emergedfrom the left, and a sippy cup
then emergedfrom the right . Finally , we tried a pair that differed in many
properties: a big, red, plastic cup with colored dots and a small, blue,
ceramic cup. The outcome displays were similar to those described in
Xu and Carey 1996: the screenwas removed to reveal either two cups
(one big, one small; or one pink, one yellow) or just one of the two cups.
Adults certainly can use these property differencesto infer two distinct
objects. However, the 12-month-olds failed to individuate the cups on the
basis of these property differences. In contrast, they succeededat indi -
viduating a cup from a ball in the study reported in Xu and Carey 1996,
and a duck from a car in the study reported in Xu , Carey, and Welch
1998. So far, only kind differenceshave beenfound to provide 12-month-
olds with criteria for individuation and identity in theseexperiments.
This seriesof studiesprovides preliminary evidencethat 12-month-old
infants representthe distinction between kinds and properties, and that
their successesin the above seriesof studies are indeed based on their
early sortals for kinds of objects, not on their representationsof proper-
Sortals
andKinds 325

ties. We tentatively conclude, then, that Quine's speculationshave been

shown to be false. At the very outset of language acquisition, children
representthe distinction betweenproperties and sortals, well before they
have even begun to master the syntactic reflection of thesedistinctions in
any natural language. As we will show in section 13.9, Quine may have
beenright in one respect; languagelearning may have a role to play in the
construction of the first sortals for kinds of objects.

13.7 PhysicalObjectasa Sortal

There is a third lesson to be drawn from the individuation studies

reviewedabove. Rememberthat the 10-month-old infants succeededin the
spatiotemporal conditions in all three seriesof studies. When shown both
objects together at the sametime separatedin space, infants established
representationsof two objects, which they maintained throughout the
event. If representationsof sortal conceptsare those representationsthat
provide criteria for individuation and numerical identity, young infants
do represent at least one sortal concept, physical object, itself. There is
now massiveevidencethat infants as young as 4 months of age establish
representationsof bounded coherent rigid objects that move on spatio-
temporally continuous paths, individuating these objects on the basis of
criteria such as " one object cannot be in two placesat the sametime; one
object cannot go out of existenceat one time and place and come back
into existenceat another time and place" (e.g., Baillargeon and Graber
1987; Kellman 1984, 1993; Spelkeet al. 1995). For example, Spelkeet ale
(1995) showed5-month-old infants objects emergingfrom the far sidesof
two screensseparatedin space, never appearing in the middle of the two
screens. That is, an object would emerge from the left side of the left
screen and return behind it , followed by a physically identical object
emerging from the right side of the right screenand returning behind it .
The infants never saw both objects together. Nonetheless, they inferred
from the fact that these two objects could not be on a single spatio-
temporal trajectory that there must be two objects involved in this event,
for when the screenswere removed, they looked longer at the impossible
outcome of only one object behind the two screensthan at the possible
outcome of two objects(seeXu and Carey 1996, for a replication with 10-
month-olds).
We conclude from studies such as these that infants distinguish one
object seenon different occasionsfrom two numerically distinct objects.
326 Carey and Xu

To do so , they must have criteria for individuation and numerical identity

for physical objects . Recall our argument above that same kind as [ gestalt

of dog ] is not a representation of the sortal dog . We have now provided

an answer to that might specify the individual in the expression , same

kind of individual as [ gestalt of dog ] , namely , the individual picked out by

the sortal physical object . This insight provides part of the answer to the

question John posed concerning how infants learn sortals like dog , but it is

only part of the answer . So far , we have shown that at 10 months , infants

can represent both the sortal physical object and the perceptual [ gestalt of

dog ] without having the sortal dog , since [ gestalt of dog ] does not pro -

vide criteria of individuation and identity .

More is needed for the infant to construct the first sortals for kinds of

objects . But before we offer our speculations about what else the infant

recruits for this task , we must address another issue . John would certainly

not have accepted our friendly amendment of his expression same kind as

[ gestalt of dog ] to same kind of individual as [ gestalt of dog ] , withphysi -

cal object specifying the relevant individual . And that is because he , like

many others , explicitly denied that physical object is a sortal concept . In

what follows , we sketch why he and others hold this position , and we

show why they need not do so .

13 . 8 Problems Arising from Considering Physical Object as a Sortal and

a Possible Solution

Along with philosophers such as Geach ( 1957 ) , Wiggins ( 1980 ) , Gibbard

( 1975 ) , Gupta ( 1980 ) , and Hirsch ( 1982 ) , John denied that object or thing

is a sortal . Their arguments have three parts . First , they claim that people

do not have clear intuitions about the criteria for judging something as the

same object or the same thing . More importantly , if object / thing is a sortal ,

then questions arise concerning the relations between this general sortal

and more specific ones . Perhaps there are no more specific sortals such as

car , cup , or dog - these more specific sortals might now be construed as

predicates of the general sortal object / thing , thus undermining the enter -

prise these philosophers have advocated . Finally , if ph } 'sical object and

dog are both sortals , paradoxes of transitivity of identity seem to arise .

Here , we focus on John ' s particular version of these worries .

In A Border Dispute , John used the following example to illustrate " the

weakness of our intuitions about how to trace identity under the descrip -

tion ' the same thing ' " ( p . 52 ) . Suppose that all of the atoms constituting
Sortals and Kinds 327

Ronald Reagan's body when he was the governor of California were re-
placed by new ones and the old oneswere reassembledto foffi1 the body
of Walter Mondale in 1984. According to John, if we trace identity under
the sortal thing, we would encounter the following problem of the tran-
sitivity of identity: Governor Reagan is identical with PresidentReagan,
Governor Reagan was identical with a set of atoms, and the sameset of
atoms is identical with Walter Mondale; therefore, by transitivity of iden-
tity , Governor Reagan was also identical with Mondale. SinceGovernor
Reagan is also identical with PresidentReagan, we would conclude that
Reaganwould stand for election against himself, " which is preposterous,
and not just on political but logical grounds" (p. 53). The solution, said
John, is to trace the identity of Reagan under the sortal person, which is
not identical with a set of atoms.
Similarly, in The Logical Foundationsof Cognition (Macnamara and
Reyes 1994), John argued that thing/object does not provide criteria for
what to count as one instance of something; that is, it does not provide
principles of individuation . If one were to count the things in the room,
one could count the chair as a chair, four legs, plus one top, hence six
things altogether!
It may indeed be the casethat the English words, " object" or " thing,"
are not sortals, given people's vague intuitions about their criteria for in-
dividation and identity. However, physical object, or Spelke-object, which
is any entity that is three-dimensional, is bounded, and retains its bound-
aries as it moves through spaceand time, is a sortal. It provides princi-
ples of individuation and identity, and it does not eliminate the necessity
for postulating other more specificsortals such as cup or dog. Finally , in
John's formalism for categories, a natural extension can be made to
specify the relation betweenphysical object and sortals such as cup and
dog, an extensionthat dissolvesthe paradoxesof identity (seeXu 1997for
further discussionof theseissues).
Let's go back to John's two examplesabove and seehow applying this
more restricted senseof object- namely, physical object- can help us
dissolvethe problems.
John's first example, involving Reagan, Mondale, and the set of atoms
constituting Reagan, is a caseof the generalproblem of the transitivity of
identity. Note that becauseJohn believed that intuitions about the same
object or the samething are vague, he stipulated that thing/ object is equal
to the set of atoms constituting Reagan. However, the more restricted
senseof object, physical object, is not the sameas a set olatoms . The set
328 Carey and Xu

of atoms constituting an object could change; the object still remains the
sameobject. Therefore, the problem of transitivity of identity disappears
when we apply the sortal object to Reagan, just as it disappearsif we
apply the sortal personto Reagan, as argued by John.
John's second example illustrates how our concept of thing/ object
is vague. Indeed, one could think of the top and the legs of a chair as
objects. However, it seemsunlikely that if one were asked to count the
things/ objectsin the room, a chair would be counted as six things/ objects.
Instead, most people would agreethat a chair is one thing or one object.
Why is this so? Our intuition is that although in everydayspeech, a leg (of
the chair) is a thing/object, when asked to count in this fashion people
spontaneouslyapply the more restricted senseof object/thing (namely,
physical object) , which eliminates the ambiguity, and the result is a count
of one thing/object given one chair.
Clearly, if physical object is a sortal, one must show that it provides
principles of individuation and numerical identity. From the empirical
work reviewed above, we seethat physical object doesprovide such crite-
ria: (1) one object cannot be in two places at the same time (e.g., two
identical cups in two distinct locations simultaneously are two numeri-
cally distinct cups); (2) objectstravel on spatiotemporally connectedpaths
(e.g., if an object appears to have "jumped" from point A to point B
without traversing a continuous path in between, there must be two dis-
tinct objects); (3) objects maintain their boundariesand move as a whole
(e.g., if one part of an object moves independentlyof the rest, there must
be two distinct objects). Thesecriteria are well specified, and both adults
and infants deploy them.
What about the second, deeperworry that if thing/object is a sortal, all
the other more specificsortals can be construed as predicatesof this gen-
eral sortal? The existenceof such a general sortal is damaging for advo-
cates of the sortal view since this general sortal may be the only one
needed. At the sametime, there are good reasonsto believethat cup, dog,
and rock are sortals since they provide principles of individuation and
identity above and beyond the principles provided by the sortal physical
object. A dog ceasesto exist when it dies, but its body continues to exist
and it is the samebody as before. The problem is in fact a rather general
one sincemore than one sortal can be applied to any given entity; besides
being an object, that thing over there (indicating Freddie) is also a pet, a
poodle, a mammal, an animal, a body, a set of atoms.
What is the relation among thesesortals? Macnamara and Reyes(1994)
argue that contrary to classicallogic, it is not one of class inclusion and
Sortals and Kinds 329

identity. For example, the set of atoms constituting a dog could change,
but one would still have the samedog; in casesof metamorphoses, a dog
could change into a mouse, or a frog into a prince, and one would still
have the same animal or the same living thing. Macnamara and Reyes
(1994) present a fonnalism for characterizingthe relations among sortals
by means of underlying maps. An underlying map is a function that
maps membersof a setA to membersof a set B. For example, a passenger
from set A is mapped onto a person from set B by an underlying map u:
passenger~ person. A member of the kind passengeris identified with,
but not identical to, a member of the kind person through the mapping
process. Similarly, a member of the kind dog is identified with a member
of the kind animal by the samemapping process. This fonnalism can be
extended to allow members of the kinds dog/animal/car to be identified
with membersof the kind physical object.
In sum, we have shown that a more restricted senseof object- namely,
physical object- is a sortal for both infants and adults, and one can
specify the relation betweenthe sortal physical object and the more spe-
cific sortals such as dog or chair by underlying mapping, thus avoiding
long-standing problems regarding transitivity of identity. Postulating the
sortal physical object is part of the solution to how infants construct sor-
tals such as dog or chair, and the hypothesisthat young infants represent
it is supported by empirical evidence.

13.9 Final Speculations

: How Are the First Sortals for Kinds of Objects
Learned?

Before infants are 10 months of age, they are able to abstract a visual
prototype from a set of dogs, and they representat least one sortal con-
cept, physical object, yet they have not constructed the sortal dog. And ,
we have argued, they have constructedsomesortal that individuates dogs,
or at least toy dogs, from other kinds of objects by 12 months of age.
How do they do so?
John posed the question in the context of word learning, asking how
children learn the meaning of a sortal term like " dog." His answer was
that infants bring to the word-learning task an understanding of refer-
ence, demonstratives, the is of identity, the sortal kind, and a sortal that
covers dogs: same kind as [ gestalt of dog] . We have offered a crucial
friendly amendment to his proposal: that the sortal that covers dogs is
same kind of physical object as [ gestalt of dog) . Physical object, in this
330 Carey and Xu

expression, provides spatiotemporal criteria for identity and individua -

tion , and these are not overridden by the more specific sortal dog, even
though dogs are not identical to their bodies . It is true that no two dogs
can be in the same place at the same time ; one dog cannot be in two places
at the same time , and dogs trace spatiotemporally continuous paths . But
physical object does more for babies than establishing individuals in their
immediate models of the world ; it enables them to establish individuals in
order to learn about their kinds .

We have not yet dwelt upon John ' s proposal that the child 's ability to
learn specific sortals requires a prior sortal , kind . When we mentioned this
proposal in section 13.2, we gave only John 's logical justification for it :
the child must eventually distinguish concepts that cover kinds of in -
dividuals , groups of individuals , individuals themselves , and properties of
individuals , and so must have some way of representing these distinctions .
We certainly agree with this logical point , and we offered empirical evi-
dence in section 13.6 that infants have distinguished kinds of objects from
properties of objects by 12 months of age.
But here we wish to emphasize the importance of the prior sortal kind
to the learning of sortals for kinds of objects. Suppose children not only
know that there will be kinds of objects, but also have expectations about
kinds that help them identify candidate kinds in their experience of the
world . It is an open question what these antecedently represented expec-
tations will be. We tentatively offer the following suggestions: ( 1) kinds are
determined by deep, often hidden properties , which determine some of the
surface properties of objects, (2) the deep properties that determine kinds
influence many intercorrelated surface properties , (3) the deep properties
that determine kinds are causal and thus interpreted in terms of whatever
causal understanding the child has constructed , (4) object kinds do not
change over time , whereas surface properties may , and (5) object kinds
are named by adults . Notice that these are properties of substance sortals
like dog and animal , not stage sortals like pet and passenger. But stage
sortals are acquired late , and children 's first interpretation of a count
noun .applied to a novel object is that it is a basic-level substance sortal
( Hall 1993).
The sortal physical object enables infants to pick out individuals and
trace them through time , so that they can observe which properties change
with time and which remain stable, and so that they can observe stable
intercorrelated properties . Such observation would help identify kinds , if
the baby is looking for them . There is now evidence that some of these
Sortals and Kinds 331

assumptionsare beginning to playa role in the infant's conceptual life

betweenages9 and 12 months. Baldwin, Markman , and Melartin (1993)
have shown that infants 12 months of age and older expecteda hidden
property of a given object (e.g., the disposition to make a given sound if
handled in a particular way) to also be a property of another object of
the same shape and size, as if kind membership predicts nonobservable
properties. Further, Sabbaghand Baldwin (1997) recently reported pre-
liminary results that 10-month-old infants in our (1996) paradigm might
succeedif the two objectsbehavein distinctive mannerspredicted by their
form when they emergefrom each sideof the screen(e.g., if a round object
rolls and a squareobject turns discretely), as if form-function correlations
provide information about object individuation .
It is the fifth assumption- that words will refer to object kinds- that
has recerived the most systematicstudy. If this connection were part of
the child's language-learning apparatus, then the infant could use adult
word usageas a clue about the kinds in the world . Three results are con-
sistentwith this possibility. First we (Xu and Carey 1996) divided the 10-
month-olds in our two studies with highly familiar kinds (bottle, book,
cup, ball) into two groups: those whose parents said they understood the
words for at least two of thesekinds (lessthan a third of the sample) and
those whose parents said they understood zero or at most one of the
words (the majority of the sample). Those who comprehendedthe words
succeededon the individuation task, even though the objects were never
named. At the very least, it appearsthat infants will not learn a word that
means bottle-shaped, bottleness , that they must construct the relevant
sortal to map the word onto. But it is also possible that word learning
plays a role in the construction of the sortal. In support of that possibility,
Xu (1997) haspreliminary findings that naming the objectsas they emerge
from both sidesof the screensin our task (Xu and Carey 1996) promotes
successat 9 months of age. And finally, in a seriesof studies, Waxman
and her collaborators have found that naming objects during both
manual habituation and visual habituation promotes categorization at
both the basic and global basic levels among both 9-month-olds and 12-
month-olds (Waxman and Markow 1995; Balaban and Waxman 1997;
Waxman and Balaban 1996).
In sum, we offer a tentative answer to the question John raised in
Namesfor Things and A Border Dispute: how do children construct the
sortals that are the meanings of count nouns and provide the criteria of
identity for proper nouns? We have argued, on both logical and empirical
 332 CareyandXu

grounds, that John's solution will not work : samekind as [ gestalt ofdogj
is not a sortal that covers dogs. Also neededare (1) an individual that
provides the first criteria of individuation and identity and (2) a stronger
notion of kind than can be provided by shape similarity. Regarding the
first, we offer the sortal physical object. And regarding the second, we
offer severalexpectationsabout kinds that infants might bring to the task
of constructing their first sortals and some tentative evidencethat these
expectationsmay actually playa role in this achievement.
Notes
1. Throughoutthe chapter, mentionsof words are in quotation marks and
mentionsof mentallyrepresented conceptsor sortalsarein italics.
2. Childrenthis younghavedifficultydisengaging attention. Followingeyegaze
requiresdisengaging attentionfrom a visibleface. In Hood, Willen, and Driver's
procedure , 3-month-old infantssucceededwhentheywerelookingat a schematic
facewhoseeyesmovedand whichthendisappeared from the screen , leavingthe
infantsfreeto follow the gaze.
References

Baillargeon , R . ( 1995 ) . A model of physical reasoning in infancy . In C . Rovee -

Collier and L . P . Lipsitt ( Eds .) , Advances in infancy research ( Vol . 9 , pp . 305 -

371 ) . Norwood , NJ : Ablex .

Baillargeon , R ., and Graber , M . ( 1987 ) . Where is the rabbit ? 5 . 5 - month - old

infants ' representation of the height of a hidden object . Cognitive Development , 2 ,

375 - 392 .

Balaban , M . T ., and Waxman , S . R . ( 1997 ) . Do words facilitate object catego -

rization in 9 - month - old infants ? Journal of Experimental Child Psychology , 64 ,

3 - 26 .

Baldwin , D . A . , Markman , E . M ., and Melartin , R . L . ( 1993 ) . Infants ' ability to

draw inferences about nonobvious object properties : Evidence from exploratory

play . Child Development , 64 , 711 - 728 .

Baldwin , D . A . , and Moses , L . J . ( 1994 ) . Early understanding of referential intent

and attentional focus : Evidence from language and emotion . In C . Lewis and

P . Mitchell ( Eds .) , Children 's early understanding of mind ( pp . 133 - 156 ) . Hove ,

England : Erlbaum .

Carey , S . ( 1994 ) . Does learning a language require the child to reconceptualize

the world ? Lingua , 92 , 143 - 167 .

Cohen , L . B . , and Younger , B . A . ( 1983 ) . Perceptual categorization in the infant .

In E . K . Scholnick ( Ed .) , New trends in conceptual representation ( pp . 197 - 200 ) .

Hillsdale , NJ : Erlbaum .

Dromi , E . ( 1987 ) . Early lexical development . Cambridge : Cambridge University

Press .
Sortals
andKinds 333

Eimas , P., and Quinn , P. ( 1994) . Studies on the formation of perceptually -based
basic-level categories in young infants . Child Development , 65, 903- 917.
Geach , P. ( 1957) . Mental acts: Their content and their objects. London : Routledge
and Kegan Paul .
Gibbard , A . ( 1975) . Contingent identity . Journal of Philosophical Logic , 4,
187 - 221 .

Gupta , A . ( 1980) . The logic of common nouns. New Haven , CT : Yale University
Press .

Hall , G . ( 1993). Basic-level individuals . Cognition , 40, 199- 221.

Hirsch , E . ( 1982). The concept of identity . New York : Oxford University Press.
Hood , B ., Willen , R ., and Driver , J. (in press) . Gaze following by 3-month -old
infants . Psychological Science.
Johnson , S., Slaughter , V ., and Carey , S. (in press) . Whose gaze will 12-month -
olds follow ? Developmental Science.
Kellman , P. ( 1984). Perception of three -dimensional form in infancy . Perception
and Psychophysics, 36, 353- 358.
Kellman , P. ( 1993). Kinematic foundations of infant visual perception . In
C . Granrud ( Ed .), Visual perception and cognition and infancy . Hillsdale , NJ :
Erlbaum .

Kripke , S. ( 1972) . Naming and necessity. Oxford : Blackwell .

Landau , B ., Smith , L . B ., and Jones, S. S. ( 1988) . The importance of shape in
early lexical learning . Cognitive Development , 3, 299- 321.
Macnamara , J. ( 1982). Namesfor things: A study of human learning . Cambridge ,
MA : MIT Press .

Macnamara , J. ( 1986) . A border dispute: The place of logic in psychology .

Cambridge , MA : MIT Press.
Macnamara , J., and Reyes, G . E . ( Eds.) . ( 1994) . The logical foundations of
cognition . Oxford : Oxford University Press.
Mandler , J., Bauer , P., and McDonough , E. ( 1991). Separating the sheep from the
goats : Differentiating global categories . Cognitive Psychology , 23, 263- 298.
Millikan , R . ( 1998). A common structure for concepts of individuals , stuffs, and
real kinds : More mama , more milk and more mouse . Behavioral and Brain Sciences ,
21 , 55 - 100 .

Oakes, L . M ., Madole , K ., and Cohen , L . B . ( 1991) . Infants ' object examining :

Habituation and categorization . Cognitive Development , 6, 377- 392.
Posner, M . I ., and Keele , S. W . ( 1968). On the genesis of abstract ideas. Journal of
Experimental Psychology , 77, 353- 363.
Prevor , M . ( 1997) . A new method for studying numerical representations in
infancy . Unpublished master 's thesis, Department of Psychology , New York
University .
 334 Carey and Xu

Quine , W . V . O . ( 1960 ) . Word and object . Cambridge , MA : MIT Press .

Quine , W . V . O . ( 1969 ). Ontological relativity and other essays . New York :

Columbia Universit ) T Press .

Quine , W . V . O . ( 1974 ) . The roots of reference . New York : Columbia University

Press .

Quinn , P . C . ( 1987 ) . The categorical representation of visual pattern information

by young infants . Cognition , 27 , 145 - 179 .

Quinn , P . C . , Eimas , P . , and Rosenkrantz , S . L . ( 1993 ) . Evidence for representa -

tions of perceptually similar natural categories by 3- and 4 - month - old infants .

Perception , 22 , 265 - 288 .

Ruff , H . A . ( 1986 ) . Components of attention during infants ' manipulative explo -

ration . Child Development , 57 , 105 - 114 .

Sabbagh , M . A . , and Baldwin , D . A . ( 1997 ) . Infants ' ability to distinguish objects

on the basis of feature information and form - function type correspondences .

Poster presented at the biennial meeting of the Society for Research in Child

Development , Washington , DC .

Scaife , J . F . , and Bruner , J . S . ( 1975 ) . The capacity for joint visual attention in the
infant . Nature , 253 , 256 - 257 .

Spe1ke , E . S . ( 1990 ) . Principles of object perception . Cognitive Science , 14 , 29 - 56 .

Spelke , E . S . , Breinlinger , K . , Macomber , J . , and Jacobson , K . ( 1992 ) . Origins of

knowledge . Psychological Revie }1-l, 99 , 605 - 632 .

Spelke , E . S ., Kestenbaum , R ., Simons , D . J ., and Wein , D . ( 1995 ). Spatio -

temporal continuity , smoothness of motion and object identity in infancy . British

Journal of Developmental Psychology , 13 , 113 - 142 .

VandeWalle, G., Prevor , M., andCarey , S. (1998

). Theuseof kinddistinctions
for objectindividuation : Evidence from reaching . Manuscript submitted for
publication
.
Vygotsky, L. S. (1962
). Thought andlanguage . Cambridge, MA: MIT Press .
Waxman , S. R., andBalaban , M. T. (1996). Ursinesandfelines: Novelwords
support object categorization in 9 - month - old infants . Paper presented at the

International Conference on Infant Studies , April , Providence , RI .

Waxman , S . R . , and Markow , D . B . ( 1995 ) . Words as invitations to form

categories : Evidence from 12 - to 13 - month - old infants . Cognitive Psychology , 29 ,

257 - 302 .

Wiggins , D . ( 1980 ) . Sameness and substance . Oxford : Blackwell .

Xu , F . ( 1997 ) . Distinct labels may provide pointers to distinct sortals in 9 - month -

old infants . Paper presented at the 22nd Boston University Conference on Lan -

guage Development .

Xu , F . ( 1997 ) . From Lot ' s wife to a pillar of salt : Evidence that physical object is a

sortal concept . Mind and Language , 12 ( 3 / 4 ) , 365 - 392 .

Sortals and Kinds 335

Xu , F ., and Carey , S. ( 1996). Infants ' metaphysics : The case of numerical identity .
Cognitive Psivchology, 30, 111- 153.
Xu , F ., Carey , S., and Quint , N . ( 1997) . Object individuation at 12 months : Shape
and other properties . Poster presented at the biennial meeting of the Society for
Research in Child Development , Washington , DC .
Xu , F ., Carey , S., and Welch , J. ( 1998). Infants ' ability to use object kind infor -
mation for object individuation . Manuscript submitted for publication .
Chapter 14
Semantics and the D . Geoffrey Hall
Acquisition of Proper Names

A Swedish court fined the parents of a young boy the equivalent of

$925 for not giving him a first name, at which point they named him
" Brfxxcxxmnpcccclllmmnprxvclmnckssqlbb1116
" (pronounced " Albin " ), which
the court has ruled to be unacceptable.
Article in the GeorgiaStraight, Vancouver, BC, December26, 1996

Having a proper name (e.g., " Tom ," " Dick ," or " Harry " ) is a funda -
mental property of being human , so basic that it has been called a uni -
versal right in a number of international declarations and covenants . For
example , Article 7-1 of Convention of the Rights of the Child , adopted
by the General Assembly of the United Nations in 1989, states that " [t]he
child shall be registered immediately after birth and shall have the right
from birth to a name" ( Valentine , Brennen , and Bredart 1996). As the
epigraph to this chapter reveals, countries may even impose restrictions
on what counts as an acceptable proper name , indicating the seriousness
with which humans treat the act of naming people .
According to ethnographers , all human societies confer proper names
upon their members (Alford 1987), and linguists have noted that proper
names crop up in all the world 's languages (Hockett 1966). Nonetheless ,
there are many striking crosscultural differences in practices associated
with naming people . For example , societies differ in terms of whether
people receive one or more first names in addition to one or more sur-
names; whether names are bestowed at birth or after a period of weeks,

This chapter was written with the support of a Natural Sciencesand Engineering
ResearchCouncil of Canada operating grant. I thank Paul Bloom, SusanGraham,
and Ray Jackendoff for their very helpful comments on previous versions. I am
immensely grateful"to John Macnamara for all that he taught me during the 10
years that I had the great fortune to have him as a teacherand friend.
 338 Hall

months, or even years; whether proper names are drawn from a large
and diverse set of possibilities or a small stock of conventional options;
whether names given near birth are used throughout life or changed at
important points (e.g., puberty); and whether proper names are used
freely in public discourseor kept private (Alford 1987). In addition, the
universality of proper names for humans does not mean that people are
the only bearers of these labels. Among other things, humans routinely
name animals (e.g., dogs), artifacts (e.g., boats), places(e.g., cities), media
products (e.g., movies), events (e.g., hurricanes), and institutions and
facilities (e.g., companiesand schools) (Hall , Veltkamp, and Turkel 1997;
Valentine, Brennen, and Bredart 1996).
This chapter examineshow children learn proper names. The task of
learning these expressionshas at least two components. First, children
must learn the mappings between individual proper names and their
appropriate interpretations in the particular languagethey are acquiring,
becausethe inventory of proper namesdiffers from languageto language.
For example, children learning English must somehow learn that the
word " Tom" should be interpreted as referring to an individual (i.e.,
Tom) in a specificmanner. Second, children must acquire knowledge of
how the syntax and morphology of their particular languagetreat proper
names, becauseproper names behave as a coherent class grammatically
within a language but do not accept the same affixes or appear in the
same serial positions within sentencescrosslinguistically. For example,
English learners must somehow discover that the word " Tom" shares
important grammatical featureswith " Mike " and " Bill " that it does not
sharewith " bike" or " mill ."
How do children learn the links between proper names and their
appropriate interpretations, and how do they gain knowledgeof the gram-
matical properties of expressionsof this class? The answer proposed in
this chapter is that children do so by exploiting implicit semanticknowl-
edgethat they possesswhen they begin to acquire language. This knowl-
edge consists of an understanding of the semantic definition of proper
names, along with an expectation that expressionsused under certain
conditions (related to this definition) should be interpreted as belonging to
this class. Children exploit this knowledge in the following way: As they
hear novel labels used by speakersto refer to things in the world , they
interpret some of them as satisfying conditions that imply they should be
taken to be proper names; as a result, children take them to be proper
names, assigningthem interpretations that are consistentwith the seman-
Semantics
andProper
Names 339

tic definition of expressionsof this class. As children learn more and

more proper namesin this way, they note how their languagetreats these
expressionsmorphologically and syntactically (e.g., how they combine
with affixesand how they interact with other words inside sentences ). As a
result, children come to learn the grammatical properties of proper names
in their language. In this way, semanticknowledge- knowledgeabout the
semanticsof any proper name, as well as knowledge about the conditions
under which speakersshould use expressionsof this class- provides chil-
dren with a way both to map individual proper namesonto their appro-
priate interpretations and to master the language-specific grammatical
properties of expressionsbelonging to this class.
Section 14.1 presentsa description of the grammatical and semantic
properties of proper names. Section 14.2 reviews previous research on
children's early understandingof theseexpressi.ons. Section 14.3 examines
in detail the hypothesis that children learn proper names by exploiting
knowledge of their semanticdefinition, along with an understandingthat
labels interpreted as satisfying certain conditions (related to this defini-
tion) should be taken to be members of this class. The section does two
things. First, it proposesfour conditions of using novel labels that learners
should see as implying that the labels could be interpreted as proper
names. Second, it reviews experimental evidence bearing on whether
young children favor interpreting a novel label as a proper name if any
one of these conditions is satisfied but not if it is violated. Section 14.4
concludesthe chapter with discussion.

14.1 What 's in a Proper Name?

14.1.1 Grammar
One way to identify proper names is through their morphology and
syntax. Even though the class of proper namesis universal, the environ-
ments in which these labels can appear within sentencesvary from lan-
guageto language. In English, proper namesare expressionsthat do not
normally take the plural morphemeand do not usually follow determiners,
quantifiers, or adjectives. For example, we may speak of " Tom" but not
" Toms" (unlesswe mean, for example, " people named Tom" ). In addi-
tion, we do not say " a Tom" or " severalToms" without leaving some-
thing to be specified (e.g., " a person named Tom" or " several people
named Tom" ). Finally , we rarely use adjectives to premodify proper
names (e.g., " happy Tom" ), restricting this practice to a pragmatically
340 Hall

limited set of contexts, such as in expressionsof approval or disapproval

(e.g., " Bad Fido !" ) or in utteranceswhere a name is usedto pick out a set
of objectsindividuated acrosssituations (e.g., " Big Tom is over there, and
little Tom is here" ) (Bloom 1990). Bloom (1990) has noted that the mor-
phological and syntactic properties of proper namesare the sameas those
of noun phrases, like " the fat dog" or " the shoe with the brown lace."
For this reason, Bloom (1990) has dubbed proper names (along with
pronouns, like " I " and " her" ) " lexical" noun phrases.
To see how an understanding of the morphological and syntactic
properties of English proper namescould help to identify theselabels in
speech, notice that theseproperties are distinct from those associatedwith
words from other lexical categories, such as count nouns (e.g., " dog" ).
Count nouns contrast with proper names in each of the ways noted
earlier. In English sentences , count nouns may appear marked with the
plural morpheme (e.g., " dog" and " dogs" ); they may also be precededby
detenniners(e.g., " a dog" ), discretequantifiers (e.g., " severaldogs" ), and
adjectives (e.g., " big dogs" ). In sum, proper names belong to a discrete
class in tenns of their grammatical properties (i.e., to the class of noun
phrases), and the properties of this classcontrast with those of the classof
count nouns (as well as with those of other grammatical classes ).

14.1.2 Semantics
Another way to identify proper namesis via their semantics. Unlike their
morphological and syntactic properties, which differ acrosslanguages, the
semanticproperties of proper namesappear to be universal (e.g., Kripke
1980; Macnamara 1986; Macnamara and Reyes 1994). Specifically, a
proper name is an expressionthat refers to an individual (in a kind), in
all the situations in which that individual appears, regardlessof the con-
ditions under which it is used. In this way, proper names contrast with
expressionsfrom other classes , such as count nouns (which refer to kinds
of individuals; e.g., " person," " dog" ). For example, the proper name
" Bill Clinton " is a label that refersto the individual , Bill Clinton, and can
be used to pick him out in all the contexts in which he figures. These
might be real situations (e.g., those associatedwith his childhood in
Arkansas or his current adulthood serving as president of the United
States), or they might be counterfactual (e.g., those associatedwith a
childhood spent in Iceland, had he moved there when he was a baby, or
an adulthood working as a veterinarian, had he chosennot to becomea
politician). An important consequenceof this fact is that proper names
Semantics
andProperNames 341

are not synonyms for definite descriptions . For example , the proper name
" Bill Clinton " cannot be defined by the description " the president of the
United States raised in Arkansas ," even though that description (in fact )
picks out the individual , Bill Clinton , uniquely . If a proper name were
simply synonymous with such a description , then it would be impossible
to account for the intuition that the individual can be picked out using
that proper name in situations where the description does not apply (e.g.,
the intuition that " Bill Clinton " can be used to pick out Bill Clinton in the
situation in which he had chosen to treat sick animals for a career and had

lived his whole life in Reykjavik ) (see Kripke 1980; for different versions
of the descriptivist view of proper names, see Frege 1892; Russell 1905;
Searle 1958) .
" Bill Clinton " also picks out the same individual regardless of who
happens to be speaking and regardless of the context in which that person
happens to be situated . Of course, there may be more than one person
named " Bill Clinton ," and so there is a sense in which the interpretation
of a proper name does depend upon the occasion of its use (see Barwise
and Perry 1983). Yet the context -specificity associated with interpreting
proper names seems to stem more from the fact that there are not enough
of them to go around than from any feature of their semantics . To ap -
preciate this argument , note the contrast between proper names and other
noun phrases, such as pronouns (e.g., " I " ) or definite descriptions (e.g.,
" the president of the United States" ) . These other expressions also can be
used to refer to individuals , even to Bill Clinton , but they require that the
context be specified in order for their referents to be fixed . For example ,
the referent of the pronoun " I " varies in a systematic way with the situa-
tion (e.g., who happens to be speaking) . Similarly , a definite description
like " the president of the United States" depends crucially upon the con -
text in which it is uttered for its referent to be identified (e.g., if the con -
versation were about the Civil War , the referent might be Lincoln ; if it
were about Watergate , Nixon ; or if it were about Whitewater , Clinton )
(see Macnamara and Reyes 1994).
Finally , a proper name refers to an individual that is typed by a kind
of individual . (In languages like English , kinds of individual are named
by count nouns - for example , " person ," " dog ." ) Macnamara ( 1986;
Macnamara and Reyes 1994) has argued that a proper name needs the
support of such a kind in order to specify the individual and in order to
account for the tracing of that individual 's identity across situations .
Consider pointing to a man who is sitting in the Oval Office in the White
342 Hall

House and saying , " That is Bill Clinton ." What is the " that " that the
proper name " Bill Clinton " is meant to name ? The answer is a person .
Yet without a kind (e.g., PERSON ) to make this specification , it is im -
possible to know whether the name should pick out Bill Clinton 's entire
body , or just his visible surface, or the set of molecules of which he is
presently constituted , or whether to include his clothing and the pen in his
hand . A kind serves to tell which individual a proper name designates. In
addition , in order for " Bill Clinton " to function as referring to the same
person when it is used to pick out both a little boy in Arkansas and a
president in the White House , there must be something to provide for the
identity of the boy with the president . This something is, again , a kind like
PERSON . By supporting statements in which the name " Bill Clinton "
appears, a kind enables a proper name to do its job of picking out one
persisting individual across these situations (e.g., the 2-year-old and 51-
year-old bearers of the name " Bill Clinton " ) . In sum, proper names form
a discrete semantic class, distinct even from other noun phrases, as
expressions that refer to individuals (typed by kinds ) in all the situations
in which they appear , regardless of the context .

14.2 What Children Know about Proper Names: Previous Research

Children first produce proper names very early in development (e.g.,

Nelson 1973; Macnamara 1982). This fact raises the question of what
they understand about these labels near the outset of language acquisi -
tion . Of the several studies that have addressed this question , some have
focused on young children 's productions of these expressions in early
speech, and others have explored children 's appreciation of links between
the syntax and semantics of these labels .
First , children as young as about 1 1/ 2 years of age appear to impose
appropriate restrictions on the sentence environments in which proper
names can appear . For example , they seem to know that proper names
cannot follow adjectives . Bloom ( 1990) studied the spontaneous speech of
English -learning children ranging in age from 18 to 34 months . Focusing
on a set of 43 adjectives (e.g., " big ," " happy ," " red" ), Bloom found that
even the youngest child 's speech appeared to respect a restriction against
using these words to premodify proper names. For example , children
rarely said things like " big Tom ." In contrast , the same children did not
show this restriction when producing these adjectives in conjunction with
count nouns , commonly and appropriately saying things like " big dog ."
Semanticsand Proper Names 343

Second , 2 -year - old children seem to know the relation between some of
the syntactic and semantic properties of proper names. Specifically , they
appear to understand that a novel word applied to an object in a proper
name syntactic context (e.g., " This is X " ) refers to an individual . In a
seminal study , Katz , Baker , and Macnamara ( 1974; Macnamara 1982)
showed two groups of children a pair of dolls differing in hair color . One
group heard a novel label presented syntactically as a proper name (e.g.,
" This is ZA V " ) for one of the dolls . The children 's task was then to
choose one of the two dolls to perform requested actions formulated using
the novel word (e.g., " Can you give me ZA V ?" ; " Can you show your
mother ZA V ?" ) . The dependent measure was the percentage of actions
performed with the named doll . The main finding was that children (girls
as young as 17 months ; boys as young as 27 months ) showed a strong
tendency to select the named doll in response to the requests: they chose it
about three quarters of the time . This result suggests that they typically
took the label to refer to an individual , consistent with their having
interpreted it as a proper name .
To demonstrate that children do not believe that all types of words
refer to individuals , Katz , Baker , and Macnamara ( 1974) presented the
same label for the same doll to a second group of children . For this group ,
however , they modeled the label in a sentence context that was appropri -
ate for a count noun (e.g., " This is a ZA V " ) . Count nouns , recall , do not
refer to individuals ; they refer to kinds of individuals . These children
heard requests to perform actions formulated in sentences that supported
a count noun interpretation of the label (e.g., " Can you give me a ZA V ?" ;
" Can you show your mother a ZA V ?" ) . This group showed no tendency
to favor the named doll in performing the actions : they chose it only
about half the time ; the remainder of the time , they picked the other
doll . This outcome is consistent with the possibility that these children
interpreted the word as naming the kind , DOLL , or the particular kind of
doll . (It is also, of course, consistent with the possibility that children were
simply guessing.) Children thus seemed to expect that a word named an
individual only if it was modeled in a proper name sentence context , not if
it appeared in a count noun sentence environment .
In another part of the study , Katz , Baker , and Macnamara (1974)
labeled one of two differently colored blocks , rather than one of two dolls ,
for two other groups of children . These children selected the named block
only about half the time in response to the experimenter 's requests,
regardless of whether the experimenters modeled the novel word as a
344 Hall

proper name (as they did for one group) or a count noun (as they did for
the other group). These results are consistent with the possibility that
children interpreted the label as naming the kind, BLOCK , or the specific
kind of block (as well as with the possibility that they guessed ) . The dif -
ferencebetweenchildren's interpretation of an intended proper name for
a doll and their interpretation of the sameword for a block suggeststhat
their willingnessto treat a word as a name for an individual dependednot
only on the syntactic context in which the word appeared, but also on the
specifickind of object (or properties of the object) being labeled.
To follow up the study by Katz, Baker, and Macnamara (1974; Mac-
namara 1982), Gelman and Taylor (1984) conducted a similar study but
changedthe method in severalways. They testedgroups containing both
boys and girls; children's averageage was about 2 1/2 years. They used
unfamiliar (previously unlabeled) toys as target stimuli, two surrogatesof
animate objects (animal-like stuffed toys) and two inanimate objects
(blocklike plastic toys). In addition, they presentedchildren with all four
objects at once. Half the time, the animals servedas target stimuli (i.e.,
one of them receiveda label modeled as a proper name or a count noun)
and the blocklike toys were distractors. The remainder of the time, the
blocklike toys played the role of target stimuli (i .e., one of them received
the label modeled as either a proper name or a count noun) and the ani-
mals served as the distractors. Adding two distractors to the two target
stimuli in the array of object choices enabled Gelman and Taylor to do
something that Katz, Baker, and Macnamara could not: clearly distin-
guish guessing (in which case, children should have chosen randomly
among all four objects) from other, more systematicinterpretations, such
as a kind of object (in which case, children should have selectedrandomly
betweenthe two target stimuli).
Gelman and Taylor 's main result replicated that of Katz, Baker, and
Macnamara. They observed the strongest tendency to select the named
toy in the group that heard the label modeled syntactically as a proper
name for an animal-like toy. This result suggeststhat children in that
group took the word to name that individual . In addition, Gelman and
Taylor found evidencethat children interpreted the novel word modeled
syntactically as a count noun, for either an animal-like or a blocklike
object, as naming a kind of object, becausechildren in thesegroups chose
randomly betweenthe two target stimuli, ignoring the two distractors. In
fact, children tended to selectone of the distractors only if they heard a
label that was presentedsyntactically as a proper name and the named
Semanticsand Proper Names 345

object was blocklike . In that group , some children appeared to forget the
act of labeling and selected exclusively one of the animal -like distractors ,
as if they interpreted the label as referring to that individual . Other chil -
dren in this group chose randomly between the two blocks , as if they took
the label to refer to a kind , such a BLOCK . Like the children in Katz ,
Baker , and Macnamara 's study , the children in Gelman and Taylor 's
study thus appeared to show an understanding of a link between proper
name syntax (e.g., " This is X " ) and proper name semantics (e.g., " X "
refers to an individual ), but only for some kinds of objects (or objects with
certain properties ) .
Liittschwager and Markman ( 1993) clarified an interpretive issue raised
by the results of the preceding studies. They argued that the earlier
findings do not , in fact , unambiguously show that children think that a
novel word modeled in the sentence context " This is X " is a proper name

when it is applied to a surrogate of an animate object . They noted that

children 's strong tendency to select the named doll (Katz , Baker , and
Macnamara 1974) or the named animal -like toy (Gelman and Taylor
1984) after learning the word for either of these objects might , as claimed ,
reflect an interpretation of the label as a proper name referring to the
named individual (e.g., " This is Sue" ) . However , they pointed out that
children could also have interpreted the word as an adjective referring to a
property of the named object (e.g., " This is blonde " ), because the named
object always had a salient property (e.g., hair color ) that distinguished it
from the other object of the same kind .
To address this issue, Liittschwager and Markman introduced a group
of 3-year-old children to an object (an animate surrogate , such as a stuffed
bear , or an inanimate , such as a shoe) with a distinctive property (e.g., the
bear was wearing a bib ). They labeled it with a novel word modeled
inside the sentence frame used in previous research to imply a proper
name interpretation (e.g., " This is DAX " ) . Next they moved the object to
a new location , removed its distinctive property (e.g., took the bib off the
bear), and brought out another identical -looking object that lacked the
distinctive property (e.g., another bear without a bib ) . They placed this
new object where the original object had been located at the beginning of
the task . They now asked the children , " Where 's DAX ?"
The authors argued that children who thought the label named a
property of the originally named object should have chosen either of the
two objects with equal probability . For example , they might have thought
the word named the distinctive property of the named object (e.g., bib -
 346 Hall

wearing); neither object now had that property, and so children should
have chosenrandomly betweenthe objects. Alternatively, children might
have taken the word to name some other property of the original object
(e.g., its shapeor color); both objectsnow had the sameproperties, and so
again children should have chosen either object. However, if children
thought that the label named the original object as an individual , then
they should have picked it , despitethe loss of its distinctive property, the
changein its location, and the fact that it was indistinguishable from the
new 0bject.
Liittschwager and Markman found that 3-year-olds did show a strong
tendency to pick the originally named object as the referent of the novel
word; subsequentanalysessuggestedthat this tendencywas specificto the
animate surrogatesthey used (i.e., bears and trolls) and did not extend
to the inanimates they used (i .e., shoesand baby bottles). These results
suggestthat children did, as previously claimed, interpret the novel word
modeled in the sentencecontext " This is X " for an animate surrogate
as naming an individual . Confirming that children do not treat words
modeled in all sentencecontexts in this way, the authors also found that
another group of 3-year-olds who heard the samenovel word modeled in
a count noun sentential context (e.g., " This is a DAX ," followed by the
test question " Where's a DAX ?" ) showedno preferencefor the originally
named object, regardlessof whether it was animate or inanimate, consis-
tent with their having interpreted the label as naming a kind of object (or,
of course, with their having guessed ).
To summarize: By the time children are 2 or 3 years old, they possessat
least some elementsof a mature understanding of proper names. Two-
year-olds produce proper names in appropriate sentential contexts (i .e.,
not following adjectives) (Bloom 1990). In addition, 2-year-olds and 3-
year-olds appear to interpret a novel word modeled in a sentential envi-
ronment appropriate for proper names (i .e., " This is X " ) and applied to
an animate surrogate object (a doll or an animal) as referring to an indi-
vidual (Gelman and Taylor 1984; Katz, Baker, and Macnamara 1974;
Liittschwager and Markman 1993; Macnamara 1982).

14.3 Using SemanticKnowledgeto Learn Proper Names

How do young children acquiretheir early understandingof proper names?

At least two types of account could be proposed. One is that children first
isolate proper names in speechvia their grammatical properties, inde-
Semantics
andProperNames 347

pendent of their semantics. For example, children might engagein a dis-

tributional analysisof the speechthey hear. They might note which affixes
can appear attached to the expressionsthat are proper namesin the adult
language and where theseexpressionscan occur inside sentencesin their
language(e.g., they might note what absolute positions they can occupy,
as well as what words can precedeand follow them). Gradually, children
might cluster these expressionsinto a distinct class on this basis. They
might then (somehow) " abstract" a formal class out of this initially dis-
tributionally defined one. By then noting the properties of the referentsof
the labels assignedto this class, children might come to tie expressions
from this classto the set of semanticproperties that define proper names.
Children might even be able to identify these labels as members of the
classof proper names(noun phrases), if they also knew that it was proper
names(noun phrases) that had thesesemanticfeatures.
This type of account runs into problems, however. One concern stems
from the fact that, as Bloom (1990) noted, proper names share their
grammatical properties with other noun phrases (including pronouns).
This fact complicates children's task of identifying proper names in
speechsolely on the basis of their morphological and syntactic features
(unless there is a set of sentencecontexts that are reserved for proper
namesand do not acceptother noun phrases, at least in speechaddressed
to children- Oshima-Takane (this volume) suggeststhat the vocative
may be such a context). In addition, unconstraineddistributional learning
procedures for acquiring lexical categories have been criticized on a
number of grounds. For example, it is unclear whether such procedures
can work (e.g., whether children have enough memory or computational
resources ) , given that convergenceupon the appropriate classdistinctions
would require children to sort through and keep track of a staggering
number of contrasting sentences(e.g., Bloom 1994; Gordon 1985; Pinker
1994). Moreover, the complexity of such procedureswould suggestthat
learnerswho used them would experiencea protracted (and error-filled)
period of learning. Yet, as already noted, very young children (i.e., 2-
year-olds) already appear to know (at least some of) the syntactic prop-
erties of English proper names (Bloom 1990) and to understand links
betweentheir syntax and their semantics(Gelman and Taylor 1984; Katz,
Baker, and Macnamara 1974; Macnamara 1982).
The precedingproblems suggestthe need for a different type of account
of how children learn proper names. Perhaps, instead of first identifying
proper namesin speechvia their grammatical properties (e.g., through a
348 Hall

distributionalleaming procedure ), young children learn proper names by

exploiting implicit knowledge about their semantics. Where would this
semantic knowledge come from ? Whereas the grammatical properties of
proper names vary from language to language and so would have to be
learned , the semantic properties of these expressions are universal and so
could be known innately . Thus , children could initially approach the task
of language acquisition with an understanding of the semantic definition
of proper names, along with knowledge of special conditions of using
labels (related to this definition ) that imply they should be interpreted as
members of this class of expression. If children did so, then the burden of
learning proper names would be much more manageable than the one
involved in acquiring them distributionally via their grammatical proper -
ties. Equipped with such knowledge , children would , then they first begin
to learn language , have a way to make mappings between individual
proper names and their appropriate interpretations . For example , if chil -
dren interpreted an expression as being used under certain conditions
(related to the semantics of proper names), then they would simply con-
strue it as a proper name , assigning it the semantic properties of labels
belonging to this class. Learning the grammatical properties of the
expressions assigned to this class would be relatively straightforward ,
because children would need to perform a distributional analysis on only
a small subset of the expressions in their language (i .e., those picked out
as proper names because speakers used them under certain conditions ) .
No prolonged period of error -ridden learning would be expected (see
Bloom 1994; Pinker 1984) .
Under what conditions should speakers use a novel label in order for
learners to see it as potentially being a proper name (i .e., as an expression
that refers to an individual in a kind in all situations in which the indi -

vidual figures , regardless of the context )? Here are four suggestions,

inspired by the semantic definition of proper names: ( 1) it should be used to
refer to only one object (rather than more than one object ); (2) it should
be used to refer to an object for which learners do not already know a
proper name (rather than one for which they do); (3) it should be used to
refer to an object seen as important as an individual rather than one not
seen in this way ); and (4) it should be used to refer to an object for which
learners know the kind of individual (rather than one for which they do
not ) .
The remainder of this section is devoted to two tasks: ( 1) explaining
)1'hy learners should see labels used under each of the four conditions as
Semantics
andProperNames 349

potentially being proper names, and (2) providing evidencebearing on

whether young children who hear new labels used under theseconditions
are inclined to interpret them as proper names (e.g., Hall 1991, 1993b,
1994a,c, 1996; Hall and Graham 1997; Liittschwager and Markman
1993). The evidence comes from experimental studies involving one of
two types of design. One type involved varying the conditions under
which a speakerused a novel word (modeled as a proper name) to pick
out things in the world . One group of children heard the label used in a
way that satisfied one of the conditions hypothesized to imply that it
could be a proper name; the other group heard the same label used in
a way that violated the condition. The prediction was that those who
heard the label usedconsistentlywith the condition would be more likely
to interpret it as a proper name than those who heard it used incon-
sistently. The secondtype of design was more indirect. It involved pre-
senting children with a novel word (modeled as a proper name) and then
asking them to chooseone of two objects as its referent. Choosing one
object satisfiedone of the conditions proposedto imply that a label could
be a proper name; choosing the other object violated it . The prediction
was that children would select the object that was consistent with the
condition.

14.3.1 Only One Object per Proper Name

One condition of using a novel label that learners should seeas implying
that it could be a proper name is that it serveto refer to only one object.
Recall that a proper name's semantic role is to refer to an individual .
Thus, learners should be more likely to view a novel word as a proper
name if it is given to only one object than if it is given to two different
objects. Notice that proper names contrast with words from many other
lexical categories in terms of this semantic property. F or example, an
adjective is a word whose semantic role is to refer to a property of an
object, and so it can be used to pick out any number of objects that
happen to have that property (seeHall and Moore 1997; Hall , Waxman,
and Hurwitz 1993; Taylor and Gelman 1988). As a result, whereaslearners
should view the existenceof two dogs both labeled " Fido" as inconsistent
with the possibility that " Fido" is a proper name, they should not find the
existenceof two dogs both labeled " striped" to be inconsistent with the
possibility that " striped" is an adjective.
Do children assumethat a label usedto refer to only one object can be
interpreted as a proper name, but a label used to refer to more than one
350 Hall

object cannot ? On the one hand , some evidence suggests the answer might
be no . In many cultures , including our own , there are far fewer proper
names than there are proper name bearers. For example , Anderson (1977,
cited in Alford 1987) estimated that there are 6 million men named
" John " and 4 million women named " Mary " in the United States. On the
other hand , anecdotal evidence suggests that children do make the
assumption . For example , Macnamara ( 1982) made the following obser-
vations about his son , Kieran :

For a time he seemed to assume that proper names were uniquely paired with in -
dividuals . The reason for believing so is that when he met his first " doppelganger "
he refused to accept the name . He was at the time [ 16 months and 13 days old ] and
had met a cousin of his , Lisa . He was then introduced to a girl of about the same
age as Lisa also called Lisa . They played for half an hour , yet , most unusual for
him , he refused to say her name , no matter how often anyone said it or urged him .
Shortly after he met three girls all named Aimee and he accepted the name for all
three . (p . 28)

Here is another anecdote reported by the mother of a boy , Matthew ,

when he was 20 months and 21 days old :

He has a friend Rebecca at the sitters ' that he loves to play with very much . He
calls her Becca because it ' s easier for him to say . I have a friend who has a three -
month old daughter named Rebecca, and Matthew will not call her by her name .
I introduced him to her saying that her name was Rebecca, and he said " No
Becca!" He has since been calling her " baby ."

The preceding anecdotes suggest that young children initially do expect

that a proper name can have only one bearer, and thus , that in order for a
novel word to be interpreted as a proper name, it should be used to refer
to onl y one object . This anecdotal evidence suggested the need for an
experiment to determine more systematically whether children have this
expectation . To meet this need, I conducted a study (Hall 1996) that
compared how one group of young children interpreted a potential proper
name used to refer to one object with how another group interpreted the
same word used to refer to two objects .
In the study , two groups of 4-year-olds heard a novel word modeled in
a sentence context (e.g., " This dog is ZA VY " ) that supported a proper
name interpretation (e.g., " ZA VY " could refer to an individual , like
" Fido " ) but that also permitted an adjective interpretation (e.g., " ZA VY "
could refer to a property , like " striped " ) . One group heard this label
applied to a drawing of a single object with a salient novel property (e.g.,
a dog with a novel striped pattern ). I expected that children in this group
Semantics
andProperNames 351

"This dog is ZA VY1J

" This dog is ZAVY and this dog is ZAVY1 '

Figure14.1
Sampleschematic
targetstimulifrom Hall 1996

would interpret the expression as a proper name (e.g., " Fido " ), given that
it was applied to only one object , given that the syntax of the sentence
allowed a proper name interpretation , and given that the object was an
animate surrogate . The other group heard the same label applied sepa-
rately to drawing of two different objects (e.g., two dogs with the same
novel striped pattern ). Here I expected that children would be more
reluctant to interpret the novel word as a proper name, because it was
applied to two objects . Instead , I suspected that they would take the word
to be an adjective (e.g., " striped " ) naming a property , because the word
could also be disambiguated in this way . (See figure 14.1.)
To assess the interpretation of the novel label , I showed children in
both groups a set of object drawings , including ( 1) the labeled object or
objects (e.g., the original striped dog or dogs), (2) an object of the same
kind that lacked the named object 's (or objects') salient property (e.g., a
different stripeless dog), and (3) an object of a different kind that pos-
sessed the labeled object 's (or objects') salient property (e.g., a striped
umbrella ). I then asked whether children would extend the novel label to
each of these objects in turn (e.g., " Is this dog ZA VY ?" , " Is this umbrella
352 Hall

ZA VY ?" ) . I intentionally phrased the questions so that the word would

remain ambiguous between a proper name (e.g., " Is this dog Fido ?" ) and
an adjective (e.g., " Is this dog striped ?" ) . I expected that children who
interpreted the word as a proper name would agree that it could apply to
only the labeled object or objects. However , I predicted that those who
took the word to be an adjective would assent to its applying to both the
labeled object (s) and the object from a different kind with the same salient
property as the labeled object (s) .
As predicted , children who heard the label for only one object showed a
strong tendency to interpret it as a proper name . In contrast , children who
heard the same label for two different objects tended to interpret it as an
adjective . These results suggest that children readily interpret a novel
word as a proper name if the word is applied to one object but are more
reluctant to do so if it is applied to two objects .
In the study reported in Hall 1996, I also demonstrated that 4-year-olds
understand that one proper name sometimes can apply to two different
objects (as they must do , of course, in learning a language like English ).
Two different groups of children heard the same novel word applied to
the same object or pair of objects, but the sentences used to introduce the
label made the word unambiguously a proper name . For example , instead
of introducing the word in the ambiguous sentence " This dog is ZA VY ,"
I now modeled it in three additional sentences that called clearly for a
proper name construal (e.g., " The name of this dog is ZA VY . This dog is
called ZA VY . This dog 's name is ZA VY " ) . In both groups , children more
readily made a proper name interpretation . Taken together , these results
thus suggest that children 's default assumption (their interpretation given
no unambiguous information to the contrary ) is that a novel word should
be used to refer to only one object if it is to be interpreted as a proper
name . The findings also indicate that children will relinquish this
assumption in the fact of unambiguous evidence (e.g., clear linguistic
cues) to do so.

14.3.2 Only One Proper Name per Object

A second condition of using a novel label that learners should see as im -
plying that it could be a proper name is that it serve to refer to an object
that has not already been assigned a proper name. The semantic function
of a proper name is to designate an individual " rigidly " : to refer to the
individual in all the situations (real or counterfactual ) in which it appears,
and to pick out the individual regardless of the context of use. A conse-
Semantics
andProper
Names 353

quenceof this fact is that two proper namesfor one individual would both
have the same interpretation. Given independent evidencethat learners
do not like to acquire more than one word with the sameinterpretation
(e.g., they tend to assumethat words contrast in their interpretations;
Clark 1987), they should be more likely to seea novel word as a proper
name if it is used to refer to an object that does not already have one of
theselabels than if it is used to refer to an object that does. For example,
they should view the existenceof one dog labeled both " Fido " and " Ro-
ver" to be inconsistent with the possibility that both these words are
proper names.
Notice that the semanticredundancyassociatedwith two proper names
for one individual would not be true in the case of multiple words from
other lexical categoriesfor the sameindividual . For example, the seman-
tic function of an adjective is to refer to a property of an individual , and
any individual has many properties. As a result, learnersshould be willing
to interpret a novel label as an adjective if it is applied to an object that
has already been labeled with another adjective. For example, learners
should view the existence of one dog designated both " spotted" and
" furry " as consistent with the possibility that both these words are
adjectives. In addition, becauseproper names and adjectiveshave differ-
ent semantics, learnersshould find the existenceof two words for the same
object, one from each of theseclasses , to be consistentwith thesewords'
belonging to thesetwo classes . For example, they should seethe existence
of one dog labeledboth " Fido" and " furry " to be consistentwith the pos-
sibility that thesewords are a proper name and an adjective, respectively.
Do children assumethat a label used to refer to an object can be inter-
preted as a proper name if the object lacks a proper name, but not if it
already has one? Some evidencesuggeststhe answermight be no. There
are societiesin which different people refer to the sameperson by using
entirely different proper names (Valentine, Brennen, and Bredart 1996),
and in our own culture, different people may refer to the sameindividual
using distinct variants of the individual 's name (e.g., in decreasingfor-
mality , we might use a surnamealone, a surnameplus first name, a first
name, and somediminutive form ofa first name) (Alford 1987). However,
some recent experimental evidencesuggeststhat young children do make
the assumption. One way to test for its existencewould be to compare
how two groups of children interpret a novel label modeled as a proper
name, using a design like that describedin Hall 1996. One group would
hear the new label used to refer to an object previously labeled with
354 Hall

another proper name ; the other group would hear it used to refer to the
same object previously labeled not with a proper name but with a word
from a different class. A proper name interpretation of the new label
should be more likely if the referent object had no previous proper name
than if it did . Susan Graham and I took another , more indirect approach
(Hall and Graham , in press). We presented children with a novel label
modeled as a proper name, along with a choice of two referents , one of
which had been previously labeled either with a proper name or with a
word from a different class . If children assumed that a label could be a

proper name if it was applied to an object that lacked a proper name (but
not if it was applied to one that had one), then we predicted that they
would choose the previously labeled object only if the previous label was
not a proper name .

In one of the studies, we taught one group of4 -year-old children a novel
word modeled unambiguously as a proper name for a familiar object
(e.g., a stuffed dog). For example , we said, " This dog is named ZA VY ."
We taught another group of 4-year-olds the same word for the same
object but modeled it unambiguously as an adjective . For example , we
said, " This dog is very ZA VY ." (See the top half of figure 14.2.) We then
brought out a second object that looked identical to the first (e.g., another
stuffed dog). We placed it to the side of , and several inches behind , the
first object , but still within the child 's easy reach. Now we provided a
second word modeled unambiguously as a proper name and asked chil -
dren to choose its referent . For example , we said, " Show me a dog that is
named DAXY ." (See the bottom half of figure 14.2.)
If children believe that a novel word can be taken as a proper name
only if it is used to refer to an object that has not already been labeled
with a proper name (e.g., one dog cannot be both " Fido " and " Rover " ),
then we predicted that those in the group who had learned a proper name
for the first object (dog A in figure 14.2) would select the second object
(dog B in figure 14.2) . In contrast , if children assume that a novel word
can be interpreted as a proper name if it is used to refer to an object that
either has or has not previously been labeled with an adjective (e.g., one
dog can be both " spotted " and " Rover " ), then we predicted that those in
the group who had learned an adjective for the first object would show no
strong tendency to select the second object . This is, in fact , what we
found . Children strongly favored the second object as the referent of the
second word (a proper name) if they had learned a proper name for the
first object , but not if they had learned an adjective .
356 Hall

an expression already assigned to that class. To rule out this alternative

we gave two other groups of 4-year-olds the same task , using the same
materials and procedure . As with the previous groups , we began by
teaching a novel word (e.g., " ZA VY " ) modeled as either a proper name
or an adjective for one object (e.g., a stuffed dog ; dog A in figure 14.2).
Again , we then introduced a second identical -looking object (e.g., another
stuffed dog ; dog B in figure 14.2), and we asked children to choose the
referent of a second word . Now , however , we presented this second word
as an adjective . For example , we said, " Can you show me a dog that is
very D AXY ?"
If children assume that a novel word can be interpreted as an adjective
if it is applied to an object that either has or has not already been labeled
with an adjective (e.g., one dog can be both " spotted " and " furry " ), then
we predicted that those who had heard an adjective for the first object
(e.g., dog A ) would show no strong tendency to select the second object
(e.g., dog B). In addition , if children believe that a novel word can be
interpreted as an adjective if it is applied to an object that either has or has
not previously been labeled with a proper name (e.g., one dog can be both
" Pido " and " furry " ), then we predicted that those who heard a proper
name for the first object would again have no clear bias to choose the
second object . The result again supported the predictions . Children did
not favor the second object as the referent of the second word (the adjec-
tive ) in either group . Children were somewhat more likely to select the
second object if they had heard an adjective for the first object than if they
had heard a proper name, but neither group selected the second object at
a level that exceeded what would be expected by chance. In addition ,
comparing these results with the results obtained when the second word
was a proper name, we found that children were less likely to select the
second object upon hearing a second adjective than upon hearing a
second proper name .
After they chose a referent for the second word (the adjective ), we again
asked children whether the label could also apply to the object they did
not choose. On those trials where children chose the second object , we
found no difference in their tendency to refuse to apply the second word
(the adjective ) to the first object , whether they had already learned a
proper name or an adjective for it . Moreover , in both groups , this ten-
dency was lower than it was in the group where children were asked about
a second proper name for the first object . Thus , children showed no
strong unwillingness to map an adjective onto an object that had already
Semantics
andProperNames 357

beenlabeledwith an adjective, and they also showedno clear tendencyto

refuse to accept an adjective for an object that already had been labeled
wi th an adjecti ve.
Taken together, the results of this study suggestthat preschoolersreject
assigningtwo proper namesto one object. Thus, the findings are consis-
tent with the hypothesis that children assumethat a novel label for an
object could be interpreted as a proper name if they lack a proper name
for it , but not if they already know one. The findings are also consistent
with the claim that this assumption does not extend to all word classes.
Specifically, children did not reject two adjectives for one object, sup-
porting the view that children are willing to permit a novel label to be
interpreted as an adjective for an object, regardless of whether they
already know another adjective for that object.

14.3.3 Objects Important as Individuals

A third condition of using a novel label that learners should see as
implying that it could be a proper nameis that it serveto refer to an object
that belongs to a kind (or has the properties) that makes that object
worthy of being referred to as an individual . The semantic function of a
proper name is to refer to an individual , without any restrictions on its
specifickind (or its specificproperties). However, learnersmay (indepen-
dently) construe only someindividuals as being important as individuals;
they may seeothers as more or less interchangeablemembers of a kind .
As a result, learnersshould view the useof a label to refer to an individual
considered significant as an individual to be more consistent with the
label's being a proper name than the use of the samelabel to refer to an
individual seento lack such significance.
Are children more likely to interpret a label as a proper name if it is
used to refer to an individual seento be important as an individual than
if it is used to refer to an individual seento lack such importance? The
answer depends upon which individuals matter to young children as
individuals. In previous work, researchershave proposed that people
believe that humans and other animate objects matter more as individ -
uals than many common artifacts (e.g., Macnamara 1982). Thus, children
might be more likely to interpret a novel word as a proper name for an
animate or animate surrogate than for an artifact. Liittschwager and
Markman (1993) recently obtained evidenceconsistent with this predic-
tion, using a two-object forced-choice procedure (as in the design of Hall
and Graham, in press).
358 Hall

Liittschwager and Markman ( 1993) asked a group of 3-year-olds to

choose the referent of a novel word modeled as a proper name (e.g.,
" Where 's BIF ?" ). Children had to choose between a novel animate sur-
rogate object and a novel inanimate object . (The authors do not report
what kinds of object these were.) Liittschwager and Markman found that
children chose the animate object significantly more often than would be
expected by chance. The authors asked another group of 3-year-olds to
choose the referent when the novel word was modeled syntactically as a
count noun (e.g., " Where 's a BIF ?" ) . These children chose the animate
surrogate object only as often as would be expected by chance. Moreover ,
children 's tendency to choose the animate object was (marginally ) sig-
nificantly higher if they heard the word modeled as a proper name than if
they heard it presented as a count noun (Woodward and Markman , in
press) . These findings provide evidence that children prefer to map a
proper name (but not a count noun ) onto an animate surrogate object
rather than onto an inanimate object . The results are thus consistent with
the claim that children assume that a novel word could be interpreted as a
proper name if it is applied to a surrogate of an animate object but not if
it is given to an inanimate object . On the assumption that children tend to
seen animates but not inanimates as important as individuals , the findings
are also consistent with the claim that children are more likely to interpret
a novel word as a proper name if it is applied to an individual seen to be
important as an individual than if it is applied to an individual seen to
lack such importance .
Several of the studies reviewed earlier in this chapter have also tested
the prediction that children are more likely to take a novel label as a
proper name if it is applied to an animate surrogate than if it is applied to
an artifact , by examining children 's interpretation of a novel label given
directly to an object (as in the design of Hall 1996). In those studies, recall ,
children showed a readiness to interpret a novel word as a proper name if
it was used to name a doll but not a block (Katz , Baker , and Macnamara
1974; Macnamara 1982); an animal -like monster but not a plastic block -
like toy (Gelman and Taylor 1984); and a bear or a troll but not a shoe
or a baby bottle (Liittschwager and Markman 1993). Thus , children have
been more likely to interpret a novel label as a proper name if the referent
is an animate surrogate than if it is an artifact . Again , on the assumption
that people and animates matter more to young children as individuals
than many artifacts , the results are consistent with the proposal that
young children are more likely to take a word as a proper name if it is
Semantics
andProperNames 359

used to refer to an individual seen as important as an individual than if it

is used to refer to an individual seen as lacking this importance .
I conducted a series of studies (Hall 1994c) to test further the hypothe -
sis that children are more likely to take a novel word as a proper name
if it is applied to an animate or animate surrogate than if it is applied
to an artifact . The method was again similar to the one described in
Hall 1996, in that it explored the interpretation children assigned to a
novel word used to refer directly to an object . Preschoolers saw a drawing
of a familiar object , an X . They then heard a novel word modeled in a
sentence frame that was ambiguous between a proper name and an
adjective (e.g., " This X is ZA VY " ) . Children heard the label applied to a
drawing of either an animal or an artifact with a salient novel property
(e.g., a striped pattern ). The question was whether children would inter -
pret the label as a proper name naming the individual (e.g., as a word like
" Tom " ) or as an adjective referring to the salient property (e.g., as a word
like " striped " ) .
To assess children 's interpretation of the novel label , I showed them
a set of object drawings , including ( 1) the labeled object (e.g., a striped
animal or artifact ), (2) an object of the same kind that lacked the labeled
object 's salient property (e.g., a different nonstriped animal or artifact of
the same kind ), and (3) an object of a different kind that possessed the
labeled object 's salient property (e.g., a striped animal or artifact of a
different kind ) . A puppet then pointed to each of the drawings , one at a
time , asking children whether they thought that the object (X ) could be
labeled with the new word (e.g., " Is this X ZA VY ?" ). The question
phrasing left the word ambiguous between a proper name (e.g., " Is this
X Tom ?" ) and an adjective (e.g., " Is this X striped ?" ) . I expected that
children who interpreted the word as a proper name would agree that it
could apply to only the named object . However , I thought that those who
construed the word as an adjective would assent to applying it to both the
labeled object and the object from a different kind that shared a salient
property with the labeled object .
The first two studies reported in Hall 1994c examined whether 4-year-
olds (experiment 1) and 3-year-olds (experiment 2) were more likely to
interpret the novel word as a proper name if the drawing was an animal
that was a typical pet (a dog , a bird , a cat, or a rabbit ) than if it was a
simple artifact (a shoe, a balloon , a cup, or a hat ). (See the top half of
figure 14.3.) As predicted , the groups who learned the words for the
petlike animals tended strongly to interpret them as proper names. In
Semanticsand Proper Names 361

Using the same procedure as in the previous experiments, I discovered

that the group of children who heard the words for the nonpet animals
were, in fact, less likely to interpret the labels as proper names than the
group in experiment 1 who heard the words for the petlike animals.
However, the group who learnedthe words for the complex artifacts were
as unwilling to interpret the labels as proper names as the group in
experiment 1 who learned the words for the simple artifacts. In addition,
I found that the group of children who learned the words for nonpet
animals were no more likely to make a proper name interpretation than
the group who learned the words for the complex artifacts. Thus, pre-
schoolersdid not seemto treat all animals as being equally suitable as
candidate recipients of proper names, although they appearedto seeall
artifacts (both simple and complex) as equally bad potential bearers of
theselabels.
Experiment 4 explored the basis of the differencebetween4-year-olds'
willingness to interpret a novel word as a proper name for a petlike ani-
mal (experiment 1) and their relative unwillingnessto do so for a nonpet
animal (experiment 3). One difference between pets and nonpets is that
the former are usually owned by people (although some nonpets may be
owned by people). I examinedwhether mentioning that the to-be-labeled
objects were possessedby the experimenterwould increasechildren's ten-
dency to interpret the labels for them as proper names. I used the same
drawings as in experiment 3. Now , however, before introducing the
ambiguous labels, I inserted an additional sentenceto indicate that I
owned the object, the X : " This is my X ."
The addition of this information about possessionincreasedchildren's
tendencyto interpret the labels as proper namesif the objectswere nonpet
animals, but not if they were complex artifacts. Children in experiment 4
were significantly more likely to interpret the words as proper names if
they were in the group for which the labeled objectswere nonpet animals
than if they were in the group for which the labeled objectswere complex
artifacts. Taken together, the results of these experiments thus suggest
that children's willingnessto interpret a novel word as a proper name for
an object is very strong if the object is an animal and if it is marked as
possessedby a person (either by being a typical pet or by having its own-
ership noted explicitly). Preschoolersdo not readily construea novel label
for an artifact as being a proper name. On the assumption that animate
objects (or objects with animate properties), especially those that are
marked as possessedby someone, are viewed as more important as
362 Hall

individuals than many common artifacts, the results suggestthat another

condition leading children to construe a novel word as a proper name is
that the labeled object be viewed as significant as an individual .
The results from this series of studies (Hall 1994c) and earlier ones
(e.g., Gelman and Taylor 1984; Katz, Baker, and Macnamara 1974;
Liittschwager and Markman 1993) do not imply that preschool children
cannot understand(or learn) a proper name in conjunction with an object
that is not animate (and possessed ). If children fully grasp the semantics
of proper names, then they should be able to comprehend(and acquire) a
proper name used in conjunction with anything that they can conceptu-
alize as an individual, including artifacts. In fact, many young children
likely do encounter(and understand) artifacts bearing proper names, such
as " Thomas" the Tank Engine or " Budgie" the Helicopter, although
theseartifacts are often anthropomorphized (e.g., they have human facial
features). In addition, Sorrentino (1997) has shown that 2-year-olds will
interpret a novel word as a proper name for an artifact if it is describedas
possessingmental states(but not if it is describedneutrally). However, the
finding that children (and adults) can learn and understand the use of
proper namesin conjunction with a wide range of individuals, including
artifacts, would not be inconsistent with the findings reported in Hall
1994cor the findings from earlier studies. When trying to identify these
labels in speech, children simply may be disposed, as a default, to seek
them out as the expressionsused to refer to those individuals that matter
as individuals, and (aside from people) these may be seento be objects
that are animate and possessedby humans.

14.3.4 Objects from Known Kinds

A final condition of using a novel label that learnersshould seeas imply-
ing that it could be a proper name is that it serveto refer to an object for
which they already know the kind of individual . Recall that, in order to
carry out its semantic role of referring to an individual , a proper name
needsthe support of such a kind , in order to specify that individual and
to trace its identit)T across situations. Becausekinds of individuals are
labeled by count nouns, learners should seea novel word (e.g., " This is
ZA V " ) as more likely to be a proper name if it is used to label an object
for which they know a count noun (e.g., a dog, for which the learner
knows the word " dog" ) than if it is usedto pick out one for which they do
not know a count noun (e.g., a novel monster, for which the child knows
no label and which the child has never seenbefore). (Seefigure 14.4.)
Semantics
andProper
Names 363

" This is ZA V '.

Figure14.4
Sampleschematic
targetstimulifrom Hall 1991

How should knowledge of a count noun for an object be assessed ? One

index of the presenceor absenceof such knowledge is the ability or
inability to producethe word for the object. However, becauseproduction
lags behind comprehensionin languagedevelopment, learnersmay know
(i.e., understand) count nouns for many objects before they produce the
words, especiallythose objects with which they have had some previous
encounters. Thus, simply equating children's having or lacking knowledge
of a count noun with their ability or inability to produce it may be
inappropriate. The ability to produce a count noun for an object does
seema reasonableindex of knowing a count noun for it ; however, a more
plausible index of not knowing a count noun for an object would be
the inability to produce a count noun for an object, along with a lack of
experiencewith the object and its kind (e.g., no previous encounterswith
objects of the kind, and so no opportunities to gain any understandingof
a count noun for the kind).
Is there any evidencebearing on the claim that children are more likely
to interpret an expressionas a proper name if it is used to refer to an
object for which they know a count noun than one for which they do not?
On the negative side, there is anecdotal evidencesuggestingthat learners
will sometimesproduce a proper name for an object before producing a
corresponding count noun. For example, Macnamara (1982) noted that
his son, Kieran, produced many proper namesfor people before produc-
ing a related count noun (such as " man," " person," or even " thing" or
" object" ). In addition, Kieran produced " Freddie" as a name for his dog
six weeks before he -produced the count noun " dog" (see also Nelson
1973). However, as noted earlier, simply failing to produce a count noun
364 Hall

for an object does not necessarilyimply that children lack knowledge of

the word, and so this evidenceis not conclusive.
On the positive side, there is other evidence suggestingthat children
have difficulty in learning proper names in conjunction with objects for
which they do not already produce a count noun. F or example, it is
commonly observedthat young children will extend a first word produced
for a parent, like " Daddy" or " Mommy ," from the labeled person to
other people (see Merriman 1986; Nelson 1973). One interpretation of
this phenomenonis that children misinterpret an intendedproper name as
referring to the kind of object (as a count noun), becausethey do not
already know a count noun for the kind, such as " man," " woman," or
" person." Again, however, this interpretation is subject to the concern
that failure to produce a count noun does not indicate that children nec-
essarily lack any knowledge of it . Another problem with the interpreta-
tion is that words like " Daddy" and " Mommy " are not unambiguous
proper names. For example, in speech to children, these words may
appear as either count nouns (e.g., " Wait till your daddy seesthat!" ) or
proper names(e.g., " Wait till Daddy seesthat!" ).
To obtain clearer evidence about the relevance of knowing a count
noun for an object to children's tendency to interpret a novel word as a
proper name, I conducted two experiments (Hall 1991). Both studies
compared how one group of children interpreted a novel word presented
as a proper name for a " familiar " object (i.e., one for which they pro-
duced a count noun) with how another group interpreted the sameword
for an " unfamiliar " object (i.e., one for which they did not produce a
count noun and with which they lacked any prior experience ).
In the first study, two groups of 2-year-olds heard a novel word mod-
eled syntactically as a proper name (e.g., " This is ZA V" ). One group
heard it for a familiar stuffed toy (a cat), and the other group heard it for
an unfamiliar stuffed toy (a novel monster). The toys were both surro-
gates of animate objects, becausechildren in previous researchwere less
willing to learn proper namesin associationwith other kinds of objects,
regardlessof their familiarity (e.g., blocks or unfamiliar blocklike toys).
I assessedfamiliarity by asking an independent group of slightly older
preschoolersto generatea count noun for the kinds of object. Children
regularly produced a count noun, such as " cat," for the cat, but they
generally failed to provide a count noun label for the novel monster (and
it is very unlikely that they had any prior experiencewith it , given that it
was an invented creature). After children in both groups heard the novel
Semanticsand Proper Names 365

label , I brought out three other stuffed animals . One was from the same
kind of object as the named one (either another cat or another monster of
the same kind ) but wore differently colored clothes . The other two were
distractors from different kinds . The task was similar to the one used by
Gelman and Taylor ( 1984) : to select one of the four toys in response to a
series of requests formulated using the new word (e.g., " Can you put ZA V
on top of your head?" ; " Can you point to ZA V ?" ). As in Gelman and
Taylor 's study , the main dependent measure was the percentage of trials
on which children chose the labeled object . A higher percentage of named
object choices suggested a stronger tendency to interpret the label as a
proper name for that individual .
Children in both familiar and unfamiliar groups showed , overall , a
strong tendency to select the named toy . This finding was not surprising ,
because both the familiar and unfamiliar labeled objects were animate
surrogates , and both were described with a word modeled syntactically
as a proper name . However , as predicted , children who learned the word
for the familiar object (the cat) were significantly more likely than those
who learned it for the unfamiliar object (the monster ) to select the named
object (i .e., to make a proper name interpretation ). Those who heard the
word for the unfamiliar object often selected the other object of the same
kind as the labeled object in response to the requests, suggesting that they
had misinterpreted the word as a name for the kind (as a count noun ). In
addition , children ' s spontaneous comments transcribed from tape record -
ings of the test sessions provided further evidence that those in the famil -
iar group were more likely than those in the unfamiliar group to interpret
the word as a proper name . In the familiar group , children sometimes
asked the experimenter what the unlabeled cat was " named " or " called ,"
suggesting that they had taken the word for the labeled object as a proper
name . In the unfamiliar group , children did not ask these questions ;
instead , they often made comments revealing that they had misconstrued
the intended proper name as a count noun . For example , after hearing the
intended proper name (e.g., " This is ZA V " ) for the labeled monster ,
children often pointed to the other monster of the same kind and
described it using the novel word , but they placed the word in sentence
contexts reserved for count nouns , such as " two ZA V s" or " another
ZA V ."

The second study included two new groups of 2-year-olds and used a
similar procedure , but different familiar (i .e., dogs) and unfamiliar (i .e.,
novel monsters of a different kind ) animate surrogates as test stimuli . The
366 Hall

main procedural change involved introducing the labeled object to both

groups as the experimenter's pet. The hypothesis was that this change
would increasechildren's tendency to view the unfamiliar object as wor-
thy of being referred to as an individual (becauseit was marked as being
possessedby someone) and so increasetheir willingness to learn a proper
name for it . However, despite adding this information about pet-hood, I
still found that children were more likely to make a proper name inter-
pretation (i.e., to select the named object and to make proper name-
appropriate spontaneouscomments) if they heard the novel word applied
to the familiar object than if they heard it for the unfamiliar object. Thus,
in both studies reported in Hall 1991, knowledge of a count noun (and
so, presumably, knowledge of a kind) for an object enhanced young
children's tendency to learn a proper name for it . These results provide
evidenceconsistentwith the hypothesis that children are more likely to
interpret a novel word as a proper name if it is used in conjunction with
an object that has already been typed by a kind than if it is used along
with an object that has not. (For an alternative interpretation of these
findings, see Hall 1991; Markman 1994; Woodward and Markman , in
press).
One intriguing finding from the second experiment reported in Hall
1991 was that, despite hearing that the labeled object was the experi-
menter's pet, children were more likely to interpret a novel word for it as
a proper name if it was familiar than if it was unfamiliar . This was a
puzzling result because" pet" is a count noun naming a kind of individual .
Children thus should have been willing to learn a proper name for both
the familiar and the unfamiliar object. Why, then, were children lesslikely
to acquire a proper name for the unfamiliar than for the familiar animal?
One possibility is that the 2-year-olds simply did not understandthe word
" pet." Another possibility, however, is that they understood the count
noun " pet" but recognizedthat it is not the right type of count noun to
support the interpretation of a proper name.
Philosophersdistinguish between" substance" count nouns (words like
" dog," " person," and " shoe" ) and " phase" count nouns (words like
" pet," " puppy," and " passenger " ) (e.g., Gupta 1980; Wiggins 1980; for
discussion of the acquisition of these words, see Hall 1993a; Hall and
Waxman 1993; Macnamara 1986). Phasecount nouns differ from sub-
stancecount nouns in that they refer to kinds that pick out individuals
defined in restricted terms (e.g., as existing over limited periods of time or
in particular contexts). It may be that phase count nouns are not good
Semantics
andProper
Names 367

candidates for supporting the interpretation of proper names, because

people assumethat proper namesrefer to individuals that are not tied to
particular conditions. In other words, when people bestow proper names,
they tend to do so on the back of substancekinds, not phasekinds. They
name individual dogs and persons, not individual puppiesand passengers .
Of course, it is possible to introduce a proper name for an object in
conjunction with a phase count noun, as in the second experiment
reported in Hall 1991(e.g., " This is my pet. This is ZA V" ). In fact, there
are many situations in which a proper name can be introduced with the
support of a phasekind . For example, a friend might introduce his spouse
through an utterance like, " This is my wife, Susan." For adults, however,
a phasecount noun like " wife" does not seemto support the interpreta-
tion of a proper name like " Susan." The reason is that it is unlikely that
the proper name " Susan" names a wife. Instead, " Susan" appears fun-
damentally to name a person (supported by a substancekind, PERSON).
For example, " Susan" probably picked out Susan long before she got
married and becamea wife. Thus, as adults, we appear to use a substance
kind to support the interpretation of a proper name, even in situations
where we know a phase kind for an object and actually hear a proper
name in conjunction with it . These considerationssuggestthat the chil-
dren who participated in this experiment may have assumedthat the
novel label was not a proper name for the unfamiliar monster, because
they required a substancekind (e.g., KIND OF MONSTER ), not a phase
kind (e.g., PET), in order to type the object suitably. I am currently
planning an experiment to test the hypothesisthat in order to interpret a
novel label as a proper name, children need knowledge of not just any
kind (count noun), but specificallya substancekind (count noun).

14.3.5 Summary
This section explored the idea that learners should be more likely to
interpret a novel label as a proper name if certain conditions of using the
word are met than if they are violated. These conditions, inspired by the
semantic definition of proper names (i.e., expressions that refer to
individuals in kinds in all situations in which the individuals appear,
regardlessof the context), are that the label be applied to only one object;
that the label be applied to an object that has not already been labeled
with a proper name; that the label be applied to an object that is marked
as important as an individual in some way (e.g., by being animate and
possessedby someone); and that the label be applied to an object for
368 Hall

which children know a (substance ) kind (e.g., a substancecount noun).

The results from studiesusing either of two types of experimental design
indicated that preschoolchildren are more likely to interpret a novel word
as a proper name if it is used in a way that upholds any of these con-
ditions than if it is used in a way that violates it . One type explored
children's construal of a novel word used to label objects in such a way
that one of the conditions was either upheld or violated; the other (more
indirect) type examinedchildren's choice of the referent of a proper name
given two alternatives, one that upheld one of the conditions, but another
that violated it .
These findings are consistent with the possibility that children learn
proper names by relying on implicit knowledge of their semantic defini-
tion, along with knowledgethat novel labelsusedunder specialconditions
(related to this definition) should be construedas belonging to this classof
expression. This knowledge could aid children in making mappings
betweenproper namesand their interpretations; and it could also provide a
way for children to learn the grammatical properties of theseexpressions ,
if they were disposed to note the contexts in which these semantically
identified labels can appear within sentencesin their language.

14.4 Discussion

By about their second birthday , children appear to understand some of

the syntactic properties of proper names, as well as links between the
syntax and semantics of these expressions. This early knowledge presents
an intriguing puzzle to researchers interested in language acquisition : how
do children acquire it ? One solution to the puzzle, explored in this chap -
ter , is that children learn proper names by exploiting tacit knowledge of
their semantic definition (e.g., Kripke 1980; Macnamara 1986; Macna -
mara and Reyes 1994), along with knowledge that labels used under certain
conditions (related to this definition ) should be interpreted as belonging to
this class of expression. Possessing this knowledge , children would have a
way both to establish appropriate links between individual proper names
and their interpretations in the language they are acquiring , and to learn
how these labels are expressed morphologically and syntactically in that
language .
An important unaddressed issue concerns whether adults ' use of proper
names when speaking to children meshes with children 's understanding of
Semantics
andProperNames 369

the semantics of these expressions, and with their understanding of the

conditions - related to these semantics- under which speakers should use
labels of this class . If children expect that proper names will be the words
used to name ( 1) only one object each, (2) objects lacking proper names, (3)
objects important as individuals (e.g., animate , human -possessedobjects),
and (4) objects already labeled by kind terms , then parents should tend to
provide children with proper names under conditions that accord with
these expectations . Furthermore , when parents introduce young children
to words in situations that violate these conditions , they should flag these
events as exceptional so as to minimize children 's mapping errors .
For example , adults should strive to use a proper name in conjunction
with only one object and attempt to signal as noteworthy those cases
where the same proper name is used to label more than one individual . In
such events, they should explain the usage directly , or they should take
measures to modify the name in order to distinguish it from a previously
learned proper name . For example , they might add a diminutive suffix to
one (e.g., " John " and " Johnnie " ), adjectives to both (e.g., " Big John "
and " Little John " ), or surnames to both (e.g., " John T ." and " John S." ).
Parents should also attempt to preserve a rule of using only one name for
each individual . They should single out for children those situations
in which one object receives more than one proper name . In addition ,
parents should try to reserve proper names for objects that children expect
to receive names (e.g., those that are animate or have animate properties ,
and are marked as possessed by people), and they should signal as ex-
ceptional the use of proper names for noncanonically namable kinds of
objects . For example , if there is a proper name for the family car , boat , or
computer , then parents should offer a special explanation of this fact to
their young children when they first tell them the name . Finally , parents
should provide proper names for objects that children have already typed
by kinds , probably those objects for which children already know a count
noun for the substance kind . If parents introduce an unfamiliar object to
their children , they should refrain from providing a proper name until
after they have first introduced a count noun for it , in particular a sub-
stance count noun . Some evidence from studies of how parents teach
words for objects suggests that parents do tend to provide substance count
nouns rather than other types of word when they first introduce unfamil -
iar objects to their children (e.g., Callanan 1985; Hall 1994b).
At present, there are no data that bear directly on the important ques-
tion of whether adults introduce proper names to young children in a way
370 Hall

that matches children '8 assumptions about which labels they hear in
speech should be interpreted as belonging to this class. However , Tracey
Burns and I are currently planning some studies to try to answer it .
Another important issue for future work to address is the origin of
children 's understanding . If young children do have knowledge of the
universal semantic definition of proper names, along with knowledge that
labels used under certain conditions (related to this definition ) should be
interpreted as proper names, then where does it come from ? The results of
the demonstrations described in this chapter are consistent with the pos-
sibility that young children possessthis understanding between the ages of
2 and 4 years, during , the period of rapid vocabulary development that
occurs in the preschool years. Yet the relatively advanced age (in terms of
the time course of language acquisition ) of the children in these studies
precludes making strong claims about the origins of this knowledge .
Further research with infants and young toddlers (i .e., those at least as
young as the 17-month -olds in the study reported in Katz et al . 1974) may
shed light on whether any or all of the knowledge also exists in children
on the brink of learning language . Evidence of the presence of this
understanding in very young children would bolster support for the view
that children initially acquire these expressions (e.g., make mappings
between them and their interpretations ; learn their grammatical proper -
ties as a class) via an unlearned semantic competence .

References

Alford , R . D . ( 1987). Naming and identity . New Haven , CT : HRAF Press.

Barwise , J., and Perry , J. ( 1983). Situations and attitudes . Cambridge , MA : MIT
Press .

Bloom, P. (1990). Syntactic distinctions in child language. Journal of Child

Language, 17, 343- 355.
Bloom, P. (1994). Syntax-semanticsmappings as an explanation for some tran-
sitions in languagedevelopment. In Y . Levy (Ed.), Other children, other languages
:
Issuesin the theory of languageacquisition(pp. 41- 75). Hillsdale, NJ: Erlbaum.
Callanan, M . (1985). How parents label objects for young children: The role
of input in the acquisition of category hierarchies. Child Development
, 56, 508-
523.
Clark, E. V. (1987 ). The principleof contrast: A constrainton languageacquisi-
tion. In B. MacWhinney(Ed.), Mechanisms of languageacquisition : The20th
AnnualCarnegieSymposium on Cognition(pp. 1- 34). Hinsdale , NJ: Erlbaum.
Frege, G. (1892 ). Uber Sinnund Bedeutung . Zeitschriftfiir PhilosophieundPhi-
losophischeKritik, 100, 25- 50.
Semanticsand Proper Names 371

Gelman , S. A ., and Taylor , M . ( 1984) . How two -year-old children interpret

proper and common names for unfamiliar objects . Child Development , 55, 1535-
1540 .

Gordon , P. ( 1985) . Evaluating the semantic categories hypothesis : The case of the
count -mass distinction . Cognition , 20, 209- 242.
Gupta , A . ( 1980) . The logic of common nouns. New Haven , CT : Yale University
Press .

Hall , D . G . ( 1991). Acquiring proper names for familiar and unfamiliar ani -
mate objects : Two -year -olds' word learning biases. Child Development , 62, 1442-
1454 .

Hall , D . G . ( 1993a) . Basic-level individuals . Cognition , 48, 199- 221.

Hall , D . G . ( 1993b) . Constraints on the interpretation of proper names. In
E. Clark (Ed .), Proceedings of the Twenty -fifth Annual Child Language Research
Forum (pp . 118- 126) . New York : Cambridge University Press.
Hall , D . G . (1994a). How children learn common nouns and proper names. In
J. Macnamara and G . Reyes (Eds.), The lagicalfaundations afcognition (pp . 212-
240) . Oxford : Oxford University Press.
Hall , D . G . ( 1994b) . How mothers teach basic-level and situation -restricted count
nouns . Journal of Child Language , 21, 391- 414.
Hall , D . G . ( 1994c) . Semantic constraints on word learning : Proper names and
adjectives . Child Development , 65, 1291- 1309.
Hall , D . G . ( 1996). Preschoolers ' default assumptions about word meaning :
Proper names designate unique individuals . Developmental Psychology , 32, 177-
186 .

Hall , D . G ., and Graham , S. A . (in press). Lexical form class information guides
word -to -object mapping in preschoolers . Child Development .
Hall , D . G ., and Moore , C . E . ( 1997). Red bluebirds and black greenflies: Pre-
schoolers' understanding of the semantics of adjectives and count nouns . Journal
afExperimental Child Psychology 67, 236- 267.
Hall , D . G ., Veltkamp , B ., and Turkel , W . ( 1997). Proper namable things.
Unpublished manuscript , Department of Psychology , University of British
Columbia .

Hall , D . G ., and Waxman , S. R . ( 1993). Assumptions about word meaning :

Individuation and basic-level kinds . Child Development , 64, 1550- 1570.
Hall , D . G ., Waxman , S. R ., and Hurwitz , W . M . ( 1993) . How 2- and 4-year -old
children interpret adjectives and count nouns . Child Development , 64, 1651- 1664.
Hockett , C . ( 1966) . The problem of universals in language . In J. Greenberg (Ed .),
Universals of language ( pp. 1- 29) . Cambridge , MA : MIT Press.
Katz , N ., Baker , E ., and Macnamara , J. ( 1974). What 's in a name? A study
of how children learn common and proper names. Child Development , 45, 469-
473 .
 372 Hall

Kripke , S. ( 1980). Naming and necessity. Oxford : Blackwell .

Liittschwager , J., and Markman , E . ( 1993). Young children 's acquisition of proper
versus common nouns. Poster presented at the biennial meeting of the Society for
Research in Child Development , New Orleans , LA .
Macnamara , J. ( 1982). Names / or things: A study of human learning . Cambridge ,
MA : MIT Press .

Macnamara , J. ( 1986) . A border dispute: The place of logic in psychology . Cam -

bridge , MA : MIT Press.

Macnamara , J., and Reyes, G . E . ( 1994) . Foundational issues in the learning of

proper names, count nouns , and mass nouns . In J. Macnamara and G . E . Reyes
(Eds.), The logicalfoundations of cognition (pp . 144- 176) . Oxford : Oxford Uni -
versity Press.

Markman , E . ( 1994) . Constraints on word meaning in early language acquisition .

Lingua , 92 , 199 - 227 .

Merriman , W . ( 1986). How children learn the reference of concrete nouns : A

critique of current hypotheses . In S. A . Kuczaj and M . D . Barrett (Eds .), The
acquisition of word meaning (pp . 1- 38) . New York : Springer -Verlag .
Nelson , K . ( 1973). Structure and strategy in learning to talk . Monographs of the
Societyfor Research in Child Development , 38( 1- 2, Serial No . 149) .
Pinker, S. (1984). Language learnability and languagedevelopment
. Cambridge,
MA : Harvard University Press.
Pinker , S. ( 1994). The language instinct . New York : Morrow .
Russell , B. ( 1905). On denoting . Mind , 14, 479- 493.
Searle, J. ( 1958). Proper names. Mind , 67, 161- 173.
Sorrentino , C . ( 1997). The role of mental state attribution on young children 's
interpretation of proper nouns. Poster presented at the biennial meeting of the
Society for Research in Child Development , Washington , DC .
Taylor , M ., and Gelman , S. ( 1988). Adjectives and nouns : Children 's strategies for
learning new words . Child Development , 59, 411- 419.
Valentine , T ., Brennen , T ., and Bredart , S. ( 1996). The cognitive psychology of
proper names. New York : Routledge .
Wiggins , D . ( 1980) . Samenessand substance. Cambridge , MA : Harvard University
Press .

Woodward , A ., and Markman , E. (in press). Early word learning . In W . Damon ,

D . Kuhn , and R . Siegler (Eds.), Handbook of child psychology : Vol. 2. Cognition ,
perception , and language. New York : Wiley .
Chapter 15
The Learning of First and Yuriko Oshima -Takane
Second Person Pronouns in
English

When I came to McGill as a graduate student, John was writing his book
Namesfor Things. John told me to read a draft of his chapter on pro-
nouns I and you for my first presentation in his graduate seminar. Since
then, I have been fascinated by the problem of indexicals, particularly
personalpronouns. I decidedto pursuethe problem of how children learn
English personal pronouns as my Ph.D . research. John gently guided me
on how to analyze the problem and showed me the complexities of the
problem that children must solve. Understanding the mind of children
solving such a complex problem is most interesting and challenging. I am
deeply grateful to John for opening my eyesto such an exciting area of
study and for sendingme off on an endlessbut fascinating inquiry of true
understanding. This chapter is about what I have learned about children's
learning of personal pronouns since I began my investigation with John.
The basic theoretical framework presentedhere was developedas part of
my Ph.D . researchand has been the guide for empirical studies I have
carried out sincethen. John helped me greatly with the philosophical lit -
erature and deepenedmy understandingof referenceand meaning.
I begin in section 15.1 by describing the problems that confront chil-
dren in learning first and secondperson pronouns in English, grammati-
cally as well as semantically. Then I analyze the semanticaspectsof per-
sonal pronouns based on Kaplan's theory of indexicals. In section 15.2,
using Kaplan's theory as a foundation, I propose a psychological theory

Preparationof this chapterwassupportedby a grant from the Natural Sciences

and EngineeringResearchCouncil of Canada . I thank Ray Jackendoff , Paul
Bloom, ThomasShultz, DavidAlexandreNicolas, DeanSharpe , MarinaTakane,
andHiroko Nakanofor veryhelpfulcommentson earlierdrafts.
374 Oshima
-Takane

of how children learn the correct semanticrules from observation and of

what kind of competencethe task of learning pronouns presupposeson
the part of children. In section 15.3, I contrast my proposal with others
that have been made in the psycholinguistic literature and discusswhy
none of the latter are satisfactory. In sections 15.4 and 15.5, I present
evidencein support of my theoretical analysis. Finally , in section 15.6, I
briefly discusswhat my investigation into the learning of personal pro-
nouns tells us about early word learning in children.

15.1 WhatHasto BeLearned

Every languagehas a group of words called indexicals, whose interpreta-

tion dependson the context of utterance. Personalpronouns are one type
of indexical: the actual person referred to by a personal pronoun changes
with the conversational role (i .e., speaker, addressee , or nonparticipants
in a conversation). This change poses a problem for children who are
learning suchpronouns. Another problem is that the model for the correct
use of these pronouns is not directly given in speechaddressedto the
children. When a mother speaksto her child- say, Mary - she saysyou
to refer to Mary and I to refer to herself. However, when Mary speaks,
shemust reversethe pronouns; shemust say I to refer to herself and you
to refer to her mother. The child must be able to understand that the
person referred to by these pronouns depends on who is speaking and
who is being spoken to. In fact, it is commonly observed that autistic
children fail to make this reversaland that blind children are often delayed
in acquiring personal pronouns, whereas normally developing children
usually master the correct use of personal pronouns by age 3 (Andersen,
Dunlea, and Kekelis 1984; Charney 1980a; Clark 1978). This chapter
investigateshow normally developing children come to understand the
meaning of first and secondperson pronouns so early.
Proper namesor count nouns as well as personalpronouns can be used
to refer to individuals (Macnamara 1982, 1986). For example, a particular
person can be referred to by the proper name David, the definite descrip-
tion the man on the corner, the function expressionfather of Tom, and
personal pronouns such as I , you, or he. These different expressionsall
refer to the sameindividual , yet they are distinct semantically as well as
grammatically. In this section, I will specify the competencethat anyone
must have in order to usepersonal pronouns.
First and SecondPersonPronouns 375

15 . 1 . 1 Grammatical Aspects

The third person pronouns substitute for noun phrases or proper names ,

not for nouns ( Lyons 1983 ; Macnamara 1982 ) . First and second person

pronouns , however , do not seem to substitute for noun phrases or proper

names ( Lyons 1980 ) . English speakers normally use pronouns to refer to

a person in the role of speaker or addressee , although the use of proper

names or kinship terms is commonly observed in child - directed talk

( Wills 1977 ; Durkin , Rutter , and Tucker 1982 ; Durkin et al . 1982 ;

Oshima - Takane 1993 ; Oshima - Takane and Derat 1996 ) . Unlike count

nouns , pronouns do not normally take determiners ( i .e . , a , the ) , quanti -

fiers ( i .e . , some , many , etc .) , or adjectives . Personal pronouns have case

distinctions such as I , me , and my , whereas neither count nouns nor

proper names do . These grammatical differences may draw children 's

attention to the semantic differences between personal pronouns and

proper names or count nouns when they learn personal pronouns ( Bloom

1990 ; Katz , Baker , and Macnamara 1974 ) .

15 . 1 . 2 Semantic Aspects

In order to understand the semantic aspects of the personal pronouns , let

us begin by comparing the following two pairs of sentences :

Mother Mommy is reading a book .

Child Mommy is reading a book .

Mother I am reading a book .

Child I am reading a book .

The sentence " Mommy is reading a book " expresses the same proposition

whether the mother or the child utters it . In order to judge whether the

proposition expressed by this sentence is true or false , we only need to

know whether or not the mother is reading a book when the sentence is

uttered . On the other hand , the sentence " I am reading a book " expresses

different propositions depending on whether the mother or the child utters

it . Thus , in order to judge whether the proposition expressed by the sen -

tence " I am reading a book " is true or false , we should know not only

who is reading a book when the sentence is uttered but also who utters the

sentence . This does not deny that the sentence " I am reading a book " has

a fixed meaning . The sentence has the same meaning whether the mother

or the child utters it , just as the sentence " Mommy is reading a book "

does . But the proposition expressed by the sentence changes depending on

whether the mother or the child utters it .

376 Oshima-Takane

In order to adequately capture such characteristicsof a sentencecon-

taining indexicals, Kaplan (1977, 1978, 1989) distinguishestwo functions:
content and character. The content is what is expressedby an expression
in a particular context of use, whereasthe characterof the expressionis a
semantic rule that is used to determine the content of the expressionin
any given context of use. If you say, " I am eating lunch," and I say the
same, the content is different, since in saying that sentenceyou are
describingwhat you are doing, whereasI am describingwhat I am doing.
If I wish to expressthe samecontent as you, I must say, " You are eating
lunch" (when talking to you) or " He/ She is eating lunch" (when talking
to someoneelse).
By contrast, the character of an expressiondoes not change with the
context of use. Rather, it is used to determine the content in any given
context of use. The character of I , for example, tells us that the content of
I used by you is you (referent) and the content of I used by me is me
(referent) . It tells us that the word I refers to the person who uses it ,
whoeverusesit . The characteris set by linguistic convention and is shared
by competent speakersof a linguistic community. The character comes
close to what we normally expressby the word meaning.
If the expressionis an assertivesentencesuch as " I am reading a book"
or " Mommy is reading a book," the content is a proposition. In Kaplan's
framework, the processof understanding an assertivesentenceused by
someoneand of judging whether it is true or false may be viewed as fol-
lows: To determinethe content of an expression, we use knowledgeabout
the character of the expressionand knowledge about the context where
the expressionis used. We then use knowledge about the content thus
picked out in order to judge whether the content of the expressionis true
or false.
Thus, when we hear the utterance " I am reading a book," if we know
who utters it and we also know the character of I, we can determine the
content of I : the content of I used by the mother is the mother and the
content of I used by the child is the child. By contrast, the content of
Mommy in the utterance " Mommy is reading a book" does not change
whether the mother or the child utters it . However, this does not mean
that the content of Mommy is context-free. If the boy next door saysthe
samesentence, the content of Mommy would probably be his mother. In
order to determine the content of Mommy on a particular occasion, we
use knowledge about the context as well as knowledge about the charac-
ter. The samething is true of a proper name and of a count noun phrase
First and SecondPersonPronouns 377

like the girl . When we hear the utterance " Mary is coming soon," the
character of the proper name Mary tells us that it refers to the individual
named Mary , whoever utters it . The content of the proper name Mary
must be the same individual at all times and in all places in which the
individual exists and in all circumstances in which the individual might
possibly exist . In this sense, a proper name is a rigid designator (Kripke
1980).1 However , there is more than one Mary in the world and the
character of the proper name is not sufficient to determine the content on
a particular occasion . Knowledge of the context is also necessary. Count
noun phrases like the girl or the dog running over there are called definite
descriptions, and we use them to refer to a particular object of which the
description is true . In order to determine the content of a definite
description in a particular context of use, again knowledge of the context
is necessary unless there is only one object in the world that fits the
description . Hence , proper names and definite descriptions as well as
personal pronouns are all dependent on the context of use. However , the
way personal pronouns depend on the context is systematically different
from the way proper names or definite descriptions do .
In order to capture such differences , we should break the context fur -
ther into three features , following Barwise and Perry (1983). These fea-
tures of the context , which are used in different ways in getting from the
character of an expression to its content on a particular occasion of use,
are discourse situation , connections , and resource situations .
A discourse situation is a situation in which an utterance occurs ; " it

involves who is speaking , when , and where , what words are being uttered ,
and to whom " ( Barwise and Perry 1983, 121) . If the mother says to her
child , " I am reading a book ," the utterance (" I am reading a book " ), the
speaker (the mother ), the addressee (the child ), the time of the utterance ,
and the location of the utterance are the elements of discourse situation .

The role of the facts of discourse situations in fixing the content is usually
referred to as indexicality , and the content of any indexical must be
among the elements of the discourse situation .2 Particular elements in the
discourse situation that are of crucial relevance to the first and second

person pronouns are the speaker and the addressee. On the other hand ,
the contents of proper names or definite descriptions are not necessarily
among the elements of the discourse situation and do not change so sys-
tematically with the discourse situation . In order to determine the content
of proper names or the contents of definite descriptions , we use the
speaker's connections and resource situations .
378 Oshima-T akane

We are related to various things in the world . For example, there are
many Marys and many dogs in the world, but we know only someof the
Marys and some of the dogs. In other words, we have connections to
certain Marys and certain dogs in the world . When we use the proper
name Mary or the definite description the dog, we refer to a particular
Mary among the Marys we know (or perceive) or we refer to a particular
dog among the dogs we know (or perceive). Connections that link an
utterance to its content on a particular occasion of use are called the
speaker's connectionsin the utterance. The addresseemust find out these
connectionsto get from the character of the utterance to its content on a
particular occasionof use, and the speakermust believethat the addressee
can recover them. For example, if a child says, " Mommy is reading a
book," the mother would think that her child refers to her by Mommy,
whereasif the boy next door saysthe samesentence , shewould think that
he refers to his mother, not to her. To determine the content of Mommy
used by her child, therefore, the mother uses her child's connections,
whereasto determine the content of Mommy used by the boy next door,
she uses the boy's connections. The character of an expression simply
givesus a generalrule to refer to things to which we and other people are
connected.
Although there are many different expressionsthat refer to the same
thing, not all of them are available to the speakerand the addresseeall the
time. For example, when I tell a friend somethingabout Mary , I can refer
to her by her name, Mary . But if I cannot rememberher name or if my
friend does not know Mary , I cannot use M al"y to refer to her. However,
if my friend and I seeMary standing at the door, I can usethe description
the girl standingat the door to refer to Mary . In this case, I am using the
fact that Mary is standing at the door to refer to Mary . This fact about
Mary is called a resourcesituation that allows me to use the description
the girl standing at the door as referring to Mary , and my own and my
friend's perceptual connections to this resource situation allow me to
successfullyrefer to Mary by that description. However, if I am talking to
my friend over the telephoneand my friend cannot seeMary standing at
the door, I have to use different resourcesto refer to Mary . For example,
if my friend knows Mary 's father, Mr . Clark, I may use the description
Mr . Clark's daughter to refer to Mary . Here I use the fact that Mary 's
father is Mr . Clark as a resource situation, and my connections to this
resourcesituation and my friend's connection to Mr . Clark allow me to
successfullyrefer to Mary by this description.
First and SecondPersonPronouns 379

To summarize : Proper names and definite descriptions as well as per-

sonal pronouns all depend on the context of use, and all have their own
character that is used to determine their content in each context of use . In

order to determine the content of a personal pronoun , we use knowledge

about a particular feature of the context , the discourse situation , whereas
in order to determine the content of a proper name or a definite descrip-
tion , we use knowledge about other features of the context , connections
and resource situations . Unlike the content of personal pronouns , the
content of proper names and definite descriptions need not be present
in the discourse situation , although it has been reported that when talking
to a child , parents most often talk about things to which the child is per-
ceptually connected (Macnamara 1972, 1982).
What does all this tell us about what has to be learned when a child

learns the singular first and second person pronouns ? It tells us that the
child has to learn at least two things : ( 1) the character of each pronoun
(i .e., the semantic rule that is used to determine the content of the pro -
noun on each occasion of use), and (2) the elements of discourse situations
that are relevant to pronouns . The characters of the singular first and
second person pronouns are given by the following rules ( Barwise and
Perry 1983; Kaplan 1977) :
(I ) A first person pronoun , in each utterance , refers to the person who
uses it .

( II ) A second person pronoun , in each utterance , refers to the person

who is addressed when it is used .

The elements of the discourse situation relevant to the first and the second

person pronouns are the speaker and the addressee, respectively .

15.2 Learning : A Proposal for a Psychological Theory

Children learn most of their words from hearing other people using them
in the course of normal conversation . The personal pronouns are no
exception . Many normally developing children master the first and second
person pronouns with few errors by the age of 3, without explicit teaching
(Clark 1978; Macnamara 1982, 1986; Oshima -Takane 1985, 1988; Shipley
and Shipley 1969; Strayer 1977) . This suggests that children learn the
meaning of personal pronouns from hearing the pronouns used in the
speech of others . Following Kaplan ' s analysis of indexicals , I assume that
each personal pronoun has a fixed meaning (character ), but the referent
380 Oshima
-Takane

(content) of the pronoun shifts depending on the context of use, that is,
who is speakingand who is spoken to. Starting from this premise, I have
proposed that the child's task of learning personal pronouns consists of
the following two parts (Oshima-Takane 1985): (1) to correctly determine
the content of pronouns used by others in each instance of use (i.e., to
comprehendpronouns correctly), and (2) to induce the correct semantic
rules- character- that would lead to the correct production of pronouns
as well as correct comprehension. In this section, I will analyze what
abilities each presupposeson the part of the child and propose a psycho-
logical theory of how children learn first and secondperson pronouns.

15.2.1 Determining the Content of PersonalPronouns

In order to determinethe content of a personalpronoun usedby others in
a particular context, a competent speaker can use knowledge about its
semanticrule as well as knowledgeabout the context of use. However, the
child learning the personalpronouns doesnot know the semanticrules. In
fact, this is preciselywhat the child has to learn. Therefore, the child must
use other means to deteffi1inethe content of a pronoun used by others.
What abilities, then, doesthis task presupposeon the part of the child? In
order to determine the content of a pronoun in each instance of use, the
child must be able (1) to identify the pronoun forms in each utterance,
and (2) to understandthat tne pronoun used by someonerefers to a par-
ticular personin the discoursesituation. Let us analyzetheserequirements
more closely.

15.2.1.1 Identifying the Pronouns In order to identity the pronoun

forms in the utterance, the child must somehow be able to segmentthe
speech stream into words. Although there is evidence suggestingthat
children learning languageare able to use prosodic cuesto identify word
boundaries(e.g., Wijnen, Krikhaar , and Den Os 1994), studiesof parental
speechto young children have shown that parents greatly help young
children to segmentthe speechstream into words or phrasesby singling
out and repeating individual words or phrases(Snow 1976). Parentsoften
isolate a proper name or a common noun and repeatit severaltimes when
they want their child to say it (Macnamara 1982). Vocative usesof proper
namesprobably also help children to identify the proper namesin utter-
ances. English personalpronouns, however, are unlikely to be singled out
in this way, and the child may have to segmentout the pronoun forms
embeddedin utterances. Developmental psycholinguistshave pointed out
First and SecondPersonPronouns 381

that some of the earliest uses of pronouns by children are part of a set
phrase such as " I do it ," and pronouns are not differentiated from other
words in utterances (Charney 1980b; Clark 1978).
Although children may not segment out the pronoun forms in the
utterances from the beginning , they must do so at a certain point in order
to figure out the character of the pronouns . The data reported in the lit -
erature show that before beginning to use any personal pronouns , all
normally developing children correctly use some nonpronominal terms
for others such as Mommy , Daddy , Jamie , and some count nouns such as
doggie, car, ball (Macnamara 1982; Nelson 1973). Further , some children
correctly use their own names and even use two -word sentences like
" Mommy sit" or " More milk ." These data suggest that personal pro -
nouns are not the first words they learn . It is possible, then , that they can
use their knowledge of the words they have already acquired to segment
out the personal pronoun forms . Interchangeable uses of proper names
and pronouns by parents (e.g., " That ' s you . That 's David " ) may also help
a child identify the pronoun foffi1s in utterances (Macnamara 1982, 1986;
Oshima -Takane 1985; Strayer 1977) .
There is abundant evidence that children use kinship terms such as
Mommy and Daddy and some proper names for others before using
any personal pronouns ( Nelson 1973; Oshima -Takane 1985; Seki 1992;
Smiley and Huttenlocher 1995). However , most children start to use their
own names at about the same time they start to use a personal pronoun to
refer to themselves. Some children call themselves by their own names
before they start to use any personal pronouns ; other children use per-
sonal pronouns to refer to themselves before they start to use their own
names (Macnamara 1982; McNeill 1963; Oshima -Takane 1985; Schiff -
Myers 1983; Strayer 1977) . Comprehension and production data on
proper names and personal pronouns that I collected for my Ph.D . research
indicated that most of the children who used personal pronouns in refer -
ence to themselves but did not yet use their own names responded to their
own names correctly in comprehension tests (Oshima -Takane 1985) .
Further , studies with Japanese children provide some evidence suggesting
that children understand their own names correctly before they begin to
use their names or any pronouns designating themselves. Uemura ( 1979),
who observed 19 Japanese children from 1 month to 20 months of age in
experimental as well as natural settings, reported that they understood
their names used by others about two months before they started to use
their names themselves. For example , from 15 months on, one child
382 Oshima-Takane

beganrespondingdifferently to hearing her own name and others' names.

Before 15 months, shesaid " Yes" or smiled when sheheard others' names
as well as her name. Starting at 15 months, shesaid " Yes" only when she
heard her own name. At 18 months, she started to use her name, and at
19 months, she started to point to herself when she was called by her
name. More recently, Seki (1992) found that about half of the 12- to 17-
month-old children that she studied respondedonly to their own names
when their caregiver at the daycare center called children's names, but
only when their caregiver looked at them while calling the names. More
than 80% of the children aged between 18 and 23 months did so even
when their caregiverdid not look at them while calling the names. Parent
report data in a separatestudy by Seki indicated that 96% of the children
between18 and 23 months called themselvesby their own namesbut none
of them usedfirst person pronouns to designatethemselves .
Sincemany English-speakingchildren begin using personalpronouns at
around 20 months, the data from Japanesechildren together with my
English data suggest that children know their own name as well as
others' names before they begin to use any personal pronouns and may
use their knowledge about proper names to identify the pronoun forms
in utterances.

15.2.1.2 UnderstandingThat a PersonalPronoun Refersto a Person In

order to understand that a personal pronoun in an utterance refers to a
particular person, the child must have understood what a personal pro-
noun designateswhen it is used, and, in particular, that what is designated
by the pronoun is a person, sinceit is only personswho can talk and take
speechroles.3 This doesnot, however, mean that the child has the notion
of referring or the notion of reference- the relation betweenwords and
objects to which they refer (Macnamara 1982, 1986). When they hear
someone using a word, children must understand the speaker's inten-
tion and identify an object the word refers to (Macnamara 1982, 1986).
Recent evidence shows that young children actively search for non-
linguistic cues to the speaker's intention and use them to infer what the
speaker intends to indicate with the utterance (Akhtar and Tomasello
1996; Baldwin 1991; Tomasello 1995). Further, parents often give young
children nonlinguistic cuesto their intention. For instance, parents often
show their child an object they are referring to by holding or pointing to it
(Clark 1978; Macnamara 1977, 1982). Although this practice helps the
First and Second Person Pronouns 383

child to identify what parents are referring to, parents rarely point to
people when they use personalpronouns (Wales 1979).
How, then, do children come to identify a person referred to by a pro-
noun? Macnamara (1986) arguesthat to identify the referent of a personal
pronoun, the child must have accessto the basic-level kind PERSON.
Kinds supply a principle of individuation , a principle of identity, and a
principle of application for what they are true of. In the caseof the kind
PERSON, the principle of individuation specifieswhat counts as an indi-
vidual person. The principle of identity provides the criteria for deciding
whether individuals considered at different points in time are the same
person or different persons. The principle of application picks out an
individual person from all other things that are not a person. Macnamara
emphasizesthat accessto the kind PERSON is particularly important
becauseit is a person that is the referent of a personalpronoun, not just a
person's face, nose, or visible surface. The ability to recognize an indi -
vidual as a member of the kind PERSON is presupposed.
Application of the kind PERSON to an individual presupposesthat the
child is able to pick out a person from all other animate and inanimate
objects and to assignthe person as a member of the kind . Thus, having
accessto the kind PERSON helps the child figure out that a pronoun
picks out a person. In fact, if the child were not able to recognizea person
as distinct from all other animate and inanimate objects, the child would
never be able to correctly determine what a personal pronoun refers to.
Recent evidencesuggeststhat children can make global kind distinctions
(e.g., animal/artifact, vehicle/ animal) and basic-level kind distinctions
(e.g., bottle/ball, cup/book) before they talk (Mandler, Bauer, and
McDonough 1991; Xu and Carey 1996). But how do children know that,
when they hear a pronoun usedby others, it refersto an entire person, not
a body part such as a nose or an object other than a person such as a
chair? This is what children must discoverby observinghow other people
use pronouns. As just noted, children are sensitiveto the speaker's inten-
tion and actively search for cues to understanding what the speaker is
expressingwith an utterance. Linguistic and nonlinguistic contexts in
which the pronouns are used may help them discover the referents. Fur-
ther, parents' interchangeableuseof personalpronouns and proper names
should help them discover that a pronoun refers to a person, not a body
part or an object other than a person (Macnamara 1986; Oshima-Takane
1985). For example, if a child already knows the word Mommy to refer to
his mother, who is a member of the kind PERSON, and if the child hears
384 Oshima-Takane

his mother saying , " Give it to me . Give it to Mommy ( while extending

her hand toward the child ) ," the child will be more likely to take me as

referring to his mother , rather than referring to a body part such as her
hand .

15 . 2 . 2 Inducing the Correct Semantic Rules

In order to induce the correct semantic rules , children must not only

determine the referent of the pronouns used by others but also discover

the relationship between the pronoun and the speech role of the person it

designates . As noted , parents often use personal pronouns and proper

names interchangeably when they talk to their child , and this practice

may help the child understand what pronouns refer to . However , this

practice does not necessarily help the child figure out the correct use of the

pronouns . This is because when children use pronouns , they must reverse

what they hear . Even if a child correctly determines the content of the

pronouns used by others , several generalizations about the semantic rules

for pronouns are logically possible . Children may think that first person

pronouns are used to refer to another person with whom they talk , or that

first person pronouns are used to refer to others who are speaking but

cannot be used to refer to themselves even when they are speaking , or that

first and second person pronouns refer to either participant in a conver -

sation ( i . e . , the speaker or the addressee ) or indeed to anybody . Children

must discover the correct relationships between the pronouns and the

speech roles from such variously interpretable data and induce the

semantic rules .

What abilities does this task presuppose on the part of the child ?

Obviously , the child must be able to identify at least two elements of the

discourse situation : who the speaker is and who the addressee is . This

presupposes that the child is able to distinguish speaker from addressee

and also able to distinguish participants from nonparticipants in a con -

versation . There is little evidence about when children come to identify

speaker , addressee , and nonaddressee as such . Several developmental psy -

cholinguists claim that children first distinguish the speaker from others ,

and then distinguish the addressee from nonaddressees ( Brener 1983 ;

Strayer 1977 ) . However , this order of differentiation of speech roles is

based on the observation that children initially come to use first person

pronouns correctly , then second person pronouns , and finally third person

pronouns . Thus , it is not clear when such differentiations occur . To

determine whether children identify the speaker , the addressee , and the
First and SecondPersonPronouns 385

nonaddresseebefore they learn personal pronouns, an approach will be

neededthat directly examinestheir knowledge of speechroles.
In a series of experiments that examined children's interpretation of
nonlinguistic signsas a clue to the speaker's intention, Macnamara (1977)
showed that children between 14 and 20 months of age were more
responsiveand gavemore correct responsesto instructions when the tester
was looking at them than when the tester was looking at their mother,
whereaschildren between 11 and 14 months of age did not show such a
differentiation. Since the tester did not give instructions until the child
returned his or her look or the child looked at the mother, inattentiveness
doesnot seemto explain the differences. Rather, the data suggestthat 14-
to 20-month-old children can interpret the speaker's eye contact with
them as a sign that the speakeris talking to them. This, in turn, suggests
that children can identify who the speaker is since they cannot use the
speaker's eye contact as a sign without knowing who the speakeris.
Children begin to usepersonal pronouns betweenthe agesof 18 and 24
months (Charney 1980b; Clark 1978; Macnamara 1982; Oshima-Takane
1985). The above results suggestthat children about to learn personal
pronouns have some knowledge about speechroles. Although 14- to 20-
month-old children may not be able to use the speaker's eye contact as a
cue in determining the addresseewhen they are not addressed , they can at
least use the speaker's eye contact to determine whether they are being
addressedor not. That is, the children can at least identify another person
as speaker, themselvesas addressee , and themselvesas nonaddressee .
Whether children can induce the correct semantic rules for first and
second person pronouns depends on whether the input they receive
includes the following situations: ones in which they can recognize that
the samepronoun used in different discoursesituations refers to different
persons, and ones in which they can recognize that different pronouns
refer to the sameperson. The first type of situation is particularly impor-
tant for distinguishing personal pronouns from proper names. This is
becausechildren can avoid interpreting a pronoun as a proper name (e.g.,
you as another name for the child) if they recognizethat the referent of the
pronoun is not limited to a specificindividual . The secondtype of situa-
tion plays an important part in learning the distinctive meanings of dif-
ferent pronouns (i.e., first, second, and third personpronouns). Overheard
speechis an input that is necessaryfor inducing the correct semanticrules
of first and second person pronouns becauseit often provides children
with these two situations, whereas child-directed speech does not
386 Oshima-Takane

(Oshima -Takane 1985, 1988, 1992). For instance, in overheard speech

children often have an opportunity to hear the pronoun you referring to a
person other than themselves , and to hear different pronouns , me and you ,
referring to the same person . Furthermore , children can witness that the
referent of me and you shifts systematically depending on who is speaking
and who is addressed. In child -directed speech, on the other hand , they
observe that you is always used to refer to themselves and me is always
used to refer to the person talking to them . Thus , children who do not
have enough opportunities to hear pronouns in overheard speech or fail
to notice them are likely to make incorrect generalizations about the
meaning of pronouns and to show errors in using them .
To recognize that the same pronoun used on different occasions refers
to the same person and that different pronouns used on different occa -
sions refer to the same person presupposes that children can trace their
own identity and that of others over time . In fact , such identity is
implicated in the use of many referring expressions, including personal
pronouns (Gupta 1980; Hall 1994; Hall and Waxman 1993; La Palme
Reyes, Macnamara , and Reyes 1994; Macnamara 1986; Macnamara and
Reyes 1994; Oshima -Takane 1985). Suppose a child hears her mother
saying you to refer to her on one occasion and her father saying you to
refer to her on another occasion . In order to understand that the mother 's

use of you and the father 's use of you both refer to the same person (the
child ), the child must be able to identify herself as herself over time . Sim-
ilarly , in order to understand that the mother 's use of me and the father 's
use of you both refer to the same person (the mother ), the child must be
able to identify the mother as the same person over time . This does not
mean, however , that mistakes cannot happen , as when an adult or a child
fails to identify a person when the person looks like someone else or the
person's appearance has changed dramatically .
Macnamara ( 1986) argues that as in the case of proper names (for
persons), access to the kind PERSON is necessary for learning personal
pronouns , since it not only helps the learner understand that the referent
of a personal pronoun is a whole person but also helps the learner handle
the identity of the person over time . This means that when children learn
personal pronouns or proper names, the kind PERSON must be available
to them and can be used to trace the identity of members of that kind . It is
likely that this knowledge is implicit in children who are learning pro -
nouns ; that is, they can recognize an individual as a member of the kind
PERSON but do not know the count noun person . Macnamara noted
First and SecondPersonPronouns 387

that his son, Kieran, did not usethe count noun person, which denotesthe
kind PERSON, until long after he used many proper namesfor persons
and personal pronouns (Macnamara 1986). Consistent with this obser-
vation, examinations of children's early vocabularies indicate that the
count noun personis not among the first 50 words (Nelson 1973; Oshima-
Takane 1985). It is possible that children correctly interpret person used
by others even though they do not produce it , although, as far as I am
aware, no systematic investigation has attempted to determine when
children come to understand this word. In any event, we have some evi-
dence indicating that children correctly use others' names and also cor-
rectly understand or use their own names before they start to use any
personal pronouns. Becausechildren cannot understand or use others'
names and their own nameswithout identifying the personsnamed over
time, the proper name data suggestthat children can trace both their
own identity and other people's by the time they start to use personal
pronouns.
Some researchers(see, e.g., Smiley and Huttenlocher 1995) have sug-
gestedthat children may initially fail to include themselvesin the kind
PERSON becausethey have limited accessto " physical appearanceof the
whole self (except in the mirror )" (p. 41). This is certainly possible, but
physical appearanceis not the only infoffi1ation children use when cate-
gorizing objects into different kinds (Baldwin, Markman , and Melartin
1993; Xu and Carey 1996). Perhaps children categorize themselvesin
the samekind as their parents but not in the samekind as dogs and cats
becausethey notice that they behave like their parents but unlike dogs
and cats. The point is that children must be able to recognizethemselves
as membersof the kind PERSON by the time they master personal pro-
nouns, eventhough they may initially fail to do so. As Macnamara (1986)
has argued, unless children know that they are also persons, it will be
difficult for them to realize that they can usefirst personpronouns to refer
to themselves .

15.2.3 Summary
The above analysis shows that the learning of first and second person
pronouns presupposesthat children have the following abilities:
. To identify the speaker, the addressee, and the nonaddresseein a dis-
coursesituation;
. To understandthe speaker's intention;
388 0 shima-T akane

. To segmentout the pronoun fofll1s in the utterance;

. To accessthe kind PERSON.

Although it is not known exactly when and how each prerequisite

becomesavailable to children, the above analysis suggeststhat at least
someof them are available from the time children begin to learn pronouns;
thus, they can use them in determining the referents of pronouns and in
inducing the semanticrules. All four prerequisitesare necessaryto deter-
mine the referent of a pronoun used by others; the first is particularly
necessaryto discoverthe relations betweenpronouns and speechroles. Of
course, these prerequisites are necessarybut not sufficient for correctly
determining the content of a pronoun usedby others and for inducing the
correct semantic rules. They may help children reduce the number of
logically possiblehypotheses,but they do not tell them which oneis correct.

15.3 Comparingthe PresentProposalwith Other Theories

The crucial requirement for understanding and using first and second
person pronouns- what one needs beyond what is needed for dealing
with other referring expressions such as proper names and definite
descriptions- is noticing the regularities in their shifting referents. I have
proposedthat overheard speechprovides situations important for discov-
ering the systematic relationships between pronouns and speech roles
(Oshima-Takane 1985, 1988). This proposal has challengeda prevailing
view in the Western literature on child language, becauseit attributes to
children the ability of analyzing overheardspeech. Much of Westernchild
language research has emphasizedthe importance of specially adapted
speechthat mothers use in child-directed talk . Overheard speechis not
consideredan important source of input from which children learn lan-
guage, becauseutterancesnot addressedto children are generally longer
and more complex than those addressedto them (see Oshima-Takane,
Goodz, and Derevensky 1996for more detailed discussionon this issue).
My argument is this: If we assumethat, without explicit teaching, children
are able to understand that the same personal pronouns (in particular,
secondperson pronouns) can be used to refer to different persons, and
that different personal pronouns (i .e., first person and secondperson pro-
nouns) can be usedto refer to the sameperson, then we must also assume
that children are able to analyze overheard speech(nonaddressedspeech )
as well as child-directed speech(addressedspeech). If children do not
First and SecondPersonPronouns 389

attend to nonaddressed speech, as the prevailing view suggests, then we

must assume that children are " biased" to make such generalizations
solely on the basis of addressed speech (Oshima -Takane 1985, 1988).
My proposal about how children learn English first and second person
pronouns differs from previous proposals in several important ways . In
this section, I will contrast my proposal with others in the psycholinguistic
literature (Charney 1980a,b; Chiat 1981, 1986; Clark 1978; Oshima -
Takane 1985, 1988, 1992; Schiff-Myers 1983; Shipley and Shipley 1969;
Strayer 1977) and discuss why none of the latter are satisfactory .
Interesting variations are observed in pronoun acquisition : most nor -
mally developing children acquire first and second person pronouns
with few production errors , whereas a few persistently make such errors .
Indeed , previous theories have focused either on children 's pronominal
errors or on their lack of pronominal errors . Clark (1978) has noted that
there are two different routes for acquiring first and second person pro -
nouns . One is that children learn these pronouns relative to the speech
roles from the outset , and they therefore make few production errors . The
other is that children initially entertain a proper name interpretation
such as you :=: child and me :=: adult , and they therefore make persistent
pronominal errors . However , Clark does not explain why some children
make persistent errors , whereas many children do not . My proposal
explains the observed variations using the same mechanism . Children who
attend to shifting reference of personal pronouns in overheard speech
would discover the relationship between pronouns and speech roles and
would acquire the pronouns with few production errors . Children who do
not have enough opportunities to hear overheard speech would not dis-
cover the relationship between the pronouns and speech roles and would
be more likely to entertain the incorrect , reversed semantic rules; as a
result , they would make persistent pronominal errors (Oshima -Takane
1985, 1988).
N one of the existing theories has seriously considered the distinction
between referent (content ) and meaning (character ). Consequently , they
have failed to recognize the importance of the kind PERSON in learning
personal pronouns . Most theories assume that children make few prono -
minal errors because they map a personal pronoun directly onto a speech
role rather than onto the person taking the particular speech role . For
instance, Shipley and Shipley ( 1969) argue that most children make few
errors in production because they immediately comprehend pronouns as
being relative to the speaker and the addressee (p . 113) . Chi at ( 1981)
390 Oshima-T akane

statesthat " the relative infrequency of pronoun reversals. . . suggeststhat

the conceptsof speakerand addresseeare so salient for most children that
they selectthese, rather than the particular individuals to which they refer,
as the content underlying 1st and 2nd person pronoun forms" (p. 85).
However, it would be dangerousto draw such a conclusion from studies
of spontaneous speechalone, as these authors have, becausechildren
can avoid making pronominal errors in production by simply replacing
pronouns with proper names(Oshima-Takane and Oram 1991). Further-
more, these theories do not explain why some children do show prono-
minal errors. According to my proposal, thesechildren are able to identify
the referent of the pronouns used in the addresseespeechcorrectly but
induce the incorrect semanticrules.
Charney (1980b) proposes a somewhat different view of why many
children do not make pronominal errors, but her account suffers from a
more serious problem. She argues that many children make few prono-
minal errors becausethey do not learn pronouns as referring to a person.
Rather, they first learn pronouns relative to their own speechroles. That
is, children initially learn first person pronouns as referring to themselves
in the role of speakerand secondperson pronouns as referring to them-
selvesin the role of addressee . Only later do children recognizethat first
person pronouns refer to the speaker and that second person pronouns
refer to the addressee . In this way, Charney suggests , many children can
avoid pronominal errors without discovering the correct semantic rules.
Charney's proposal, then, assumesthat children come to use first person
pronouns correctly without understandingthesepronouns usedby others.
This assumption is logically impossible, because, as Chiat (1981) also
points out, if children use first person pronouns correctly without errors,
they must know that first person pronouns can refer to others as well as
themselves. Children must be able to recognize that others use a first
person pronoun as referring to themselvesbefore or no later than they
correctly use the same first person pronoun as referring to themselves .
However, Charney claims that her proposal is based on the observed
developmental discrepancy between comprehension and production
(Charney 1980b; Strayer 1977). That is, some children correctly use first
personpronouns to refer to themselveswithout understandingthem when
used by others, whereas children understand second person pronouns
referring to themselves , though they do not use them. Careful examina-
tion of Charney's data suggeststhat the children classified as correctly
using first person pronouns without comprehendingthem failed the com-
First and SecondPersonPronouns 391

prehension test, not because they understood first person pronouns as

referring to themselves as speaker, but because they produced them as a
part of rote phrases such as " I got it " in which I does not have any self-
reference at all (Chiat 1981, 1986; Oshima -Takane 1985). A major prob -
lem with Charney ' s claim lies in the assumption that children do not
comprehend personal pronouns simply because they do not pass con-
trolled comprehension tests. The literature indicates that compared with
spontaneous speech studies, such controlled comprehension studies tend
to underestimate children 's linguistic competence, and the age estimates
for particular kinds of linguistic competence change according to proce -
dures employed to test them ( Brown 1973). Charney 's proposal has ig-
nored these difficulties with comprehension tests. There seem to be no
empirical grounds for claiming that children can use a first person pro -
noun correctly without knowing that the same pronoun can refer to an-
other person in the speech of others (see Oshima -Takane 1985, 1988 for
more detailed discussions on this issue).
Finally , several investigators have suggested that for children to
understand the shifting reference of first and second person pronouns , they
must be able to take the speaker's point of view (Andersen , Dunlea , and
Kekelis 1984; Charney 1980a; de Villiers and de Villiers 1974; Loveland
1984; Tager -Flusberg 1989). For instance, Loveland ( 1984) proposes that
understanding spatial viewpoints is a cognitive prerequisite to under -
standing the speaker's point of view , which , in turn , is a prerequisite to
the correct use of personal pronouns . However , the evidence for a causal
link between understanding of spatial viewpoints and acquisition of these
pronouns is not conclusive (Girouard , Ricard , and Decarie 1995; Issler
1993 ; Loveland 1984 ; see Oshima - Takane , Takane , and Shultz 1996
for further discussion on this issue). Others , especially those focusing on
autistic children 's pronominal errors , take a similar view . For instance,
Charney (1980a) and Tager -Flusberg ( 1989) attribute autistic children 's
difficulty in learning personal pronouns to their inability to take others '
perspective . It should be noted that although pronominal errors are often
observed in autistic children , some normally developing children also
show persistent pronominal errors . Further , there is some evidence that
at about 9 months of age, long before they begin to learn personal
pronouns , normally developing children show behaviors suggesting that
they can understand another 's perspective (Baron -Cohen and Ring 1994;
Tomasello 1995). Macnamara has remarked that " infants learn their
language by first determining , independent of language , the meaning
392 0 shima-T akane

which a speakerintends to convey to them, and by then working out the

relationship betweenthe meaning and the language" (Macnamara 1972,
1; seealso Macnamara 1982). In fact, unless children have the capacity
to grasp the intention of other speakers, they cannot determine the re-
ferent of any words they hear (Akhtar and Tomasello 1996; Baldwin
1991). As discussedearlier, the major problems with first and second
personpronouns are not only that the referent of the pronouns shifts with
the speakerbut also that the model for correct usageis not provided in
speechaddressedto children. Thus, even if children could correctly iden-
tify the person that a pronoun used by someoneelse referred to, they
might be unable to produce the pronoun correctly. Understanding the
speaker's intention is not sufficient for learning the semanticrules under-
lying the correct use of these pronouns. The present proposal stresses
that observing shifting referenceof the pronouns in overheard speechis
necessary .

15.4 EmpiricalEvidence

In this section, I will present evidence in support of the proposal that

children first determinethe referent of a pronoun usedby others and then
induce the semantic rules. In particular, I will show that in learning
personal pronouns, children do not map pronouns directly onto speech
roles from the outset. Instead, they first determine the person a pronoun
refers to in the speechof others and then make generalizationsabout the
semantic rules. I will also present empirical findings in support of the
claim that observing shifting reference of first and second person pro-
nouns in overheard speechis necessaryfor inducing the correct semantic
rules for thesepronouns.
According to my analysis, children who do not have enough oppor-
tunities to observe shifting referenceof personal pronouns in overheard
speechshould entertain a proper name interpretation at least for second
person pronouns and use secondperson pronouns to refer to themselves .
However, some researchershave questionedwhether children who make
pronominal errors in fact treat pronouns as proper names (Chiat 1981;
Macnamara 1982). For instance, Chiat (1981) has argued that if children
entertain a proper name interpretation, they should use a pronoun refer-
ring to a particular person consistently and therefore should make con-
sistentpronominal errors. Nonetheless, the data reported in the literature
suggestthat their pronominal errors are inconsistent.
First and SecondPersonPronouns 393

A major difficulty in determining whether children make pronominal

errors becauseof an incorrect semantic representationsuch as a proper
name interpretation is that noffi1ally developing children rarely make
pronominal errors (Charney 1980b; Chiat 1981; Shipley and Shipley
1969). In fact, I might have reacheda conclusion similar to Chiat's, had
I not come across two normally developing children who went through
periods of making consistentpronominal errors (Oshima-Takane 1985).
One child was a boy named Donald . At the age of 18 months, he began
using the secondperson pronoun you to refer to himself (e.g., " Daddy
play with you" ) and the first person pronoun me to refer to his mother
(e.g., " Give it to me" ). His errors persistedfor about 4 months. The other
child, a boy named David, began making consistenterrors with both first
and secondperson pronouns at around 23 months. His errors persisted
for about 10months. Both children were firstborns, and eachwas the only
child in his family . Their comprehensionand production data indicated
that they entertained a proper name interpretation for second person
pronouns and used them interchangeablywith their proper namesin self-
reference. For first person pronouns, evidencefor the proper name inter-
pretation was not as clear-cut as that for secondperson pronouns. For
instance, David's comprehension data indicated that he initially inter-
preted first personpronouns to refer only to his mother; then to any other
person; and then to any other personwho was speaking. At the sametime
that David interpreted first person pronouns as referring only to his
mother in the comprehensiontask, he was using them to refer to any
person other than himself. The rapid shift from his initial misinterpreta-
tion of first person pronouns as referring to his mother to his later mis-
interpretation of first person pronouns as referring to any speakerother
than himself and the fact that he held the latter misinterpretation for a
longer period of time has led me to the following hypothesis: David ini -
tially assumedthat first person pronouns referred only to his mother (i.e.,
proper name interpretation), probably becausehis mother was the person
who used first person pronouns in self-referencemost frequently when
talking to him. He soon noticed other personsbesideshis mother using
first person pronouns in self-reference, and began using first person pro-
nouns to refer to them. However, it took him a long time to realize
that first person pronouns could refer to any speaker including himself
and that second person pronouns could refer to any addresseeother
than himself. All this suggeststhat David did not have any problem with
deteffi1ining the referent of first and second person pronouns in child-
394 Oshima
-Takane

directed speech. His problem was that he had induced the incorrect
semantic rules, which were based solely on the information provided in
child -directed speech (see Oshima -Takane 1985, 1992 for more detailed
discussions) .
A subsequent longitudinal study of 10 girls (Oshima -Takane and Oram
1991) provides further evidence that children determine the referent of
pronouns used by others before they learn the relationship between pro -
nouns and speech roles . Children 's comprehension and production data
were collected at three different ages: 21, 24, and 30 months . Compre -
hension was tested both when the child was an addressed listener and

when she was a nonaddressed listener . Spontaneous production data were

collected by videotaping spontaneous speech in a 50-minute free play sit-
uation . Analyses of spontaneous language showed that children made few
production errors in the course of acquisition . However , at the earlier
stages (21 and 24 months ), in nonaddressed listener trials half of the
children exhibited systematic comprehension errors indicating that they
thought that the pronoun you referred to themselves. No child showed
systematic comprehension errors for me. These results suggest that even
though children make few production errors , they do not necessarily
understand the correct meaning of the pronouns . It appeared that children
made few production errors because they did not use pronouns until they
understood the correct meaning ; they used proper names instead of pro -
nouns . Since the children who showed systematic comprehension errors
comprehended you correctly when they were addressed, they did not have
any problem with determining the referent of you in speech addressed to
them . Rather , they showed sy'stematic errors because they induced the
incorrect semantic rule based on child -directed speech only . These find -
ings suggest that in learning personal pronouns , children do not immedi -
ately relate pronouns to speech roles from the outset . Rather , the data
seem to support the proposal that children first determine the referents of
pronouns used by others and then make generalizations about the mean-
ings of the pronouns . Observational studies with hearing -impaired chil -
dren and an experimental study with autistic children have also provided
evidence supporting this proposal (Cole , Oshima -Takane , and Yaremko
1994; Oshima -Takane and Benaroya 1989; Oshima -Takane , Cole , and
Yaremko 1993) .
Now let us examine empirical evidence concerning another aspect of
the present proposal , namely , that observing shifting reference of personal
pronouns is necessary for inducing the correct semantic rules for first and
First and Second Person Pronouns 395

second person pronouns. It was hypothesized that children induce the

correct semantic rules by observing shifting referenceof pronouns used
in overheard speech, whereas they induce incorrect semantic rules by
observing pronouns in child-directed speech. To test this hypothesis, I
conducted a teaching experiment with 19-month-olds who were about to
learn personalpronouns (Oshima-Takane 1985, 1988). In this experiment,
I askedparentsto play two pronoun games(the me-me gameand the me-
you game) with their child to model the pronouns me and you twice a
week for six weeks under two different experimental conditions. In the
addresseecondition, the parents modeled the pronouns in turn while
looking at the child. The mother pointed to herself and said me, or
pointed to the child and said you. The father did the same thing, saying
me when pointing to himself or saying you when pointing to the child. In
the nonaddresseecondition, the parents modeled the pronouns while
looking at each other in half of the sessionand at the child in the other
half. That is, while looking at the father, the mother said me while point-
ing to herself or said you while pointing to the father. The father then did
the samething. When the mother addressedthe child, she looked at the
child and said me while pointing to herself or said you while pointing
to the child. In both conditions, the child was expected to imitate the
parents' use of the pronouns and their pointing gestures. The resultswere
consistentwith my proposal. The nonaddresseecondition facilitated the
correct imitation pattern (i.e., without errors, the child said me while
pointing to himself or herself and/ or said you while pointing to the person
he or she was looking at). The addresseecondition, on the other hand,
often led to the incorrect imitation pattern (i.e., at least once, the child
said me while pointing to the person he or shewas looking at and/or said
you while pointing to himself or herself). However, only a few children
imitated incorrectly all the time, and there was no significant condition
effect. The most interesting finding was that not a single child in the
addresseecondition showed the correct imitation pattern; only children
in the nonaddresseecondition did. The finding suggeststhat in learning
the correct meaning of personal pronouns, children need to observepro-
nouns occurring in overheard speech. Further, it suggeststhat even chil-
dren under 2 years of age attend to and can extract information from
speechaddressedto another person.
A subsequentobservational study has provided naturalistic evidence
consistent with this experimental finding (Oshima-Takane, Goodz, and
Derevensky 1996). The study demonstrated that secondborn children
396 Oshima
-Takane

produced correct pronouns earlier than firstborns, even though the two
groups of children did not differ on other languagemeasuressuchas mean
length of utteranceand vocabulary size. Our argument is that secondborn
children acquire the correct usageof pronouns earlier than firstborn chil-
dren becausethey have relatively more opportunities to hear pronouns
used in overheard speech, that is, in conversation between their parents
and their older siblings. A correlational analysisof secondbornchildren's
input and pronoun development (Oshima-Takane 1997) provides more
direct evidence for the importance of overheard speech in pronoun
acquisition. The more frequently the secondbornsheard first person pro-
nouns in overheard speech, the earlier they mastered the use of these
pronouns. Furthermore, the fewer unconventional nominal references
(e.g., useof Mommy or the older sibling's name instead of first and second
person pronouns) the mothers or older siblings used in overheard speech,
the earlier the child masteredthe use of secondperson pronouns.
In sum, the evidencereviewedso far is consistentwith the proposal that
children first determine the referentsof the pronouns used by others and
then induce semanticrules. Experimental and observational studies pro-
vide converging evidence that, in inducing the correct semantic rules,
children benefit from observing shifting referenceof pronouns in over-
heard speech. Furthermore, observational studies have shown that some
children make pronominal errors in production becausethey hold incor-
rect semantic representations, and thesemirror the way first and second
person pronouns are used in child-directed speech: me refers to any other
person and you refers to themselves . Another important finding is that
although a proper name interpretation for second person pronouns
appearsto be common among children at an earlier stage, many of them
do not show production errors becausethey produce proper names in-
steadwhen referring to themselvesas speakerand to others as addressee .
Nonetheless, the evidence that children induce the incorrect, reversed
rules (i.e., me means any other person and you means themselves ) by
observing pronouns in child-directed speechis limited. This is becausein
the teaching experiment, there were more children in the addresseecon-
dition who showedan inconsistenterror pattern than thosewho showeda
consistenterror pattern. Further, the number of children in the addressee
condition who showed the consistent error pattern was not significantly
greater than the number of children who showedthis pattern in the non-
addresseecondition. One problem with the teaching experiment is that
children who observedpronoun models in the addresseecondition could
First and SecondPersonPronouns 397

observethe samepronouns used in overheard speechoutside the experi-

ment. Therefore, there is no guaranteethat their imitative behaviorswere
based solely on the infoffilation provided in the addresseecondition.
Some of them might have been influenced by information provided in
overheard speechand thus might have begun showing the correct imita-
tive behavior, which would result in the inconsistent incorrect imitation
pattern.

15.5 ComputerModeling

For ethical reasons, we cannot manipulate the linguistic input outside the
teaching experiment in such a way that children in the addresseecondi-
tion are not exposedto any overheard speech. Further, even though we
analyze the children's task in order to specify the prior knowledge or
abilities that they must have, we cannot directly test the effects of the
hypothesizedprior knowledge or abilities in empirical studies. Computer
modeling may provide a way to complement such limitations associated
with human empirical studies. One advantage of computer modeling is
that we can test hypothetical learning mechanismsby setting up ideal
environmental conditions that would be ethically impossible with human
children. In this section, I present evidence obtained from computer
modeling studies that I have been conducting with my collaborators
(0 shima-Takane, Takane, and Shultz 1995, 1996; Shultz, Buckingham,
and 0 shima-Takane 1994). In thesestudies, we have concentratedon the
secondpart of children's task in learning pronouns: how children induce
the semanticrules of first and secondperson pronouns after determining
I
the content of a pronoun in the speechof others. In particular, the studies
were designedto obtain more clear-cut evidencethat children induce the
incorrect, reversedrules by observing pronouns used in addressedspeech
only, whereasthey induce the correct rules by observingpronouns usedin
nonaddressedspeech. In addition, we examined the role of knowing the
kind PERSON in this induction process.
We used the cascadecorrelation (CC) learning algorithm (Fahlman
and Lebiere 1990), with which networks can grow dynamically to adapt
to more complicated problems. The algorithm starts with a network
without hidden units. It first tries to improve its performance within a
given network topology by adjusting the connection weights. When it can
no longer improve its performance by merely adjusting the weights, it
changesthe network topology by adding a hidden unit that is good at
398 Oshima-T akane

detecting the current error. It then readjuststhe connection weights going

into the output units. This cycle is repeateduntil all output activations are
within score-threshold of their targets for all training patterns. The topo-
logical changes in the network define distinct developmental stagesin
learning.
We used the CC learning algorithm becauseit can represent " rules"
that connect inputs to outputs by learning regularities present in the
training patterns. In the case of learning first and second person pro-
nouns, the rules involve an interaction betweentwo variables: the speech
roles of the personsin the discoursesituation and the intended referentsof
the pronouns, becausethe personreferredto by a pronoun shifts depending
on who is speaking or who is spoken to. The CC learning algorithm is
particularly good at capturing interaction effects among input variables
such as thesewithout being told which interactions are important .
All modeling experimentsdescribedin this section involve two training
phasesto simulate how children learn to produce the two pronoun forms
me and you by listening to other persons producing them. In phase I
training, networks learn other-speaking patterns. In phase II training,
networks learn child-speakingpatterns that are added to phaseI training
patterns. PhaseI training can be seenas the period during which children
hear pronouns used by others but have not yet produced any pronouns,
and phase II training as the period during which they not only hear the
pronouns but also start producing them. The initial CC network used in
our simulations had input units representingspeaker, addressee , and ref-
erent, and output units representingthe pronoun. In eachtraining pattern
(i.e., eachinstanceof use), semanticcues(i .e., discourseas well as referent
information) relevant to the pronoun produced were given: the speaker
unit indicated who the speakerwas, the addresseeunit indicated who the
addresseewas, and the referent unit indicated who the speakerintended to
refer to.
In the first set of simulations (Shultz, Buckingham, and Oshima-
Takane 1994), networks learned other-speaking patterns with only the
mother and the father using the pronouns in phaseI training before they
learned child-speaking patterns in phase II training, but information
concerning the kind PERSON was not implemented as prior knowledge.
We used binary distributed coding to representmother (01), father (10),
and child (11) as distinct individuals but not as membersof the samekind .
Two output units were used to code me (+ .5 - .5) and you (- .5 + .5). We
predicted that if the notion of the kind PERSON is not crucial for
First and Second Person Pronouns 399

inducing the correct semantic rules, networks trained with addressee pat -
terns (i .e., patterns in which the child was always addressed) during phase
I training would learn the incorrect , reversed rules by the end of this
training and would need phase II training to master the correct child -
speaking patterns . By contrast , networks trained entirely with non -
addressee patterns (i .e., patterns in which the child never was addressed)
during phase I training would learn the correct rules and would produce
the correct child -speaking patterns without phase II training . However , if
the notion of the kind PERSON plays an important role in this induction
process, as assumed in the present proposal , then even networks trained
with nonaddressee patterns would not induce the fully correct semantic
rules . They would need at least some phase II training to produce the
correct child -speaking patterns . The results support the latter prediction .
N one of the networks trained with nonaddressee patterns in the first set of
simulations could learn to produce the correct child -speaking patterns
without requiring explicit error -correcting feedback (phase II training ),
although they corrected errors quickly . On the other hand , the networks
trained with addressee patterns learned the incorrect , reversed rules and
needed extensive phase II training to correct their errors .
The second set of simulations was designed to test whether adding prior
knowledge of the kind PERSON , the information that the child (self ) is a
member of the same kind as mother and father (other ), would facilitate
the induction of the correct semantic rules without error -correcting feed-
back (Oshima -Takane , Takane , and Shultz 1995, 1996). We used analog
coding to implement prior knowledge about PERSON . That is, the child
was coded as 0 , the mother was coded as + 2 , and the father was coded
as - 2. One output unit was used to code me (+ .5) and you (- .5) . The
analog coding allows us to represent members and the class to which they
belong by assigning a number to the members on the same unit .4 Further ,
the distinction between self (child ) and others can be derived from the
regularities of pronoun use in the training patterns . With the previous
binary distributed coding , on the other hand , the networks treat each
person as a distinct object with no relation to other objects in the training
patterns . In the second set of simulations , we also investigated whether
experiencing many examples involving various persons was another
important factor facilitating induction of the correct semantic rules . To
do this , we added two other persons, coded as + 1 and - 1, respectively .
We compared the conditions in which the networks heard only two other
persons using pronouns (3-person conditions ) with those in which they
400 Oshima-Takane

heard four other persons using pronouns (5-person conditions ). Children

learning personal pronouns noffi1ally hear persons other than parents
using pronouns (e.g., older siblings , grandparents , baby -sitters) even
though their parents ' utterances are the major source of input they hear .
Thus , we thought it reasonable to assume that overhearing pronouns used
by more than two persons may help children to realize that a pronoun
refers not only to a specific person who has interacted with them but also
to any person , depending on that person' s speech role .
As predicted , when prior knowledge about the kind PERSON was
implemented with the analog coding , the CC networks were able to pro -
duce the correct child -speaking patterns without any error -correcting
feedback if they heard pronouns used by a variety of speakers in non -
addressed speech. The nonaddressee-trained networks in the 3-person
conditions showed better generalizations than the addressee-trained net-
works . However , they needed some phase II training to master the child -
speaking patterns . On the other hand , all of the nonaddressee-trained
networks in the 5-person conditions showed perfect generalization to the
child -speaking patterns without any error -correcting feedback , just like
human children . Analysis of the knowledge representation of these net-
works indicated that generalization to untrained , other -speaking patterns
was very good . By contrast , the networks trained with addressee patterns
learned the incorrect , reversed rules and needed extensive phase II train -
ing to correct errors .
The child 's natural language environment involves some mixture of
addressed and nonaddressed speech. For instance , a firstborn child with
no sibling may hear mostly addressed speech while one parent is away at
work and some nonaddressed speech when that parent returns . A second-
born child , on the other hand , may hear more nonaddressed speech (e.g.,
conversations between the at-home parent and older sibling while the
other parent is at work ). To simulate variations in children 's natural lan -
guage environment , another 5-person simulation experiment with differ -
ent mixtures of addressee and nonaddressee materials included in phase I
training was conducted . Interestingly , networks that were given non -
addressee patterns 20% of the time or more showed perfect generalization
to child -speaking patterns without any phase II training , just like the pure
nonaddressee networks that are given nonaddressee patterns only . On the
other hand , networks that were given nonaddressee patterns 10% of the
time needed some phase II training to master the child -speaking patterns ,
just like the -pure addressee networks that were given addressee patterns
First and Second Person Pronouns 401

only . Network analysis revealed that without nonaddressed speech , net -

works would learn incorrect , reversed rules and make consistent reversal

errors . A subsequent remedial experiment using the pure addressee net -

works with three different mixtures of addressee and nonaddressee mate -

rials provided evidence that opportunities to hear shifting reference of

pronouns in nonaddressed speech facilitate the unlearning of incorrect
semantic rules . Further , the results indicated that if the incorrect , reversed
rules are learned solely from addressed speech , substantially more such
opportunities are needed to unlearn the incorrect rules ( see 0 shima -
Takane , Takane , and Shultz 1996 for more detailed information ).
If children learning personal pronouns fail to include themselves in the
kind PERSON , it will be difficult for them to realize that they , too , can
use first person pronouns to refer to themselves ( Macnamara 1986 ) . We
tested this with another modeling experiment in which the distinction
between self (child ) and other was explicitly coded but individuals were
not coded as members of the same kind . The result supported our pre -
diction . Generalization tests immediately after nonaddressee training
revealed that none of the networks produced the first person pronoun me
in reference to the child ( self ) , although some did produce the second
person pronoun you in reference to a person other than the child . In order
to induce the correct semantic rules without error - correcting feedback ,
children need to have knowledge of the kind PERSON , in which both self
and other are included ( Oshima - Takane , Takane , and Shultz 1995 , 1996 ;
Shultz et al ., 1995 ) .
In sum , evidence from the computer modeling studies clearly shows
that addressed speech is a source of incorrect , reversed semantic rules and
that nonaddressed speech is a necessary ingredient for inducing the correct
rules . Knowing the kind PERSON and being exposed to many examples
involving various persons as the referents of the pronouns are also
important factors for inducing the correct semantic rules without produc -
tion errors .

15.6 ConcludingRemarks

Following Kaplan's (1977, 1978, 1989) analysis of indexicals, I have

assumed that a personal pronoun has a fixed meaning (character) ,
although its referent (content) systematicallyshifts dependingon the dis-
course situation. On the basis of the semantic rules for first and second
person pronouns defined by Kaplan, I have proposed a psychological
402 Oshima
-Takane

theory of how children learn the semantic rules as they observe other
people use such pronouns. That is, children first determine the person
referred to by a pronoun in each instanceof use and then make generali-
zations about the semanticrules for first and secondperson pronouns. In
order to induce the correct semanticrules, children must observeshifting
reference of pronouns in overheard speech. Child -directed speechis a
source of pronominal reversal errors; children who do not have enough
opportunities to observe shifting pronominal reference in overheard
speechinduce the incorrect, reversedrules.
The empirical studies together with the computer modeling studies
provide converging evidencein support of this proposal. The data indi-
cate that children determinewhich person a pronoun usedby others refers
to in each context of usebefore they discover the relationship betweenthe
pronouns and speechroles. There is no indication that children map pro-
nouns directly onto speechroles. Further, the empirical studies demon-
strate that children even younger than 2 can extract information given in
overheard speechto induce the correct semanticrules for first and second
person pronouns. The computer modeling studies also provide clear-cut
evidencethat children induce the incorrect, reversedrules by observing
pronouns used in addressedspeechonly, whereasthey induce the correct
rules by observingpronouns usedin nonaddressedspeech.In addition, the
studiesdemonstratethat knowing the kind PERSON and being exposed
to many examplesinvolving various personsas the referents of the pro-
nouns are important in this induction process.
In this chapter, I have not dealt with the problem of how children
determine the content of pronouns in the speechof others. However, this
does not mean that I assumethat children have no problem in determin-
ing the content of the pronouns used by others. As discussedin section
15.1, there are severalproblems children must solve in order to correctly
determine the content of pronouns used by others, and none of these
problems is as simple as it looks. I have assumedthat children use their
prior knowledgeof proper names(their own name and others' names) and
of the kind PERSON in determining the content of personal pronouns
used by others. Prior knowledge of proper namesis important in several
respectsfor determining the content of pronouns used by others. First, if
children know their own and other people's names, we can assumethat
they are able to trace their own identity as well as that of others over time
before they learn personal pronouns. Second, we can assumethat the
interchangeableuse of proper names and personal pronouns by parents
First and SecondPersonPronouns 403

helps children identify both the pronoun forms in utterancesand the ref-
erentsof the pronouns. Knowing the kind PERSON is important because
children's ability to recognizethemselvesand others as members of the
same kind PERSON is essential for understanding that personal pro-
nouns refer to a person.
This doesnot mean, however, that with accessto their prior knowledge
of proper names and the kind PERSON, children have no problem in
determining the content of pronouns used by others. For instance, one
mother, who participated in the teachingexperimentdescribedabove, told
me that her son appearedto think that me referred to a neck, probably
becauseshe and the father pointed toward their neck when they played
the me-me game (0 shima-Takane 1985). Since this boy had already
called himself by his name and his parents by Daddy and Mommy, his
problem was not that he did not know any word that designatedhim or
his parents as a whole person. Instead, he did not yet seemto have dis-
covered that the particular word me refers to a whole person, not just a
person's neck. Interestingly, this boy seemedto solve this problem the
day after the final sessionof the me-me game. His mother's diary stated
that when she went to her son's bedroom in the morning, he began
pointing to himself and said, " Joe, me" several times with a confident
look . This was the first time he used the pronoun me with his name, Joe.
The mother interpreted this event as indicating that her son discovered
that when he uttered me, it picked out himself as an entire person, just like
his name. Sinceparentsused me only in the me-me game and did not use
names along with me at all, the boy must have discovered the corefer-
ential relationship between names and the pronoun me used by adults
outside the teaching experiment. In normal conversations, unlike the
teachingexperiment, the boy could hear pronouns usedin sentences ; thus,
syntactic cues, too, might have helped him reject the body part interpre-
tation. Further, the experienceof observinghis mother and father playing
the me-me game with each other (i .e., in the nonaddresseecondition)
might have helped him to discover that me refersto the person who utters
it . Perhapshis implicit assumption that he is of the samekind PERSON
as his parents might have led him to use me to refer to himself without
hesitation. Unless children actively search for cues to the speaker's in-
tention, as this boy did, they will never be able to arrive at the correct
interpretation.
The present approach does not assumeany fast mapping mechanisms
specific to word learning as do the special lexical constraints approaches
404 Oshima-T akane

popular in psychology (e.g., the object -kind bias or the mutual exclusivity
assumption - Markman 1989; Merriman and Bowman 1989). Findings
from the studies described here suggest that the learning of personal pro -
nouns is gradual and cannot be explained by these fast mapping mecha-
nisms. It could be that the constraints and biases posited to be specific to
word learning are not yet fully developed during the early stage of pro -
noun learning . Or these constraints may be operating only for certain
types of word learning . Logical analysis of children 's task in learning
words other than personal pronouns and detailed empirical studies of
their solutions to the task such as those discussed here will elucidate the

early language -learning process and will provide further insights into the
ability children bring to language learning .
Notes

1. Note that other referring expressions such as indexicals and definite descrip -
tions refer rigidly (i .e., the referent is taken as fixed for all possible circumstances ),
once the referent is detennined . Kaplan uses directly referential for these referring
expressions (see Kaplan 1989) .
2. In the case of the direct -quotational use of indexicals , as in " Teddy Bear said ,
'I 'm sleepy,' " the person who reports what Teddy Bear said does not use the
words in quotation marks but mentions them . Thus , the discourse situation of the
expression being quoted is the one Teddy Bear was in when he uttered it .
3. This does not mean that we cannot use a personal pronoun to refer to a thing
other than a person . In fact , people often talk to a dog or a cat or even an inani -
mate object using personal pronouns . However , it is considered that animals and
inanimate objects in such cases are personified , and second person pronouns are
used to refer to them .

4. The number assigned to each person is on a nominal scale, and it simply

identifies the members with respect to the property in question . With this coding ,
networks represent the members (tokens ) and the kind (type ) to which they
belong and trace their identity over time . It should be noted , however , that
networks do not know what each number indicates and what the property in
question is.

References

Akhtar , N ., and Tomasello , M . ( 1996). Two -year -olds learn words for absent
objects and actions . British Journal of Developmental Psychology , 14, 79- 93.
Andersen , E . A ., Dunlea , A ., and Kekelis , L . S. ( 1984) . Blind children 's language :
Resolving some differences . Journal of Child Language , 11, 645- 664.
Baldwin , D . A . (1991). Infants ' contribution to the achievement of joint reference.
Child Development , 62, 875- 890.
First and SecondPersonPronouns 405

Baldwin , D . A ., Markman , E . M ., and Melartin , R . L . ( 1993) . Infants ' ability to

draw inferences about nonobvious object properties : Evidence from exploratory
play . Child Development , 64, 711- 728.
Baron -Cohen , S., and Ring , H . ( 1994) . A model of the mindreading system:
Neuropsychologicaland neurobiological perspectives
. In C. Lewis and P. Mitchell
( Eds.), Children 's early understanding of mind : Origins and development ( pp . 183-
207) . Hillsdale , NJ : Erlbaum .
Barwise , J., and Perry , J. ( 1983) . Situations and attitudes . Cambridge , MA : MIT
Press .

Bloom , P. (1990). Syntactic distinctions in child language . Journal of Child

Language , 17 , 343 - 355 .

Brener , R . ( 1983) . Learning the deictic meaning of third person pronouns . Journal
of Psycho linguistic Research, 16, 330- 352.
Brown , R . ( 1973) . A first language: The early stages. Cambridge , MA : Harvard
University Press.
Charney , R . ( 1980a) . Pronoun errors in autistic children : Support for a social
explanation . British Journal of Disorders of Communication , 15, 39- 43.
Charney , R . ( 1980b) . Speech roles and the development of personal pronouns .
Journal of Child Language , 7, 509- 528.
Chiat , S. ( 1981). Context -specificity and generalization in the acquisition of pro -
nominal distinctions . Journal of Child Language , 8, 75- 91.
Chiat , S. ( 1986) . Personal pronouns . In P. Fletcher and M . Garman ( Eds .),
Language acquisition : Studies in first language development (2nd ed., 339- 355) .
Cambridge : Cambridge University Press.
Clark , E . V . ( 1978) . From gesture to word : On the natural history of deixis in lan -
guage acquisition . In J. S. Bruner and A . Garton ( Eds .), Human growth and devel-
opment: Wolfson College Lectures ( pp . 85- 121) . Oxford : Oxford University Press.
Cole , E ., Oshima -Takane , Y ., and Yaremko , R . ( 1994). Case studies of pronoun
development in two hearing -impaired children : N onnal , delayed , or deviant ?
European Journal of Disorders of Communication , 29, 113- 129.
de Villiers , P. A ., and de Villiers , J. G . ( 1974) . On this , that , and the other : Non -
egocentrism in very young children . Journal of Experimental Child Psychology , 18,
438 - 447 .

Durkin , K ., Rutter , D . R ., Room , S., and Grounds , P. ( 1982) . Proper name usage
in maternal speech: A longitudinal study . In C . E . Johnson and C . L . Thew ( Eds.),
Proceedings of the Second International Congressfor the Study of Child Language
( pp . 405- 412) . Washington , DC : University Press of America .
Durkin , K ., Rutter , D . R ., and Tucker , H . ( 1982) . Social interaction and language
acquisition : Motherese help you . First Language , 3, 107- 120.
Fahlman , S. E ., and Lebiere , C . ( 1990) . The cascade correlation learning architec -
ture . In D . S. Touretzky ( Ed .), Advances in neural information processing systems 2
( pp . 524- 532) . San Mateo , CA : Morgan Kaufmann .
406 Oshima-Takane

Girouard , P. C ., Ricard , M ., and Decarie , T . G . ( 1995). Perspective-taking skills

and the acquisition of pronouns . Paper presented at the biennial meeting of the
Society for Research in Child Development , Indianapolis , April .
Gupta , A . ( 1980). The logic of common nouns. New Haven , CT : Yale University
Press .

Hall , D . G . ( 1994). How children learn common nouns and proper names. In
J. Macnamara and G . E. Reyes ( Eds.), The logical foundations of cognition
( pp . 212- 240) . New York : Oxford University Press.
Hall , D . G ., and Waxman , S. R . ( 1993). Assumptions about word meaning :
Individuation and basic-level kinds . Child Development , 64, 1550- 1570.
Issler , D . ( 1993). Comprehension of spatial points of view and acquisition of
personal pronouns in Brazilian Portuguese. Unpublished master ~s thesis, Pontifical
Catholic University of Rio Grande do SuI, Brazil .
Kaplan , D . ( 1977) . Demonstratives . Paper read at a symposium on demonstratives
at the March 1977 meetings of the Pacific Division of the American Philosophical
Association .

Kaplan , D . ( 1978) . On the logic of demonstratives . Journal of Philosophical Logic ,

8 , 81 - 98 .

Kaplan , D . ( 1989) . Demonstratives : An essay on the semantics, logic , meta -

physics , and epistemology of demonstratives and other indexicals . In J. Almog ,
J. Perry , and H . Wettstein ( Eds .), Themesfrom Kaplan ( pp . 481- 614) . New York :
Oxford University Press.

Katz , N ., Baker , E ., and Macnamara , J. ( 1974) . What 's in a name? A study of

how children learn common and proper names. Child Development , 45, 469- 473 .
Kripke , S. ( 1982) . Naming and necessity. Oxford : Blackwell .
La Palme Reyes, M ., Macnamara , J., and Reyes, G . ( 1994) . Reference, kinds and
predicates . In J. Macnamara and G . E . Reyes ( Eds.), The logicalfoundations of
cognition (pp . 91- 143). New York : Oxford University Press.
Loveland , K . A . ( 1984) . Learning about points of view : Spatial perspective and
the acquisition of 'J/you '. Journal of Child Language , 11, 535- 556.
Lyons , J. ( 1983) . Deixis and modality . Sophia Linguistica , 12, 77- 117. Tokyo :
Sophia University .
Macnamara , J. ( 1972) . The cognitive basis of language learning in infants .
Ps} 'chological Review, 79, 1- 13.

Macnamara , J. ( 1977). From sign to language . In J. Macnamara ( Ed .), Language

learning and thought (pp . 11- 35) . New York : Academic Press.
Macnamara , J. ( 1982). Namesfor things: A study of human leal'ning . Cambridge ,
MA : MIT Press .

Macnamara , J. (1986
). A borderdispute
. The place of logic in psychology . Cam -
bridge, MA : MIT Press .
First and SecondPersonPronouns 407

Macnamara , J., and Reyes, G . E . ( 1994) . Foundational issues in the learning of

proper names, count nouns and mass nouns . In J. Macnamara and G . E . Reyes
( Eds.), The logical foundations of cognition ( pp . 144- 176) . New York : Oxford
University Press.
Mandler , J. M ., Bauer , P. J., and McDonough , L . ( 1991). Separating the sheep
from the goats : Differentiating global categories . Cognitive Psychology , 23, 263-
298 .

Markman , E . M . ( 1989) . Categorization and naming in children . Cambridge , MA :

MIT Press .

McNeill , D . ( 1963). The psychology of you and I : A case history of a small

language system. Paper presented at the annual convention of the American
Psychological Association .
Merriman , W ., and Bowman , L . ( 1989). The mutual exclusivity bias in children ' s
word learning . Monographs of the Society for Research in Child Development ,
54(3- 4, Serial No . 220) .
Nelson , K . ( 1973). Structure and strategy in learning to talk . Monographs of the
Societyfor Research in Child Development , 38( 1- 2, Serial No . 149).
Oshima -Takane , Y . ( 1985) . Learning of pronouns . Unpublished doctoral disserta -
tion , McGill University .
Oshima -Takane , Y . ( 1988) . Children learn from speech not addressed to them :
The case of personal pronouns . Journal of Child Language , 15, 94- 108.
0 shima -Takane , Y . ( 1992) . Analysis of pronominal errors : A case study . Journal
of Child Language , 19, lll - l3 ! .
Oshima -Takane , Y . ( 1993) . Why do children make nominal reference? Poster pre-
sented at the biennial meeting of the Society for Research in Child Development ,
New Orleans , March .

Oshima -Takane , Y . ( 1997) . Effects of overheard speech on pronoun development in

secondbornchildren. Paper presentedat the biennial meeting of the Society for
Research in Child Development , Washington , DC , April .
Oshima -Takane , Y ., and Benaroya , S. ( 1989). An alternative view of pronominal
errors in autistic children . Journal of Autism and Developmental Disorders , 19, 73-
89 .

Oshima -Takane , Y ., Cole , E ., and Yaremko , R . ( 1993) . Semantic pronominal

confusion in a hearing -impaired child : A case study . First Language , 13, 149- 168.
Oshima-Takane, Y ., and Derat, L . (1996). Nominal and pronominal referencesin
maternal speech during the later stage of language acquisition : A longitudinal
study . First Language , 16, 319- 338.
Oshima -Takane , Y ., Goodz , E ., and Derevensky , J. L . ( 1996) . Birth order effects
on early language development : Do secondborn children learn from overheard
speech? Child Development , 67, 621- 643.
Oshima -T akane , Y ., and Oram , J. ( 1991) . A cquisi tion of personal pronouns : What
do comprehension data tell us? Poster presented at the Biennial Meeting of the
 408 Oshima-Takane

InternationalSocietyfor the Studyof BehaviouralDevelopment

, Minneapolis
,
July.
Oshima-Takane, Y., Takane, Y., and Shultz, T. R. (1995
). Learningof personal
pronouns
: Networkmodelandanalysis(Tech. Rep. No. 2095). Montreal: McGill
University.
Oshima-Takane, Y., Takane, Y., and Shultz, T. R. (1996
). Learningof first and
secondpersonpronounsin English: Network modeland analysis . Unpublished
paper, McGill University.
Schiff - Myers , N . ( 1983 ) . From pronoun reversals to correct pronoun usage : A case

study of a normally developing child . Journal of Speech and Hearing Disorders ,

48 , 385 - 394 .

Seki , M . ( 1992 ) . Acquisition of the first person terms in infants : A comparison

between normally developing children and mentally retarded children . Unpublished

master ' s thesis , Tsukuba University .

Shipley , E . F . , and Shipley , T . E . ( 1969 ) . Quaker children ' s use of thee : A rela -

tional analysis . Journal of Verbal Learning and Verbal Behavior , 8 , 112 - 117 .

Shultz , T . R . , Buckingham , D . , and Oshima - Takane , Y . ( 1994 ) . A connectionist

model of the learning of personal pronouns in English . In S . J . Hanson , T . Petsche ,

M . Kearns , and R . L . Rivest ( Eds . ) , Computational learning theory and natural

learning systems : Vol . 2 . Intersection between theory and experiment ( pp . 347 -

362 ) . Cambridge , MA : MIT Press .

Shultz , T . R . , Schmidt , W . C . , Buckingham , D . , and Mareschal , D . ( 1995 ) .

Modeling cognitive development with a generative connectionist algorithm . In

T . J . Simon and G . S . Halford ( Eds . ) , Developing cognitive competence : New

approaches to pl ' ocess modeling ( pp . 205 - 261 ) , Hillsdale , NJ : Erlbaum .

Smiley , P . , and Huttenlocher , J . ( 1995 ) . Conceptual development and the child ' s

early words for events , objects , and persons . In M . Tomasello and W . Merriman

( Eds . ) , Beyond names for things : Young children ' s acquisition o / verbs ( pp . 21 - 61 ) .

Hillsdale , NJ : Erlbaum .

Snow , C . E . ( 1976 ) . Mother ' s speech . In W . von Raddle - Engel and Y . Lebrun

( Eds . ) , Baby talk and infant speech ( pp . 263 - 264 ) . Amsterdam : Swets and

Zeitlinger .

Strayer , J . ( 1977 ) . The development of personal references in the language of two -

year - olds . Unpublished doctoral dissertation , Simon Fraser University .

Tager - Flusberg , H . ( 1989 ) . An analysis of discourse ability and internal state lex -

icons in a longitudinal study of autistic children . Paper presented at the biennial

meeting of the Society for Research in Child Development , Kansas City .

Tomasello , M . ( 1995 ) . On the interpersonal origins of self - concept . In U . Neisser

( Ed . ) , The perceived self ( pp . 174 - 184 ) . New York : Cambridge University Press .

Uemura
. ,M.(1979
44 ).Development
ofego
ininfancy
.Shinrigaku
Hyoron
,22
,28
-
First and SecondPersonPronouns 409

Wales, R. (1979 ). Deixis. In P. Fletcherand M. Garman(Eds.), Languageacqui-

sition(pp. 401- 428). New Yark: CambridgeUniversityPress .
Wijnen, F., Krikhaar, E., and Den Os, E. (1994 ). The (non) realizationof un-
stressedelementsin children's utterances : Evidencefor a rhythmic constraint.
Journalof ChildLanguage , 21) 59- 83.
Wills, D. D. (1977 ). Participantdeixisin Englishandbabytalk. In C. Snowand
C. Ferguson(Eds.), Talkingto children(pp. 271- 295). Cambridge : Cambridge
UniversityPress .
Xu, F., andCarey, S. (1996 ). Infants' metaphysics
: Thecaseof numericalidentity.
CognitivePsychology , 30) 111- 153.
Chapter
- 16

Kinship...and Mathematical F. William Lawvere

Categories

Those conceptsthat are historically stable tend to be those that in some

way reflect reality, and in turn tend to be those that are teachable.
" Mathematical" should mean in particular " teachable," and the modern
mathematical theory of flexible categoriescan indeed be used to sharpen
the teachability of basic concepts. A concept that has enjoyed some his-
torical stability for 40,000 years is the one involving that structure of a
society that arises from the biological process of reproduction and its
reflection in the collective consciousnessasideasof genealogyand kinship.
Sophisticatedmethods for teaching this concept were devised long ago,
making possible regulation of the processitself. A more accurate model
of kinship than heretofore possible can be sharpenedwith the help of
the modern theory of flexible mathematical categories; at the sametime,
established cultural acquaintance with the concept helps to illuminate
severalaspectsof the generaltheory, which I will therefore try to explain
concurrently.
Abstracting the genealogicalaspectof a given society yields a mathe-
matical structure within which aunts, cousins, and so on can be precisely
defined. Other objects, in the samecategory of such structures, that seem
very different from actual societies, are nonethelessshown to be impor-
tant tools in a society's conceptualizing about itself, so that for example
gender and moiety become labeling morphisms within that category.
Topological operations, such as contracting a connected subspaceto a
point, are shown to permit rationally neglectingthe remote past. But such
operations also lead to the qualitative transformation of a topos of pure
particular Becoming into a topos of pure general Being; the latter two
kinds of mathematical toposesare distinguishedfrom each other by pre-
cise conditions. The mythology of a primal couple is thus shown to be
a naturally arising didactic tool. The logic of genealogyis not at all 2-
412 La wv ere

valued or Boolean, becausethe truth-value space naturally associated

with the ancestorconcept has a rich lattice structure.
As John Macnamara emphasized, an explicit mathematical framework
is required for progressin the scienceof cognition, much as multidimen-
sional differential calculushas beenrequired for 300 yearsas a framework
for progress in such sciencesas thermomechanicsand electromagnetics .
I hope that the exampIes consideredhere will contri bute to constructing
the sort of generalframework John had in mind.
Although a framework for thinking about thinking would be called
" logic" by someancient definitions, in this century logic has come to have
a much more restrictive connotation that emphasizesstatementsas such,
rather than the objects to which the statementsrefer; this restrictive con-
notation treats systemsof statementsalmost exclusively in terms of pre-
sentations(via primitives and axioms) of such systems, thus obscuring the
objective aspectsof abstract generalsthat are invariant under change of
presentation. About 100yearsago the necessaryobjective support for this
subjectivelogic began to be relegatedto a rigidified set theory that shares
with mereology the false presupposition that given any two sets, it is
meaningful to ask.whether one is included in the other or not. Although
the combination of narrow logic and rigidified set theory is often said to
be the framework for a " foundation of mathematics," it has in fact never
servedas a foundation in practice; for example, the inclusion question is
posed in practice only betweensubsetsof a given set, such as the set of
real numbers or the set of real functions on a given domain. There are
many different explicit transformations betweendifferent domains, but the
presumption that there is a preferred one leads to fruitless complications.
The development of multidimensional differential calculus led in par-
ticular to functional analysis, algebraicgeometry, and algebraictopology,
that is, to subjectswhose interrelationships and internal qualitative leaps
are difficult to account for by the narrow and rigidified " foundations."
This forced the development of a newer, more adequate framework
(which of course takes full account of the positive results of previous
attempts) 50 years ago, and 20 years later the resulting notions of cate-
gory, functor, natural transformation, adjoint functor, and so on had
becomethe standard explicit framework for algebraic topology and alge-
braic geometry, a framework that is even indispensablefor the commu-
nication of many concepts. Then, inspired by developmentsin mechanics
and algebraic geometry, it was shown how a broader logic and a lessrigid
notion of set are naturally incorporated in the categorical framework.
Kinship and Mathematical Categories 413

Toposesare categoriesof setsthat have specifiedinternal cohesionand

variation and whose transformations are continuous or equivariant. The
infinite variety of toposesnow studied arisesnot in order to justify some
constructivist or other philosophical prejudice, but in order to make use-
fully explicit the modes of cohesion and variation that are operative in
multidimensional differential calculus. The internal logic of toposesturned
out to be that which had beenpresentedearlier by Heyting. Heyting logic
(and co-Heyting logic) describesinclusions and transfoffi1ationsof subsets
of such sets in a way that takes account of the internal cohesion and
variation, refining the fragmented and static picture that the still earlier
approximation by Boolean logic provides.
I will explore here two examplesshowing how the categorical insight
can be usedto make more explicitly calculable, not only the more overtly
spatial and quantitative notions, but also other recurring generalconcepts
that human consciousnesshas perfected over millennia for aiding in the
accurate reflection of the world and in the planning of action. The two
examplesare kinship, in particular, and the becoming of parts versusthe
becoming of wholes, in general. Since the narrowing of Logic 100 years
ago, it has been customary to describe kin ships in teffi1s of abstract
" relations" ; however, if we make the theory slightly less abstract, the
category of examplesgains considerablyin concretenessand in power to
representconceptsthat naturally arise. Becoming has long been studied
by resolving it into two aspects, time and states, with a map of time
into the states; but when we consider the becoming of a whole " body"
together with that of its parts (e.g., the development of collective con-
sciousnessor the motion of the solar system), there are several" conflict-
ing" aspectsthat we need to manage: the collective state " is" nothing but
the ensembleof the statesof the parts, yet the constitutive law deteffi1in-
ing what the state (even of a part) " becomes" may dependon the state of
all; the time of the whole may be taken as a prescribedsynchronization of
the times of each part, yet the news of the collective becoming may reach
different parts with differing delays. A general philosophical idea is that
the cohesivenessof being is both the basis in which these conflicts take
place and also partly the result of the becoming (Hegel: " Wesen ist
gewesen " - that is, the essenceof what there is now is the product of
the processit has gone through). I will try to show more precisely how
the being of kinship in a given historical epoch relatesto the becoming of
reproduction and of intermarriage betweenclans, even when this kinship
and reproduction are consideredpurely abstractly, neglectingtheir natural
and societal setting.
414 Lawvere

16.1 A ConcreteCategory Abstracting the Notion of Reproduction

I want to consider general notions of kinship system, as suggestedby

known particular examples. The following fundamental notion was ela-
borated in collaboration with Steve Schanuel (Lawvere and Schanuel
1997).
An analysis of kinship can begin with the following biological
observation:

(T1) Each individual has exactly one mother and exactly one father,
who are also individuals.

As a first approximation, we can explore the ramifications of this idea,

taken in itself, as an abstract theory T1. Though we may have thought
originally of " individual " as meaning a member of a particular tribe,
there is in fact in a " theory" no information at all about what individuals
are; the mathematical concretecorrespondingto the statedabstract theory
Tl is the category whose objects are sets, consisting of abstract elements,
but structured by two specifiedself-mappings m and f Such a structured
set can be considered as a " society." We can immediately define such
concepts as " a and b are siblings" by the equations am == bm and af ==
bf , or " b's maternal grandfather is also c's paternal grandfather" by the
equation bmf == cff . (In the Danish language 'morfar' means mother's
father, 'farfar' means father's father, similarly 'farmor' and 'mormor';
and young children commonly address their grandparents using those
four compound words.) There is a great variety of such systems, most of
which seemat first glance to be ineligible, even mathematically, for more
than this verbal relation to the kinship concepts; for example, I cannot
be my own maternal grandfather. However, if we take the theory and the
category seriously, we will find that some of these strange objects are
actually useful for the understandingof kinship.
As is usual with the concrete realizations of any given abstract theory,
theseform a category in the mathematical sensethat there is a notion of
structure-preservingmorphism X - t Y betweenany two examples: in this
casethe requirement on ljJis that jJ(xm) == jJ(x)m and ljJ(xf ) == jJ(x)f for
all x in X. (It is convenient to write morphisms on the left, but structural
maps on the right of the elements; in the two equations required, the
occurrencesof m and f on the right sides of the equality refer to the
structure in the codomain Y of the morphism.) There are typically many
morphisms from given domain X to given codomain Y. If X - t Y ! z
Kinship and Mathematical Categories 415

aremorphisms , thereis a well-determinedcompositemorphismX ~ Z ,

and for eachobjectX, thereis alwaysthe identity morphismfrom X to
itself. Two objectsXl and X2 are isomorphicin their categoryif there
existsa pair of morphismsXl --t X2 --t Xl with both composites equalto
therespective identitymorphisms . Most morphismsarenot isomorphisms ,
and they indeeddo not even satisfythe requirementof " injectivity,"
which just meansthat the cancellationlaw, ljJXl== fjJx2impliesXl ==X2,
holdsfor fjJ. If a morphism~ is in fact injective, it is alsocalled" a sub-
objectof its codomain,whichtakenin itself is the domain." (Contraryto
somerigidified versionsof set theory, a given objectmay occur as the
domainof manydifferentsubobjects of anothergivenobject. A subobject
may be thoughtof as a specifiedway of "including" its domaininto its
codomain.) An elementy is saidto "belongto" a subobject~ (notation:
y G~) if there is x for which y == ~x; this x will be unambiguousbecause
of the injectivity requirement on ~. Here by an element of X is often
meant any morphism T - t X ; such elementsare also referred to as figures
in X of shape T, but often we also restrict the word 'element' to mean
'figure of a sufficient few preferred shapes'. Sometimeswe have between
two subobjects an actual inclusion map ~I S; ~2 as subsetsof Y (i .e.,
Xl ~ X2 for which ~l == ~2Ct ). It is theseinclusions that the propositional
logic of Y is about.

16.2 GenealogicalTruth, SeenTopos-Theoretically

The category sketchedabove, that is, the mathematically concretecorre-

spondent of the abstract theory TI of m and/, is in fact a " topos." That
means, among other properties, that there is a " truth-value" object Q that
" classifies," via characteristicmorphisms, all the subobjectsof any given
object. In the simpler topos of abstract (unstructured) sets, the set 2,
whose elementsare " true" and " false," plays that role; for if X ~ Y is
any injective mapping of abstract sets, there is a unique mapping Y ~ 2
such that for any elementy of Y, rPy == true if and only if y belongs to
the subset~; and if two maps rPI, rP2from Y to 2 satisfy rPIentails rP2in
the Boolean algebra of 2, then the corresponding subsetsenjoy a unique
inclusion map ~l ~ ~2 as subsetsof Y (i .e., Xl ~ X2 for which ~l == ~2a).
However, in our topos, 2 must be replaced by a bigger object of truth-
values in order to have those features for all systemsY and for all sub-
systemsX, ~, as explained below.
In our tapas of all Tl -societies, there is a particular object I , which as an
abstract set consistsof all finite strings of the two symbols m and f, with
416 Lawvere

the obvious right -action that increases length by 1, giving its cohesiveness
and variation ; in particular , the empty string 1I is the generic individual ,
all of whose ancestors are distinct . A concrete individual in a society X is
any morphism I ---+ X ; in particular , the infinitely many endomorphisms
I ~ I are easily seen to be in one-to -one correspondence with the strings
in I . Composing with the two endomorphisms m and f , which correspond
to the two strings of length 1, completely describes, in the context of the
whole category , everything we need to know about the particular internal
structure of X : the actions xm , x f are realized as special cases of the
compositionof morphisms
. The conditionthat everymorphismX ! Y
in the topos must satisfy is then seen to be just a special case of the asso-
ciative law of composition , namely , the special case in which the first of
the three morphisms being composed is an endomorphism of I . The facts
stated in this paragraph constitute the special case of the Cayley -Y oneda
lemma , applied to the small category W that has one object , whose
endomorphisms are the strings in two letters , composed by juxtaposition ;
a category with one object only is often referred to as a monoid .
With loss in precision we can say that x ' is an ancestor of x in case there
exists some endomorphism w of I for which x ' == xw . In the particular
society X , the w may not be unique ; for example , xmff == xfmf might
hold. But the precision is retained in the category W/ X , discretelyfibered
over W, which is essentially the usual genealogical diagram of X . The
correlation of the reproductive process in X with past solar time could be
given by an additional age functor from this fibered category to a fixed
ordered set; a construction of Grothendieck ( 1983) would permit inter -
nalizing such a chronology too in Tl 'S topos .
N ow we can explain the truth -value system. The subsets of I in the
sense of our topos are essentially just sets A of strings that , however , are
not arbitrary but subject to the two conditions that if I ~ A is a member
of A , then also wm and wf must be members of A . Thus , A can be
thought of as all my ancestors before a certain stage, the antiquity of that
stage depending however on the branch of the family . These As constitute
the truth -values! To justify that claim , we must first define the action of m
andfon As ; this is what is often thought of as " division " :
A : m == { w s W : mw sA } ,
A : f == { w G W : fw GA } .
It is easily verified that these two are again subobjects of I if A is. Note
that A : f consists of all my [ather 's ancestors who were in A . The truth -
Kinship and Mathematical Categories 417

value true is A :=: W, the whole set of strings, and union and intersection
A u B, A n B are the basic propositional operations or, and on truth -
values. The operation not A of negation does not satisfy all the Boolean
laws, sinceit must again yield a subobjectof I :
not A == { WB W : for all v B W, wv does not belong to A }
is the set of all my ancestorsnone of whose ancestorswere in A, so (A
or not A) is usually much smaller than true. Why does the object Q thus
defined serveas the unique notion of truth -value set for the whole topos?
Consider any subset X ~ Y in the senseof the topos; then there is a
unique morphism Y ~ Q such that for all individuals I ~ Y in Y,
y G~ if and only if cjJ
(y) == true.
The value of </>on any elementy is forced to be
tj>(y) == { W8 W : yw 8 ~} .
Thisis themeasuredanswer , to the question''Preciselyhow Irish is y?"

16.3 How to Rationally Neglect the RemotePast

Every X is a disjoint union of minimal componentsthat have no mutual

interaction; if we parameterizethis set of componentsby a set IIo (X ) and
give the latter the trivial (identity) action of m, f, then there is an obvious
morphism X --+ IIo (X ). If llo (X ) :=: 1, a one-point set, we say X is con-
nected; that doesnot necessarilymean that any two individuals in X have
some common ancestor, becausemore generally connecting can be veri-
fied through cousinsof cousinsof cousins, and so on.
An important construction borrowed from algebraic topology is the
following :
Given a subobject A ~ X , we can form the " pushout" X mod A that
fits into a commutative rectangular diagram involving a collapsing mor-
phism X --t XmodA and an inclusion llo (A) ~ XmodA as well as the
canonical A --t llo (A) and satisfiesthe universal property that, if we are
given any morphism X - t Y whose restriction to A depends only on
no (A ), then there is a unique morphism XmodA ~ Y for which t/I is the
composite, t/I0 following X - + X mod A, and which also agreeson no (A)
with that restriction.
This X mod A is a very natural thing to consider becausepractical
genealogicalcalculations cannot cope with an infinite past. For example,
the Habsburg X are mainly active over the past 1,000 yearsand the Orsini
 418 Lawvere

ovrer the past 2 , 000 years . But this collapsing of the remote ancestors A to

IIo (A ) introduces an interesting idealization ; for simplicity we consider

the case where A is connected , IIoA == 1 . Then the construction has given

rise to a morphism 1 - * X mod A that is the residuum of A . But what is a

" point " ( = = figure of shape 1) in a topos like ours ? Since m and fact

as identity in 1, a point ( e . g . , the point of X mod A that has arisen ) is a

special sort of individual x that is its own mother and father :

xm == x == xi .

This idea of a superindividual is of course merely a convenience in genea -

logical bookkeeping .

16 . 4 Deepening the Theory to Make Gender Explicit

A reasonable gender structure is not definable in the abstract theory T 1

above , since there is no requirement that the values of the operator m be

disjoint from the values of the operatorfand also no way of determining

the gender of infertile individuals . So as usual we refine our theory to a

richer one :

( T 2) In addition to T 1 , every individual has a definite gender , every

individual ' s mother is female , and every individual ' s father is male .

The corresponding category of all concrete applications of T2 will also be

a topos and in fact have a direct relation to the topos of TI ' Na ~ ely , in

the first topos there is a particular object G with only two individuals

called male andfemale and where the two structural operations act as the

constant maps . Our refined topos has objects that may be described as

pairs X , y where X ~ G is a morphism in the first tapas and has as mor -

phisms all triangular diagrams [X ~ G , X ~ Y , Y ~ G ] in the first tapas

for which Jtj ; == y ; the codomain of this triangle is defined to be the pair Y ,
J . The conditions

y ( xm ) == ( yx )m

and

y ( xf ) = = ( yx ) f

that morphisms must satisfy state in this case that y is a compatible

gender - labeling , and the further condition on a triangular diagram means

that morphisms in our refined tapas moreover preserve gender . That

applies in particular to the individuals in X that happen to be ( in our

Kinship and Mathematical Categories
419

purely reproductive sense ) spinsters or bachelors . This tapas has two

generic individuals , because labeling maps I - + G can map 11 either to m

or to ! The terminal object of this new tapas is actually IG so that a point

in the new sense is really the " locus of a point moving virtually between m

and f " ; more exactly , a morphism IG - + Y in this tapas amounts to an

Eve / Adam pair for which

e f = = a f = = a & em = = am = = e .

If we construct the pushout along A - + IIoA of a subobject A < : - + X in the

sense of this topos , we get in particular lloA C - ? - X / A , which specifies a

different Eve / Adam pair for each of the mutually oblivious subsocieties

among the specified ancestors .

16.5 Moieties

A still further refinement makes precise the idea of a society equipped

with a strategyfor incestavoidance. The simplestidea T3 of suchis that of
(matrilineal) moiety labeling, which can be explained in terms of another
two-individual application C of T 1. The two " individuals" are called Bear
and Wolf in Lawvere and Schanuel1997. The idea is that the moiety of
any individual x is the sameas that ofx 's mother and distinct from that of
x's father. Thus, on the two individuals in C, m acts as the identity,
whereasf2 == id but f # id. Any TI-respecting morphism X - + C is a
labeling with the expectedproperties, and commuting triangles over C
constitute the morphisms of a further topos that should be investigated.
However, I will denote by T3 the richer abstract theory whose applica-
tions are all the TI-applications X equipped with a labeling morphism
X - + G x C. Thus, in T3's topos there are four genericindividuals I (one
for each pair in G x C), morphisms from which are concreteindividuals
of either genderand of either Bear or Wolf moiety.

16.6 Congealing
Becominginto Being

According to the criteria proposed in Lawvere 1991, the three toposes

thus far introduced representparticular forms of pure becoming, because
they satisfy the condition that every object X receivesa surjective mor-
phism from another one E that has the specialproperty
ZlW== Z2W implies ZI == Z2
 420 Lawvere

for any given w in Wand any two individuals Zl , Z2 in E . This " separable

covering " E of X is achieved through refining each individual ( if neces -

sary ) into several individuals with fonnally distinct genealogies ( e . g . ,

xmf # yf in E even if y = = xm in X ) . However , by restricting to certain

subcategories consisting only of objects that contain something roughly

like the Eve / Adam pair , we obtain toposes that have instead the qualita -

tive character of pure being ; for example , the Habsburg era had a certain

quality of being ( and was the basis for a lot of motion ) .

More precisely , let us consider various theories corresponding to

monoid homomorphisms

W ~ M

that are epimorphic . Epimorphic homomorphisms may be surjective

( induced by a congruence on W ) , or may adjoin at most new operator

symbols that are two - sided inverses of operator symbols coming from W ,

or may involve a combination of these two kinds of " simplifications " of

W . Such a homomorphism gives rise to a well - defined subcategory of Tl ' S

topos , namely , the one consisting of all those systems X satisfying the

congruence or invertibility conditions that become true in M . ( For anal -

ogous subcategories of the other two toposes , we would need to consider

epimorphic functors

W / G ~ D

and

W / Gx C ~ E

to small categories D or E of two ( respectively four ) objects . ) The result -

ing subcategories , although toposes , are not subtoposes but " quotient

toposes . " Here a quotient morphism of toposes , compressing a bigger

situation into a smaller situation , involves a full and faithful inverse

functor p * ( the obvious inclusion in our example ) and two forward func -

tors p ! and p * that are respectively left and right adjoint to p * in the sense

that there are natural one - to - one correspondences

PIX - * Y Z - * p * X

X - * p * Y p * Z - * X

between morphisms , for all sets X defined over the big situation and all

Y , Z defined over the smaller one . Note that the notion of morphism is

the same in both situations ; that is what it means to be full and faithful .
Kinship and Mathematical Categories 421

These adjointnesses are the refined objective version of relations whose

subjective reflections are the rules of inference for existential quantifica -
tion p !, resp. universal quantification p * of predicates relative to a sub-
stitution operation p * (indeed , p * preserves coproducts and products of
objects, as is subjectively reflected in the fact that a substitution preserves
disjunction and conjunction of predicates ; it may fail to preserve the
analogous negations however , for example in the case of continuous
predicates in topology ). Specifically , p *X extracts those individuals from
X whose genealogy happens to conform to the requirements specified in
M , whereas PIX applies to all individuals in X , the minimal forcing re-
quired to merge them into a new object that conforms to M .
These facts will be used below in investigating elementary examples
of quotient toposes, in particular , some of those that seem to congeal
becoming into being .
To distinguish a topos of general pure being , I proposed in La wvere
1986 and 1991 the two criteria that llo (X x Y) == llo (X ) x llo ( Y) and
that every object is the domain of a subobject of some connected object .
The first of these means that any pair consisting of a component of X and
a component of Y comes from a unique single component of the Cartesian
product X x Y ; in particular , the product of connected objects should be
again connected (since 1 x 1 == 1), which is rarely true in a tapas of par -
ticular becoming . That condition will be important for the qualitative ,
homotopical classification of the objects, but I will not discuss it further
here .

The second criterion mentioned above, that every object can be

embedded as a subobject of some connected object , has the flavor " We
are all related ," but is much stronger than merely requiring that a partic -
ular object be connected ; for example , the disjoint sum of two connected
objects, which is always disconnected , should be embeddable in a con -
nected object .
It follows from a remark of Grothendieck ( 1983) that both proposed
criteria will be satisfied by a topos of M -actions if the generic individual I
(== M acting on itself ) has at least two distinct points 1 ~ I . The distinct -
ness must be in a strong sense that will be automatic in our example
where 1 has only two subobjects; it is also crucial that the generic indi -
vidual I itself be connected . Part of the reason why this special condition
of Grothendieck implies the two general criteria can be understood in the
following terms : it is not that becoming is absent in a topos of general
being ; rather , it is expressed in a different way . An object X may be a
422 Lawvere

space of locations or states in the tradition of Aristotle ; that is , X is an

arena in which becoming can take place. In particular , some such spaces
T can measure time intervals and hence a particular motion in X during T
may give rise to a morphism T ~ X describing the result of the becoming
process; that is, if T is connected , if we can distinguish two instants I 4 T
called to , tl and if jltk == Xk for k == 0 , 1, then we can say that Xo became
Xl during the process 11; this is the common practice in mathematical
engineering (at least for certain categories within which it has, in effect,
become customary to work in the past 300 years). But then if we consider
tl as a subobject of T, its characteristic map from T to the truth -value
space will be a process along which false becomes true ; this in turn implies
that the truth -value space is itself connected since T is , and because of the
propositional structure of truth -values, it follows that indeed the hyper -
space, whose points parameterize the subspaces of X , is connected too . On
the other hand , in any topos any object X is embedded as a subobject of
its hyperspace via the " singleton " map . Taking T == I , it follows that in
the cases under discussion every object can be embedded in an object that
is not only connected , but moreover has no holes or other homotopical
irregularities . Note that now each individual I ~ X involves also a pro -
cess in his or her society. There are many such situations , but what can
such a process mean ?

As a simple example , suppose that we want a category of models of

societies in which genealogical records go back to grandparents . Thus , we
consider the homomorphism W ~ M onto a seven-element monoid in
which

m3 == mlm == m2
and

ml2 ==m2I ==ml

hold , together with similar equations withfand m interchanged . Drawing
the obvious genealogical diagram , we see that the generic individual / in
the topos of M actions has parents that represent the first intermarriage
between two distinct lines, and that / 's grandparents are the " Eves and
Adams " of those two distinct lines . Indeed , this object [ permits a unique
gender structuring and so we may pass instead into the topos whose theory
is M / G, finding there that each of the two genericindividuals [ G has two
distinct points IG : 4 [ G. The " process" involved in an individual boy
[ G ~ X in a society X in this tapos is that whereby his father 's line united
with his mother ' s.
Kinship and Mathematical Categories 423

16.7 Finer Analysis of the Coarser Theories

Given any object L in a topos, one can form a new topos of objects fur-
ther structured by a given labeling morphism to L . However, it is worthy
of note that in our examples, one L == G (leading to T2) and the other one
L == G x C (leading to T3), these labeling objects belong to a much
restricted subcategory(quotient topos) within the topos of all Tl -objects.
Although the Tl topos, consisting of all right actions of the free monoid
Won two symbols, is too vast to hope for a complete survey of its objects
(as is borne out by theorems of Vera Trnkova stating that any category
can be embeddedas a full subcategoryof it ; seePultr and Tmkova 1980),
by contrast thesesmaller toposesare more tractable.
Consider first the case G of gender in itself. It belongs to the sub-
category correspondingto W - + W3, the three-element monoid in which
xl == I , xm == m for all three x == 1, I , m. If we restrict attention to the
right actions of W3 only, we seethat the genericindividual I in the sense
of that topos has only one nontrivial subobject, which is G itself. There-
fore, the truth -value space.Q3for that topos (in contrast to the infinite .Q
for W) has only three elements also. The most general " society" X is
a sum of a number of noninteracting nuclear families, some with two
parents and somewith a single parent, and a society is determined up to
isomorphism by the double coefficient array that counts families of all
possiblesizesof thesetwo kinds. Cartesian products are easily computed.
Second, consider the simple moiety-labeler C in itself; it belongsto the
quotient topos determinedby W - + W2, the two-elementgroup generated
by I with 12 == 1, where we moreover interpret m as 1. Here C is itself
the generic individual and has no nontrivial subobjects, so that 02 is the
two-element Boolean algebra. All right W2-actions are of the form X ==
a + bC, where a is the number of individuals fixed, and b half the number
moved, by ! Theseobjectsare multiplied by the rule c2 == 2C.
Third , we can include both G and C in a single subcategoryas follows:
the infinite free monoid W maps surjectively to each of W3 and W2;
therefore, it maps to the six-element product monoid W3 x W2, but not
surjectively since the image Ws is a five-element submonoid (the missed
elementis the only nonidentity invertible elementin the product monoid).
The category of right Ws-setscan be probed with the help of its generic
individual I , which again (surprisingly) has only three subobjectsso that
Qs has again three truth-value-individuals. In I (i.e., Ws) there are besides
me my four distinct grandparents, the Tl -structure coming from the fact
424 La wvere

that my parents are identical with my maternal grandparents in this

theory . The unique nontrivial subobject of I is precisely G x C, the label
object for our central noncollapsed theory T 3; it is generated by any single
one of its four elements . I leave it to the reader to determine a structural

description of all right Ws-actions , with or without G x C labelings .

16 .8 Possible Further Elaborations

To the T2 topos (whose abstract general is the two -object category W / G)

we can apply separately our concrete construction idealizing remote
ancestors and also our refinement of the abstract general by moiety -
labeling . However , it does not seem consistent to apply both of these
simultaneously , owing to the simple group -theoretic nature of our C. A
reasonable conjecture is that tribal elders in some part of the world have
devised a more subtle abstract general C ' that relaxes the incest avoidance
for the idealized remote ancestors, thus permitting a smoothly functioning
genealogical system enjoying both kinds of advantages.
Some more general examples of simplifying abstractions include that
based on a homomorphism W ~ M n that makes every string of length
greater than n (where , e.g., n == 17) congruent to a well -defined associated
string of length n. An interesting construction to consider is the appli -
cation of the pushout construction to an arbitrary Tl -object X , not with
respect to some arbitrarily chosen remoteness of ancestry A , but with
respect to the canonical morphism p * p * X --+ X .
The need for Eve and Adam arises from the central role of self -maps as
structure in the above theories and might be alleviated by considering
instead a two -object category as the fundamental abstract general, although
possibly at the expense of too great a multiplicity of interpretations . This
category simply consists of three parallel arrows , as described in Lawvere
1989. A concrete application of this theory To involves two sets (rather
than one) and three internal structural maps
X no~' - - t X up to now

called m, f , and 8, where s specifies, for each contemporary individual , the

place of his or her " self " in genealogical history . There is a natural notion
of morphism between two such objects (involving two set-maps subject to
three equations ) yielding again a topos , whose truth -value object is finite
and illuminating to work out . To's topos has Tl 's topos as a quotient ,
because forcing the structural map 8 to become invertible yields a small
Kinship and Mathematical Categories 425

category equivalent to the monoid W . If s is not inverted , iteration of

m andfis a more complicated partial affair : in case xm has the property

that there is a contemporary mother y for which ys == xm , then yf is a

maternal grandfather of x ; but there may not be such a y , and on the

other hand , mathematical experience counsels against excluding in gen -

eral the possibility of several such y for a given x . If we do invert s , we

make explicit that which , in a larger sense , is the theme of this book , the

continuing role of each past individual in our cognition at present .

References

Grothendieck , A . ( 1983 ) . Pursuing stacks . Unpublished manuscript , University of

Montpellier , France .

Lawvere , F . W . ( 1986 ) . Taking categories seriously . Revista Colombiana de

Matematicas , 20 , 147 - 178 .

Lawvere , F . W . ( 1989 ) . Qualitative distinctions between some toposes of general -

ized graphs . In J . Gray and A . Scedrov ( Eds .) , Logic , categories , and computer

science ( Contemporary Mathematics 92 , pp . 261 - 299 ) . Providence , RI : American

Mathematical Society .

Lawvere , F . W . ( 1991 ) . Some thoughts on the future of category theory . In A .

Carboni , M . C . Pedicchio , and G . Rosolini ( Eds .) , Category theory ( Lecture

Notes in Mathematics 1488 , pp . 1- 13 ) . Heidelberg : Springer - Verlag .

Lawvere, F. W ., and Schanuel, S. ( 1997 ) . Conceptual mathematics . Cambridge :

Cambridge University Press.

Macnamara, J., and Reyes, G. E. (1994). The logical foundations of cognition.
Oxford: Oxford University Press.
Pultr, A ., and Tmkova , V . (1980). Combinatorial, algebraic, and topological rep-
resentationsof groups, semigroups , and categories. Amsterdam: North -Holland .
Chapter 17
Count Nouns, Mass Nouns, Marie La Palme Reyes,
and Their Transformations: John Macnamara,
A Unified Category- Gonzalo E. Reyes, and
Theoretic Semanti cs H 0timan Zo If aghari

17.1 Introduction

17.1.1 Prototypical Count Nouns and Mass N onns

All natural1anguagesseemto distinguish at the semantic level between
count nouns (CNs) and massnouns (MNs ). Somenatural languages, like
English, mark the distinction at the syntactic level. Prototypical of CNs is
'dog' and of MN s is 'matter' (in the senseof physical stuff, not in the
senseof concern or affair). One syntactic difference is that usually CN s
take the plural ('dogs') whereasMNs do not. Other syntactic distinctions

This chapterhasbeen"in the making" for a numberof years, and it would be

difficult to thank all the peopleand institutionsthat have given us criticism,
advice, and encouragement . In spiteof this, we would like to singleout the fol-
lowing: John Macnamaraand GonzaloE. Reyesgratefully acknowledgethe
supportof individual grantsfrom Canada ' s Natural Sciencesand Engineering
ResearchCouncil(NSERC) and of a collaborativegrant from Quebec 's Fonds
pour la formationdechercheurs et l'aidea la recherche (FCAR). MarieLa Palme
Reyesis gratefulfor the supportof a SocialSciences and HumanitiesResearch
Council of Canada(SSHRC) postdoctoralfellowshipduring the early yearsof
research , as well as supportfrom Macnamara ' s NSERC grant and the FCAR
grant to Macnamaraand Reyesduring the later years. HoumanZolfaghariwas
partly supportedfrom the NSERC grantsof Macnamaraand Reyes . GonzaloE.
Reyeswould like to acknowledge an invitationfrom the Departmentsof Mathe-
maticsandPhilosophyof the UniversiteCatholiquede Louvain-la-Neuve, where
partsof this chapterwerepresented . BrendanGillon and JeffPelletierhelpedus
to find our way throughthe massof paperson massnounsandcountnouns, and
we aregratefulto them. JoyceMacnamarawaskind enoughto improveour text
anddiscussintuitionsrelatingto manylinguisticexamples . Finally, we wouldlike
to acknowledge valuableconversations with AndersKock, Mihaly Makkai, and
RichardSquireon the subjectof this chapter.
428 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

relate to determinersand quantifiers. One can say 'a dog', 'another dog',
'many dogs', 'two dogs'; one cannot correctly say *'a matter', * 'another
matter', * 'many matter', * 'two matter'. It seemsthat the distinction in
English grammar was first describedby Jespersen(1924, 198).
Languagesdiffer in morphology, agreementrules, and phrasestructure,
so one does not expect to find in every natural language a count/mass
distinction with the same linguistic correlates as in English. Although
many European languagesare like English in this connection, not all
are. Irish and Latin , for example, lack the indefinite article, and so one
cannot distinguish CN s from MN s in those languagesby the possibili ty
or impossibility of adding the indefinite article to the noun. Japanesehas
neither definite nor indefinite articles and lacks a uniform way to build
plurals. Yet Japaneserequires classifiers to license the application of
certain quantifiers such as numerals to nouns. Thus, in order to apply the
numeral 'ni' ('two') to the noun 'inu' ('dog'), the classifier'hiki ' is required
to form the expression'Inu-ga ni-hiki iru ' ('There are two dogs'). This
classifier is required also for words denoting fishes and insects, although
not for words denoting birds, for which the classifier 'wa' is required. On
the other hand, with almost any count noun denoting inanimate entities
the classifier 'tsu' is required, as in the expression'Ringo-o yo-tsu tabeta'
('I ate four apples'). This classifiercannot be usedwith massnouns. With
mass nouns such as 'mizu' ('water'), a different classifier 'hai' or 'bai' is
required to form expressions such as 'Mizu-o san-bai (hai) nonda'
('I drank three glassesof water'). This classifiercannot be usedwith count
nouns or with other massnouns such as 'nendo' ('clay'). Thus, the correct
use of some classifiers seemsto require a count/mass (as well as an
animate/inanimate) distinction at the level of nouns. (We owe the details
about Japaneseto Yuriko Oshima-Takane.)
Incidentally, if we inquire what guides linguists to the decision that
there are common nouns in languagesother than English, the answer
cannot simply be grammar. As just illustrated, grammar varies greatly
from language to language. Something other than grammar must be
contributing to the decision. We submit that the type of referencefor
prototypical words plays a major role. We think that acrossvariations in
grammar there is a semantic uniformity in the interpretation of at least
such prototypical count nouns as 'dog' as well as in the interpretation of
such prototypical mass nouns as 'matter' and that this semantic unifor-
mity is a good guide to the relevant grammatical facts in each language.
We propose to exploit the semanticuniformity as far as possible.
Count Nouns and Mass Nouns 429

This should not be taken as an assumptionthat the perceptual experi-

ence of a noun's extension determineswhether a noun applied to it is
count or mass. It clearly does not. The very sameperceptual experience
that licensesthe application of the MN 'gravel' licensesthat of the CN
'pebble' . Moreover, languagesvary in what they chooseto presentin the
first instance as mass and what as count. French speakers, for example,
apply the CN 'meuble' where English speakerswould normally apply the
MN 'furniture'. Notice that 'furniture' doessupply natural units, or as we
say articles of furniture (chairs, tables, lamps, etc.). The importance of the
prototypical examplesalso showsup in work on child languagelearning.
Young children tend to take a new word taught them for a stuff that is
like sand in its consistencyas a MN , and a word for unfamiliar creatures
reminiscent of animals as a CN (see especially McPherson 1991; Soja,
Carey, and Spelke 1991).
Pelletier and Schubert (1989) document the difficulty in deciding what
precisely falls under the classificationsmass or count: nouns and noun
phrases, adjectives and adjectival phrases, verbs and verb phrases, or
adverbs and adverbial phrases? Even if one could settle on the relevant
syntactic category or categories, there is still the issueof what preciselyin
thosecategoriesis massor count: the expressionsthemselves , their senses
,
or their occurrences ? All this in addition to the problem of how to decide
whether something, whatever that somethingis, is massrather than count.
But notice that to start work on thesequestionsassumesthat we under-
stand the count/mass distinction. On the basis of the prototypical exam-
ples discussedabove, we take the view that this distinction applies to
nouns, whether or not it applies to other grammatical categories. Our
work is concernedonly with the count/massdistinction for nouns, whose
solution is presupposedin formulating some of the previous questions.
Despitetheir obvious interest, thosequestionswe do not proposeto tackle.

17.1.2 Cumulativeand Distributive Reference

The usual way of distinguishing MNs from CNs is by specifyingsemantic
properties that MNs have but CNs lack.
Two or more dogs do not together constitute a larger dog, whereas
two or more quantities of matter together constitute a larger quantity of
matter. It is customary, following Quine (1960, 91), to refer to this prop-
erty of the extensionof MNs as cumulativereference. Likewise, a (proper)
part of a dog is not a smaller dog, whereasa part of matter is matter in
the sensethat a part of a portion of matter is a smaller portion of matter.
430 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

Many writers refer to this semanticproperty of MN s as homogeneous(or

distributed or divided or divisive) reference. All the authors that we have
consulted accept the cumulative reference(as we do), but the property of
divisivenessof referencedivides them!
What is our position in this debate? Our position can be best formu-
lated by paraphrasing the great Mexican comic Mario Moreno, Cantin-
flas: " We are neither for nor against, but quite the opposite." In fact, if in
accordancewith the test of cumulative referencewe interpret a MN as a
sup-lattice M (i .e., a poset suchthat every subsethas a least upper bound),
then it makes no senseto ask whether a part of an elementof M is again
an element of M . What could be a part over and above the elementsof
M ? We believe that this discussionpresupposes(perhapsinadvertently) a
universal substance, a sup-lattice containing every interpretation of a MN
as a sub sup-lattice. With such a substanceit makes senseto ask whether
every element of the universal substance, which is a part of an elementof
M , is an element of M . We reject the notion of a universal substance,
however, and so we are led to reject the whole debateas meaningless . This
rejection, however, seemsto go too far. Take for instancethe MN 'foot-
wear'. It applies to a shoe and to any collection of shoes, but does not
apply to a proper part of a shoe. A shoe's heel is certainly not footwear,
and a leg of a table is not furniture. Thus, there is a truth of the matter
whether a part of a shoe is a shoe and a part of a table is a table. All of
this seemsperfectly meaningful, and such knowledge is assumedin the
everydayuse of the language. Indeed, later we will introduce a notion of
substancerelative to a systemof interpretations of nouns (like 'heel' and
'leg'), and thesequestionswill in fact becomemeaningful. The answerwill
dependof courseon the systemof interpretations considered.
Notice that a related although different question may be formulated
meaningfully without invoking a universal substance: is the sup-lattice M
atomless? The reasonis that the notion of an atom (or primitive element)
is definable in any sup-lattice. We believe that these two questions
(whether the part of an element of M is again an element of M and
whether M is atomless) have sometimesbeen confounded, and this has
obscured the fact that the first can be formulated only if a universal
notion of substanceis assumed. At any rate, under the assumption of a
universal substance, divisivenessof referenceimplies that the extensions
of MN s are atomless. This is the only issue that we will addressat this
point .
Several authors (see, e.g., Roeper 1983; Lonning 1987; Bunt 1985)
have erecteddivisivenessof referenceinto a characteristic of MN s. Bunt
Count Nouns and Mass Nouns 431

(1985, 45- 46), for example, allows that in the actual physical extensions
of many MN s there are minimal parts, but dismissesthe fact as linguisti-
cally irrelevant. Quine (1960), Parsons(1970), Gillon (1992), and others
reject this claim for the reason that we formulate as follows. It may well
be that for many centuries users of the English word 'water' have not
realized that there are minimal portions of water- namely, molecules-
whoseparts are not water. (Notice how both questions, already discussed ,
are here confounded.) The effective use of the word does not depend on
knowledge of that scientific fact. This seemsto have led some writers to
the conclusion that the facts of the matter have no bearing on the exten-
sions of common nouns. But extension is extension, independentof our
knowledge. There is a truth of the matter that there are minimal parts
in extensionsof some MNs at least: witness the examples of 'footwear'
and 'furniture' already discussed . Thus, we do not seereasonsto set limi -
tations a priori to extensions of MN s. Furthermore, excluding atomic
extensionswould rule out the possibility of considering plurals as MN s.
Not only is there linguistic evidenceto considerplurals as MN s, discussed
by Carlson and Link (see below), but we will see that the plural con-
struction may be viewed as a basic link between CN sand MN s. The
fruitfulness of this approach will be apparent in the chapter.

17.1.3 Categoriesof Nouns and Their Interpretations

One of the main novelties of our approach is our taking into account the
connectionsbetween MNs , the connectionsbetween CNs, and the ways
in which MNs can be transformed into CNs and vice versa. For instance,
there is a connectionbetweenthe MN 'iron ' and the MN 'metal' described
in colloquial languageby the sentence'Iron is metal', analogous to that
between 'a dog' and 'an animal' that we studied in La Palme Reyes,
Macnamara, and Reyes 1994a. Furthermore, there must be some way
of connecting the MN 'metal' and the CN 'a metal' to validate some
arguments involving both expressions . This net of connections may be
organized in an objective way by means of a system of categoriesand
functors: the nominal theory. This system includes the category ~ JV'
whose objects are CNs and whose morphisms are axioms of the form
'a dog is a mammal', 'a mammal is an animal', which may be thought
of as a systemof identifications. By composing those axioms, we obtain
the new identification 'a dog is an animal'. In a similar vein, the nominal
theory includes the category JIt JV' of MNs such as 'iron' and 'metal',
whosemorphisms are axioms such as 'iron is metal'. The transfonnations
432 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

mentioned before between CN sand MN s are described by functors . One

example is the plural formation that takes the CN ' dog ' into ' dogs ' . Since

the extension of this term obviously has the property of cumulative refer -

ence , we categorize ' dogs ' as a MN . Carlson ( 1977 ) and then Link ( 1983 )

called attention to the affinity between plurals and MNs . In fact , this

affinity even extends to the syntax of both expressions . As an example , just

like MN s , plural CN s do not take the pI ural . Also on a par with MN s,

such combinations as * 'a dogs ' and * ' another dogs ' are ungrammatical .

We interpret the nominal theory as follows : CN s are interpreted as sets

( or kinds ) . Morphisms between CN s are interpreted as set - theoretical

maps . Such maps we call underlying . MN s are interpreted as sup - lattices

( to be defined precisely later ) and morphisms as sup - preserving maps .

Functors of the nominal theory are interpreted as functors of the corre -

sponding interpretati .ons . For instance , formation of the plural is inter -

preted as the power set functor between the interpretations of C( / JV and

those of J / t JV , that is , between the category of sets , Sets , and the category

of sup - lattices , st .

17 . 1 . 4 Grammatical Transformations and Syllogisms

The nominal theory and its interpretation will be used in this chapter for a

single goal : to discuss the validity of a sample of eight syllogisms dis -

cussed by Pelletier and Schubert ( 1989 ) , involving CN s , MN s , and predi -

cables . To achieve this purpose , we need to build semantic counterparts to

grammatical transformations of CN s into MN s, MN s into predicables ,

and so on , extending our previous work ( La Palme Reyes , Macnamara ,

and Reyes 1994a ) . As an example , consider the syllogism

Claret is a wine , wine is a liquid , so claret is liquid .

In the first premise , ' claret ' is a NP ( noun phrase ) and ' is a wine ' a VP

( verb phrase ) ; in the second , ' wine ' is a NP and ' is a liquid ' a VP . In the

conclusion , ' claret ' is a NP ( as in the first premise ) and ' is liquid ' is a VP .

The NPs in this example are , however , like PNs ( proper names ) or

descriptions . Crosslinguistic evidence suggests that ' claret ' is not a PN . In

fact , in French , one uses the expression ' Ie bordeaux ' , and French does

not allow the definite article in front of a PN . This suggests categorizing

' claret ' as a descriptive noun phrase ( DNP ) , that is , as a NP whose inter -

pretation is a member of a kind . As in La Palme Reyes , Macnamara , and

Reyes 1994a , we consider ' is a wine ' as a predicable derived from the term

' wine ' (a CN , as indicated by the occurrence of ' a ' ) . The second premise
Count Nouns and Mass Nouns 433

can be analyzedin a similar way. On the other hand, we will assumethat

'is liquid ' in the conclusion is a predicable derived from the term 'liquid '
(a MN , as indicated by the absenceof 'a').
At the level of interpretation, these grammatical transformations (or
derivations) will take us into computations of colimits of interpretations
in our categoriesto define notions of relative entity, relative substance,
and so on.

17.1.5 Category Theory versusSet Theory

A final word about the use of category theory (rather than set theory) in
this chapter. (For a more thorough discussion, see Magnan and Reyes
1994.) A first observationconcernsgenerality: setsthemselvesconstitute a
particular category. The use of category theory allows us to formulate
semanticsthat are free of the particular determinations imposed by a too
rigid adherenceto set theory. From this point of view, we would like
to emphasizethat the semanticsintroduced here is one among several
possibleones.
A second observation concerns abstractness . Contrary to widespread
belief, set-theoretic semanticsare more abstract than category-theoretic
ones. This is easily understood when we compare usual set-theoretic
semanticsof CNs and MNs with our category-theoretic semantics. As we
pointed out, one of the main novelties of our approach is to take as the
basic ingredients of our semantics the connections between dogs and
animals, iron and metal, wines and wine, and so on. These connections
guided the choice of the categoriesusedto interpret CNs and MNs .
To give an example, any sup-lattice is also an inf -lattice; thus, from a
set-theoretical point of view, it makes no difference whether one works
with sup-lattices or inf -lattices. On the other hand, morphisms of sup-
lattices are quite different from morphisms of inf -lattices. The followin ~
example shows that underlying relations do not preserve /\ in general:
take a person who has traveled twice, say, and consider the underlying
map at the level of sets of persons(which are sup-lattices as well as inf-
lattices),
2PASSENGER
~ 2PERSON
,
which associateswith a set of passengersthe underlying set of persons.
Clearly, u preservesV (arbitrary unions in this case). On the other hand,
u does not preserve binary 1\: Let Pi ' P2 be the two passengerswhose
underlying person is John. Then { PI } 1\ {P2} = e5 and hence U({PI } 1\
434 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

{P2} ) == 0 . But U{ Pl } 1\ U{P2} == {John} . Other examples of this kind

can be given to justify our choice of categories.
From this point of view, the trouble with set-theoretical semanticsis
simply this: they are too abstract, since they abstract away thesefunda-
mental relations, which are therefore not properly represented(in these
semantics). As a consequence , set-theoretical constructions are not con-
strained in a natural way: there are just too many, and when a choice is
required, extraneousprinciples, usually ora pragmatic nature, are brought
in to decide the issue. On the other hand, categorical constructions are
highly constrained through the use of universal properties: among all
possibleconstructions, one is distinguishedas satisfying a universal prop-
erty. Becauseof this feature, it has been argued that the theory of cate-
goriesconstitutesa theory of concreteuniversals, set theory being rather a
theory of abstract universals(Ellennan 1988). This explains the ubiquity
of universal constructions in this chapter. In a more speculativeway, we
are tempted to believe that universal constructions may capture what is
universal in the human mind, including what is fundamental in human
language.

17.2 TheNominalTheory

We now describethe nominal theory. This theory describesin an objective

way the count nouns (CNs), the mass nouns (MNs ), and their trans-
formations by meansof a systemof categoriesand functors

satisfyingthe followingrelation: p -i u, namely,p is left adjoint to u. We

will illustratethis relationbelow. We think ofp as 'plural of' and of u as
'a portion of', 'a numberof', 'a setof', andsoon, sincethereis no unified
way (and sometimes simplyno way) of expressing u at the surfacelevel.
The adjointnessrelationstatesthe equivalence betweenan axiom of the
form p( I a portion of meatl) ~ ~ in JtIJV' ('portions of meat are
food') andtheaxi~m r a portion of meatI-!:. u(~ ) in C
(jJV' ('a portion
of meatis a portion of food'). This is clearenough. Sometimes , however ,
we lack lexical itemsto expressu, as in the equivalence(also obtained
from the adjunction) betweenp ([~~ ~~~~]) ~ [~~!~~~!~ in JtIJV' ('dogsare
Count Nouns and Mass Nouns 435

animals ') andtheaxiom[~~~~~J~ u([:~~!~~~~]) in rcJV(whichcouldbe

expressed ratherclumsily as'a dogis a numberof animals ', withthe
understanding thatthisnumber couldbeone ). Axioms in thisrelation are
saidto betransposes of eachother .
Noticethatunderthisadjunction , pu(~ )~ ~ ('portions of
meataremeat ') is anaxiomof /ItJV, sinceits transpose u(~ )~
u(~ ) ('a portionof meatisa portionof meat ') is anaxiomof rcJV,
indeedanidentityaxiom . Axioms obtained in thiswayarecalledco-units
of theadjunction . Similarly , [~~~~~J ~ up([:~~~~J) ('a dogis a number
ofdogs ') isanaxiomofrcJV, since itstranspose p([~~~~~J) ~ p([~::~~! ])
('dogsaredogs ') is anidentityaxiomof J!{JV' Axiomsobtained in this
wayarecalledunitsof theadjunction .
Thiswayof conceptualizing thepluralformation seems compatible
with that of Jackendoff (1991 ). Themaindifference between thetwo
approaches isthatweorganize nouns intocategories (rcJVandJ!{JV) and
sowecanconsider pluralformation asa functorbetween them , whereas
Jackendoff dividesnounsinto sets : 'individuals' ('a dog'), 'groups ' ('a
committee '), 'substances ' ('water '), and'aggregates ' ('buses ', 'cattle'). Plural
formation in histheoryisa map(between some of these sets ) thatsends ,
forinstance , 'a dog'into'dogs ' and'a committee ' into'committees '. This
is precisely thewayourfunctoractsonobjects of thecategory rcJV. A
furtherdifference concerns hisdivisionof nouns . It maybepossible to
introduce thesefurtherdivisions in ourapproach , a question
wehavenot
investigated . Weplanto returnto thisquestion in a forthcoming
paper.
Eachcategoryof the nominaltheorymaybeconsidered asa system
of identifications that replaces
the '==' of a singlesortedtheory. These
categories seemto beposetalin thesense thatthereis at mostonearrow
between twoobjects , butwewill notrequirethemto beso.
~JV' is thecategory whoseobjectsare"genuine " CNs suchas[~~~~~J,
I a mammal I, I ananimalI andwhosemorphisms are axiomsof identi-
ficationof theform

The identity morphisms are particular axioms of the form

[~~~~J~ [~~~~J'
436 La Palme Reyes~Macnamara~Reyes~and Zolfaghari

and composition is given by modus ponens. For instance , from

and

[_
-amammal
I~ Iananimal
I,
we obtain

[~~ ~i ] ~ I ananimal
!.
Similarly
~ ,~,~,JV
~isthe
,~category
whose
andobjects
are
whoseMNsof
such
morphisms
are as
the[:~~!~!J,
form
~ ~ ~ ,
~ ~ ~ .
Identity morphisms and composition of morphisms are as above .
The maps p and u are assumed to be functorial . The functoriality ofp
amounts to saying that from

~ ~ ~ ,

we may conclude that

rA : l are ru "; l
~ ~ ~ .

As an example , from

we may conclude that

The functorialityof u saysthat from

~ ~ ~,
we obtain
Count Nouns and Mass Nouns 437

In particular , from

~ ~ ~ ,

we obtain

both categories. Thus, we may say 'Claret is a wine' and 'More wine was
served after the dessert'. In the first sentence
, 'wine' appears as a CN ,
whereasin the second, it appearsas a MN . At the level of the nominal
theory, this meansthat we have two objects:

[~~ ~~~ !~~ E ~ JV' and ~ E JttJV' .

But surely
~ there must be a connection between them. We represent the
connection by a morphism of ~ JV :
I ~ ' "ll1n ~ I is I ' "ll1n ~ I
I a WIlle l - - t U L.~ ~~~J'

Equivalently , since p is left adjoint to u, the connection may be expressed

by a morphism of JtI/ JV :
II'WIlleS
"l~1
nA
~IIare
I~
~L'"l~
~
1
nJI'A
These morphisms expressthat 'a wine is wine' and 'wines are wine',
respectively(theseassertionsare indeedobviously equivalent) . In section3
we will interpret the first as a morphism of the category of sets. The sec-
ond, on the other hand, we will interpret as a morphism of the category of
438 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

sup - lattices , its transpose . Notice t ~ at the morphism [~~~ ~~ ~~ ~~] ~ u ~

is quite different from [~ ~~ !~ ~J ~ [~ ~:~ !~ ~ , a morphism of CC.. IV that
expresses that ' a wine is a wine ' and that will be interpreted as an identity
map in the category of sets .

The case of wine is not an isolated one . Another example is ' metal ' .
Thus , we can say ' Iron is a metal ' , but also ' Iron is metal ' . Once again , the
connection between the two occurrences of 'metal ' is expressed in the
nominal theory by a morphism in CC.. IV:

[~~~~~!~J-.!!-. u[~~~!~!].
Equivalently , we may express the connection by a morphism in JI: t .. IV:

11"nl3t'::11~ I are I n"\ pt "::ll I

L~~ ~! ~ !~ ~ l__~ ~: ! ~ ~ .

Note that in addition to the previously mentioned endofunctor

pu : JI:t . . IV - t JI:t . . IV, which may be read as ' portions of ' or ' numbers of ' ,
up : CC.. IV - t CCJV is an endofunctor that may be read as ' a set of ' ( or ' a
number of ' ) , whose interpretation will be the power set operation .

REMARK 1 It is interesting to observe that the functorp is not full in gen -

eral . This means that there might be morphisms p(~ ) ~ p(~ )

in Jt { . . IV that do not come from any morphism ~ !!.. ~ of rc . . IV.

In more logical terms : the sentences ' A ' s are B ' s' and ' An A is a B ' are

not equivalent in general ( although the second implies the first ) . This
fact seems to contradict a widespread belief . As an example , consider
' Living room sets are items of furniture ' and ' A living room set is an item
of furniture ' . The first is true ; the second is false .

17 .3 Interpretations of the Nominal Theory

Our goal in this section is to define the notion of an interpretation of the

nominal theory . As we explained in section 17 . 1.3 , CN s are interpreted as
sets and MNs as sup -lattices . Notice , however , that sets and sup -lattices
constitute categories . Indeed , the category of sets , Sets , has sets as objects
and set - theoretical functions as morphisms . Composition is the ordinary
notion of composition between functions . The category of sup - lattices , Sl ,
has sup - lattices as objects and sup -preserving functions as morphisms .
More precisely , an object of Sf is a triple ( M , ,::;: , V ) where ( M , ,::;: ) is a
poset and V : 2M - t M associates with every subset of M its supremum .
Thus , VA ~ m iff Vm ' E A m ' ::; m . A morphism of Sl
CountNounsand MassNouns 439

f : (M , :::;;, V ) - t (N , :::;;, V )

is a set-theoretical map between f : M - t N that preserves suprema

in the sense that f (VA ) == V3f (A), where 3f (A ) == { y EN : 3x E
A (y == f (x))} , that is, the image of A under the function f Once again,
composition is the usual composition betweenfunctions.
REMARK2 In La Paille Reyes, Macnamara, and Reyes 1994b, inter-
pretations of CNs are kinds, namely, setswith a relation of constituency,
associatingwith each member the set of situations of which that member
is a constituent. This allows us to study predicableslike 'sick' or 'run' that
may hold of a member of a kind at a given situation but fail to hold at
another. To simplify the presentation, we will leave situations out of the
picture and interpret CNs as setsand MNs as sup-lattices. Furthermore,
thesesimplified notions will sufficefor the limited aim of this chapter.
Thereis a connection between. these categories that may be described
by meansof two functors

such that P ~ U . P is the (covariant ) power set functor and U is the for -
getful functor . More precisely , P is the functor that associates with a set
X the sup-lattice P X of the subsets of X , and with a function / : X --+ Y
the function :Jf : P X --+ P Y defined above . This is clearly a Sf-morphism .
On the other hand , U is the functor that associates with a sup - lattice

(M , ~ , V ) the set M and with the morphism / : (M , ~ , V ) --+

(N , ~ , V ) the function/ itself.
We will usually denote by ' O(A ) ' , ' O(B ) ', ' O(C) ', . . . the objects of Sf
and write ' IO(A) I' instead of ' U(O(A))' .
An interpretation of the nominal theory in the system of categories just
described is given by two functors (both denoted by 1) making the fol -
lowing diagram commutative
I
((j . . IV ~ Set

pliu p11
u
JIit JV - --~-_.- + SI

in the obvious sense; for instance, PI == Ip , and so on .

 440 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

This meansthat the objects and the morphisms of C (i/AI" are interpreted
as setsand morphisms betweenthesesets, the objects and the morphisms
of JI{ % as objectsand morphisms of Sl. Furthermore, the interpretations
should " behave well" with respect to functors of the nominal category.
Thus, for instance, PI == Ip meansthat if the interpretation of [ ~~~~~J is
the set of dogs, then the interpretation of ~ should be the set of sets
of dogs, and so on. (It is worthwhile to notice that axioms in the nominal
theory are interpreted not as truth-valuesbut as maps.)
Functoriality of I : rc% ~ Setsmeansthat the interpretation of 'is' in
the axiom

is a morphism of sets- intuitively the one that assigns to a dog its

underlying animal. Furthermore, the interpretation of 'is' in the axiom

[~~~~~J~ [~~~~J
is the identity map ; and so on .

REMARK 3 Lawvere ( 1992 ) has suggested identifying categories of

space with distributive categories and categories of quantity with linear

categories . In the appendix we will show that the target category of the

interpretations of CN s , Sets , constitutes a category of space whereas the

target category of the interpretations of MN s , Sf , constitutes a category of

quantity . We feel that any determination of the interpretations of CN s

and MNs should have these properties . Ours is just one such determina -

tion , chosen mainly for simplicity .

17 . 4 Syllogisms and Their Validity

17 . 4 . 1 Some Types of Syllogisms

After defining the notion of interpretation we may study validity of all

kinds of reasoning involving CNs , MNs , DNPs , and predicables , as is

usually done in logic . Lacking a systematic way of generating these kinds

of reasoning , we will limit ourselves mainly to a few syllogisms that have

been studied in the literature ( Pelletier and Schubert 1989 ) . The notion of

validity for syllogisms is the usual Tarski 's notion of truth , defined by

recursion on complexity of formulas .

Count Nouns and Mass Nouns 441

(a) Syllogisms involving CN s, PN s, and predicables. These are the

usual Aristotelian syllogismssuch as
All Greeks are men, all men are mortal, so all Greeks are mortal .
In La Palme Reyes, Macnamara, and Reyes 1994a, we noticed two
problems connected with such syllogisms: one due to the change of
grammatical role of some of the terms (e.g., 'men' appears as part of
a predicablein the first premiseand as a CN in referring position, subject
of the sentence , in the second); the other due to the changeof sorting of
predicables(e.g., 'mortal ' is sorted by the CN 'men' in the secondpremise
and by the CN 'Greek' in the conclusion).
These problems were dealt with in the work cited, and we will say
nothing more about them here beyond remarking that the main idea was
to use the colimit of a systemof kinds to construct a coincidencerelation
that in turn provides a semantical counterpart to the transformation of
a CN into a predicable. We will extend thesenotions here to systemsof
sup-lattices giving semantical counterpartsto the transformations of MNs
into predicables, essentialfor the validity of the syllogisms.
(b) Syllogismsinvolving MN s, DNPs, and predicables. An example of
theseis
Claret is beer, beer is alcoholic, so claret is alcoholic.
Notice that here we have problems similar to those found in the Aristo -
telian syllogisms: 'is beer' is a VP (or predicable) in the first premiseand a
NP- in fact, a DNP - in the second. A further problem concernsthe sort
of the predicable 'alcoholic': it seemsto be sorted by 'beer' in the second
premise but by 'claret' in the conclusion. However, 'claret' is categorized
only as a MN in the nominal theory. If this is so, should we allow MNs as
well as CN s to sort predicables?
Another example of such a syllogism is
Claret is a beer, beer is a liquid , so claret is liquid .
In this syllogism a new difficulty presentsitself. In fact, 'liquid ' appears
as a predicable derived from the CN 'liquid ' in the second premise (as
witnessedby the particle 'a'), but as a predicable derived from the MN
'liquid ' in the conclusion.
We must give a semantical account of thesetransformations if we ever
hope to test the validity of this type of syllogism.
The way to tackle theseproblems is parallel to that employed with the
Aristotelian syllogisms and depends on a notion of coincidence that
442 La PalmeReyes
, Macnamara
, Reyes
, andZolfaghari

allows us to transform a MN into a predicable, for instance, from 'beer'

to 'to be beer'. We will deal with thesequestionsin section 17.4.4.
REMARK4 (1) We have given as examplessyllogismswhosepremisesare
not necessarilytrue to avoid falling into the trap of believing that they
should be taken as axioms of the nominal theory, thereby limiting the
validity of syllogismsto a very restricted class. (2) We have given what we
believe to be a " literal" reading of the syllogisms. This does not seemto
be the only possiblereading. In fact, we could also read the first premise
as 'claret is a kind of beer'. It is important to keep this in mind when
testing the validity of syllogisms.

17.4.2 CoincidenceRelationsand Predicates

In the appendix we will construct notions of entity (a set E) and sub-
stance (a sup-lattice O(S)), relative to an interpretation, with a map
e : E ~ O(S) and underlying maps canx : X ~ E from each set X that
interprets a CN and underlying sup-lattice maps canO(A) : O(A ) ~ 0 (8)
from each sup-lattice O(A ) that interprets a MN .
In turn, thesenotions will allow us to define two coincidencerelations
betweenmembersof kinds. Both of thesenotions are reflexive, symmetric,
and transitive. According to the first, a and bare E-coincident if they have
the sameunderlying entity. As an example, a dog is E-coincident with its
underlying animal. According to the second, a and bare O(S)-coincident
if they have the sameunderlying substance. As an example, a portion of
iron is O(S)-coincidentwith its underlying portion of metal. Becauseof the
existenceof the map E ~ O(S), E-coincidenceimplies O(S)-coincidence.
Thus, a dog and its underlying animal have the sameunderlying substance.
In terms of thesenotions, we will define the semanticalcounterpart of
the grammatical transformations betweenCNs, MNs , and predicablesin
section 17.4.4.

17.4.3 Grammatical Analysis

Following Pelletier and Schubert (1989), we will limit ourselvesto the
following list of eight syllogismsand compare our resultswith theirs.
1. Claret is a wine, wine is a liquid , so claret is a liquid .
2. Claret is a wine, wine is a liquid , so claret is liquid .
3. Claret is a wine, wine is liquid , so claret is liquid .
4. Claret is a wine, wine is liquid , so claret is a liquid .
5. Claret is wine, wine is a liquid , so claret is a liquid .
Count Nouns and Mass Nouns 443

6. Claret is wine, wine is a liquid , so claret is liquid .

7. Claret is wine, wine is liquid , so claret is a liquid .
8. Claret is wine, wine is liquid , so claret is liquid .
Before testing the validity of thesesyllogisms, we need to assigngram-
matical roles to their teffils. As noted in remark 4, we will make our
reading ofthem asliteral aspossible .
First , a general remark to motivate theanalysis . Syllogisms are
grounded inthenominal theory , inthesense thatterms other thanpredi -
cables belong tothenominal theory andarethuscategorized aseither a
CNoraMN. (Some termsmayappear inmore thanonecategory .)
Wewillassume thatthenominal theory categorizes 'claret' asa MN
([~!~~~]), but'wine ' and'liquid ' asbothCNs([~~~!~~, [~:~~~~~!~J) and
MNs(~ , [~~~!~J).
Letusconsider thesecond syllogism . Thefirstpremise , 'claret is a
wine ', isanalyzed grammatically asNP+ VP.Inturn , NPisanalyzed as
DNP(descriptive noun phrase ) withlexical item[~~~~]. Ontheother
hand , VPis analyzed asV+NPwithlexical items 'is' and[~~~!~~J.
Following proposals inLaPalme Reyes , Macnamara , andReyes 1994a ,
theVP'isawine ' isdeduced fromtheCN'wine '. Thesecond premise ,
'wine isaliquid ', isanalyzed similarly . Ontheother hand , theconclusion ,
'claret is liquid
', willbeanalyzed asDNP+ VP, where 'is liquid' is
derived fromtheMN'liquid '.
17.4.4 InterpretationandValidity
As we saidin section17.3, a CN is interpretedas a kind (or setin this
chapter), a MN asa sup-lattice, a DNP asa memberof a kind, and a VP
asa predicate .
Comingbackto the secondsyllogism , the lexicalitem [~~~~~ in J!lJV,
analyzedas a DNP, shouldbe interpretedas a memberof a kind. But
what memberof what kind? We proposeto interpretit asthe largestele-
ment 1o(C) of the sup-lattice O(C) that interprets(the MN ) [~~~~!]-
namely, asthelargestportion of claret, in the kind IO(C)I. Thepredicable
'is a wine' is interpretedas the predicateof beingE-coincidentwith a
particularwine. Thus, 'claretis a wine' is true iff lo (c) E-coincideswith
someWE W, whereW is the kind of wines. Similarly, 'wine is a liquid' is
true iff 1o(w) E-coincideswith someI EL. whereL is the kind of liquids.
On theotherhand, theconclusion , 'claretis liquid', is trueiff lo (c) hasthe
propertyof beingliquid. We considerthis propertyas sortedby the CN
444
La Palme Reyes, Macnamara, Reyes, and Zolfaghari

that is the

morphIsm
are wine ' in the nominal theory . This morphism is interpreted as a map
f : W -jo O( W ) , which to w associates f (w), a portion of wine in O( W ).
Similarly , the morphism 'liquids are liquid ' is interpreted as a map
g : L -jo O(L ).
Since E-coincidence implies O(S)-coincidence , from the premises we
may conclude

(1) lo (c) is O(S)-coincident with f (w);

(2) 1o(w) is O(S)-coincident with g(I).
By definition of 1o(w), f (w) ~ 1O(w), and this implies the corresponding
inequality between the underlying substances,

(3) cano(w)(f (w)) ~ canO(W) (Io (w)),

since canO ( w ) is a sup -lattice map .
From ( I ), (2), and (3) we conclude easily that

(4) cano(c)( lo (c)) ~ canO(L)(g (/ )) .

To go further, we needanother postulate. We shall assumethat 'an O(S)-

portion of liquid is liquid ' in the precisesensethat canO(L) : O(L ) -jo O(S)
is downward surjective: if s ~ canO(L) (A), then there is A' ~ A such that
s == canO(L) (A'). If this property is satisfied, we will say that O(L ) is
homogeneous (or distributed or divisive or divided ).
Putting all of this together , we have

PROPOSITION1 If the interpretation O(L ) of 'liquid ' is homogeneous ,

then the second syllogism is valid .

On the other hand, the first syllogism is not valid, even if O(L ) is
homogeneous : 'an O(S)-portion of a liquid , although liquid , need not be
a liquid ' . In fact , proceeding in this way , we can prove the following :
THEOREM
1 If the interpretation O(L ) of 'liquid ' is homogeneous, then
the syllogisms 2, 3, 6, and 8 are valid , and the syllogisms I , 4, 5, and 7 are
not valid .

REMARK 5 Our conclusions put us at odds with Pelletier and Schubert

( 1989) . Indeed , these authors consider the first syllogism as valid . How -
Count Nouns and Mass N Duns 445

ever, we agree on the validity (and nonvalidity) of the remaining ones.

The disagreementcomesabout from the different interpretations assigned
to the CN 'a wine'. Unlike us, Pelletier and Schubertinterpret it as 'a kind
of wine' and they prove validity from the postulate that 'kinds of kinds
are kinds'. Two remarks are in order. First, their interpretation is higher-
order, sincethey quantify over 'kinds of wine'. Second, we do not believe
that this rather vague notion allows us to found a logical theory. In par-
ticular, the postulate of transitivity is not evident to us.

Appendix

In this appendix , we formulate and prove some mathematical statements

used in the main text .

A . I ( Co ) cones and ( Co ) limits in a Category

Given a functor F : .!I - + d , a co cone for F consists of an object A E d

and a family { F ( i ) L A } i of morphisms of d such that for every arrow

i ~ j E .!I the diagram

F ( i ) ~

F (a ) 1 ~ A

F (j )

commutes . This cocone will be denoted ( A , { h } i ) . Cocones for F consti -

tute a category by defining a morphism

<I > : ( A , { h } i ) - + ( B , { gi } i )

to be a morphism of d , <I > : A - + B , such that <I > 0 1; = gi for all i E .!f .

We define the colimit of F, colim F, to be the initial cocone of this cat-

egory. By spelling this out, co/im F is a family { F (i) ~ A } i of morphisms
of d such that
1. For every arrow i ~ j E .!Jithe diagram

commutes .
446 La PalmeReyes
, Macnamara
, Reyes
, andZolfaghari

2. (Universal Property) Whenever { F (i) ~ B} i is a family of maps such

that for every i ~ j E Jl! the diagram
F (i)
~
F(c)1x~ B
F (j )

F
(F
i(a
))l~
-:~
..y
.I!A
for all i:

i~-~
commutes, there is a unique morphism () : A --+ BEd

-
.7 B
such that OJ; == gi

F(j)':~ ...
We say that the category d is cocompleteif every functor F : .!Ii -1- d ,
.!f being a small category, has a colimit .
REMARK6 (1) Since terminal objects are unique up to unique iso-
morphism, we are justified in talking about " the" colimit of a functor. (2)
Dually , we can define conesand limits in a category by considering dop ,
rather than d . We say that d is completeif every functor F : J1 - + d , .!f
being a small category, has a limit .

A.2 Exactness
Propertiesof SetsandSl
THEOREM 2 The category of sets, Sets, is complete, cocomplete, and dis-
tributive in the strong sensethat coproducts commute with pull-backs.
Furthermore, surjectionsare stable.
Proof All of this is well known (see, e.g., Mac Lane 1971).
As an example of colimits, let .!Ii == C(6..;v and F == I : C(6..;v -j. Setsbe an
interpretation of the category of CN s. The colimit of I may be described
as follows:

1. Take the disjoint union Eo == UiI (i). Thus, an elementof Eo is a

couple (i , a) where a E I (i).
2. Take the smallestequivalencerelation =: on Eo generatedby the pairs
((i, a), (j , b)) where there is a : i -j. j E CC
% such that I (i) (a) = b.
3. Define E = Eo/ =: and maps, for each i,

I (i) ~ Eo.uE,
Count Nouns and Mass Nouns 447

as follows: a ~ (i, a) ~ [(i, a)] where [(i, a)] is the equivalenceclassof

(i, a) underthe relation :=.
Then (/ (i) [~ i E)i is the colimit of I .
To explainthis constructionin oneexample , assumethat
la quadruped
[
y
[~:~~~!!~J
~
[~::~~ ~!~J
is a diagramin ((lJV ' that expresses
that 'a turtle is a quadruped
' and
'a turtle is a reptile'. In Eo we havecoupleslike (I a 'quadruped
I, q),
([~ ~!~~~~J, f), and ([:~~~~ !!~J, r), whereq E/ (1a quadruped !), t E
/ ([:~~!~~~~J), and r E/ ([~~~ ~!~ ), respectively . Assumethat q is the
quadruped underlyingthe turtle t andr the reptileunderlyingt. Then
(I a quadruped I, q), ([~~!~~!!~J, f), and ([~:~ :~~E!!!~]' r) are amongthe
pairsgenerating =: . On theotherhand, I a quadruped I, q), ([~~E~ !!!~J,
r)), althoughnot a generating pair, is in =: .
THEOREM3 The category Sf is complete , cocomplete , and linear .

Proof This is again well known (see, e.g., Joyal and Tierney 1984) .
Recall that a category is linear if it has the following properties :
1. C(i has finite products and coproducts (including terminal and initial
objects : 1 and 0, respectively ) .
2. The unique morphism 0 - + 1 is an isomorphism with inverse 1 - + o.
3. The canonical morphism A + B - + A x B obtained from the mor -
phisms

( 1A, 0AB) : A - 4 A x B
and

(0BA, 1B) : B - t A x B

is an isomorphism , where 0AS is the composite of A - + 1, 1 - + 0, and

0 -+ B.

We proceed as in Joyal and Tierney 1984. In this category 1 = 0 =

({ *} , s , V ).
448 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

(M
,~M
,V)
Products are diagrams

/
(M x N, :S, V )
~ (N , ::::; N , V )

where ::::; is defined pointwise :

V i ( Xi , Yi ) == ( V iX i , V iY i ) , n M ( x , Y) == x , n N ( X, Y) == Y .

Coproducts are diagrams

(M , ~ M , V )

~
/ " (M x N , ::::; , V )
/ iN
(N , ~ N , V )

whereiM(x) == (x, 0) andiN(Y) == (0,y).

Equalizersarediagrams

(E, ~ E, V ) '-t
e (M, ~ M, V ) ~fg (N, ~ N, V )

such that

e f
E ' -tM4N
g

is an equalizerin Sets, ~ == (e x e)- l ( ~ M).

Coequalizers arediagrams
f q
(M , ~ M, V ) ~g (N, ~ N, V ) -+ (Q, :$: , V )

where

i
Q == { z EN : Vx E M (f (x ) ::::; Z +-+ g (x ) S z) } '-;). N

and q ~ i . (See Joyal and Tierney 1984 for this characterization .)

The universal property of the coequalizer is easy and left to the reader .

As an example of colimit (substance ) in Sf , assume that .!Ii is the

category
Count Nouns and Mass Nouns 449

IS

where identities have been omitted . Let I : ! f - + Sl be a functor . We will

compute O ( S ) == coliml , following the description of co limits in terms of

coequalizers ( see Mac Lane 1971 ) , as the coequalizer

I([ j )~gI([:~~~!~~)xI(~ )~0(8),

where f ( a ) == ( u ( a ) , 0 ) and g ( a ) == ( 0 , v ( a ) ) and l ( [ j ) ~ 1 ( [::~~ ~ ! ~ ! J ) ,

l ( ~ ) ~ l ( ~ ) are the interpretations of ' is ' in AJV .

But O ( S ) is then obtained by dividing out the coproduct

I([~~~!~!J)xI(~ )
by the smallest sup - lattice congruence relation == such that

VaEI (~ )(u(a), 0) := (0, v(a)).

As described in the preceding theorem ,

O ( S ) == { ( b , c ) E 1 ( [ ~~ ~ ! ~ ! J ) x l (~ ) :

~ a E l (~ ) ( u ( a ) S b ~ v ( a ) s c ) }

and q ( b , c ) = = the smallest ( b ' , c ' ) E 0 ( 8 ) such that b s b ' and c s c ' .

The connection between Sets and Sf is given by the couple of functors P

and U . These functors are of a special kind :

THEOREM 4 The - functor P is left adjoint to U . ( In symbols : P -1 U .)

Proof This is well known ( see , e .g . , Mac Lane 1971 ) . The theorem says

that there is a natural bijection between Sl ( P X , M ) , the set of sup -

preserving maps from P X to M , and Set ( X , U M ) , the set of maps between

X and the underlying set of M . This is a consequence of the following uni -

versal property of the singleton map X il UP X : fo ! every set map

f : X - + U ( M ) there is a unique sup - preserving map f : PX - + M such

that u ( f ) 0 { } == f . The proof _of this assertion is rather immediate : the

definition of the sup - preserving f is forced to be

450
La Palme Reyes, Macnamara, Reyes, and Zolfaghari
-

f (A ) == V f (a)
aEA

since
] mustpreserve
V 's andA = UaEA
{ a}. (Of courseonemustprove
that the map , so defined , is sup-preserving .)

This connection is reflected in a connection between entity and

substance :

THEOREM
5 There is a unique map e : E - + 0 (8) such that for each i the
diagram
{}
l (i) ~
IPI (i) 1== IIp (i)1

canI
(~ E ()
~~CanIP(j)I
I0 ( 8 ) 1

is commutative .

Proof According to the definition of E == coliml , it is enough to show

the commutativity of

I (i) _____il _-+ IPI (i) I

ICanlp
(i) I

!(a)l 510 (8)1

l (j ) ---{-}"-+ IPl (j ) I ICanlp
(j)I

for all i ~ j E ~ JV ~. But the diagram may be decomposed into a diagram

in Sets and another in Sf :

I (i ) _ _il -- t IP I ( i ) I

I(a)1 113
/(1)I X
I (j ) - { } -t IPI (j ) 1

PI (i) ==Ip (i )
------~ ~
3I(a)=/p(a)1 - - - --:=---_..~ 0 (S) can Ip ( j )

P l ( j ) == Ip (j )

The first commutes trivially : 3/ (a) { a } == { l ( a ) ( a ) } . The second commutes

by definition of ' O ( S ) . By property ( 2 ) of the definition of E == coliml ,
Count Nouns and Mass Nouns 451

(i)10{ } , which
thereis a uniqueE ~ 10(8)\ suchthat eo canI(i) == ICanlp
is the conclusion of the theorem .

References

Bunt , H . C . ( 1985) . Mass terms and model-theoretic semantics. Cambridge : Cam -

bridge University Press.
Carlson , G . (1977). A unified account of the English bare plural . Lmguistics and
Philosophy , 1, 413- 457.
Ellerman , D . P. ( 1988) . Category theory and concrete universals . Erkenntnis , 28,
409 - 429 .

Gillon , B . S. ( 1992) . Towards a common semantics for English count and mass
nouns . Linguistics and Philosophy , 15, 597- 639.
Jackendoff , R . ( 1991) . Parts and boundaries . Cognition , 41, 9- 45 .
Jespersen, O . ( 1924). The philosophy of grammar . London : Allen and Unwin .
Joyal , A ., and Tierney , M . ( 1984) . An extension of the Galois theory of Gro-
thendieck (Memoirs of the A .M .S., 309) . Providence , RI : American Mathematical
Society .
La Palme Reyes, M ., Macnamara , J., and Reyes G . E . ( 1994a) . Functoriality and
grammatical role in syllogisms . Notre Dame Journal of Formal Logic , 35, 41- 66.
La Palme Reyes, M ., Macnamara , J., and Reyes, G . E . ( 1994b) . Reference, kinds
and predicates . In J. Macnamara and G . E . Reyes ( Eds .), The logicalfoundations
of cognition (pp . 91- 143) . Oxford : Oxford University Press.
La Palme Reyes, M ., and Reyes, G . E . (to appear) . Generic figures and their
glueings: A constructive approach to functor categories .
Lawvere , F . W . ( 1992) . Categories of space and of quantity . In J. Echeverria , A .
Ibarra , and T . Mormann ( Eds .), The space of mathematics (pp . 14- 30) . Berlin : de
Gruyter .
Lawvere , F . W ., and Schanuel , S. ( 1997). Conceptual mathematics . Cambridge :
Cambridge University Press.
Link , G . ( 1983) . The logical analysis of plurals and mass terms : A lattice -theoretic
approach . In R . Bauerle , C . Schwarze, and A . von Stechow ( Eds.), Meaning , use
and interpretation of language ( pp . 302- 323) . Berlin : de Gruyter .
Lonning , J. T . ( 1987) . Mass terms and quantification . Linguistics and Philosophy ,
10 , 1- 52 .

Mac Lane , S. ( 1971) . Categories for the working mathematician . New York :
Springer -Verlag .
Magnan , F ., and Reyes, G . E . ( 1994). Category theory as a conceptual tool in the
study of cognition . In J. Macnamara and G . E . Reyes (Eds .) , The logical founda -
tions of cognition ( pp . 57- 90) . Oxford : Oxford University Press.
McPherson , L . M . P. ( 1991) . A little goes a long way : Evidence for a perceptual
basis of learning for the noun categories count and mass. Journal of Child Lan -
guage , 18 , 315 - 338 .
452 La Palme Reyes, Macnamara, Reyes, and Zolfaghari

Parsons, T . ( 1970). An analysis of mass and amount tenus . Foundations of Lan -

guage , 6 , 363 - 388 .

Pelletier , F . J., and Schubert , L . K . ( 1989) . Mass expressions. In D . Gabbay and

F . Guenthner ( Eds .), Handbook of philosophical logic (vol . 4, pp . 327- 407).
Dordrecht : Reidel .

Quine , W . V . o . ( 1960) . Word and object . Cambridge , MA : MIT Press.

Roeper , P. ( 1983) . Semantics for mass terms with quantifiers . No us, 17, 251- 265.
Soja, N . N ., Carey , S., and Spelke, E . S. ( 1991). Ontological categories guide
young children 's inductions of word meaning : Object terms and substance terms .
Cognition , 38, 179- 211.