Frege and Names
Frege and Names
Frege and Names
Pirmin Stekeler-Weithofer
CONTENTS
sense or ‘way’ leading to it. There might be no meaning without some sense.
In other words: When we use expressions in (object level) sentences or state-
ments, we do not reflect on their usage, nor do we think about any ‘rule’
of their use. Rather, we ‘implicitly assume’ that we already know how to
handle the criteria defining the identity of meaning. By doing this, we do
not presuppose criteria of identity of sense nor, in the case of ‘rule-follow-
ing’, identities of rules. Naming senses (or rules or functions) is a metalevel
enterprise.
Many modern theorists of language who see themeselves in the tradition
of Frege’s analysis go further and think that the sense of an expression ‘is’
something like a rule or a function. As possible arguments or inputs of such
a function one could assume, then, n-tupels of ‘things’ (or ‘entities’) like
the following: an expression taken as a linguistic form or type, a context, a
generic (‘abstract’) and/or an individualized (‘concrete’) situation (turning
the type of usage into a token of use), perhaps a speaker, a hearer, and
perhaps other ‘things’ like interests or goals. A value or output of such a
function could be an object if the expression was a name, or a truth value if
it was a sentence. – We know the function or rule resp. the sense, it seems,
if we know the values or outputs resulting from any of the possible inputs.
But how can or do we know something like this? What is a function? How
do we grasp a rule? These were some of the central questions of the later
Wittgenstein. Indeed, these question are strongly related to the question
how to grasp the sense of an expression and how its meaning is determined
by this sense.
We may object against this idea claiming that ‘formal languages’ and
‘mathematical functions’ differ from ‘natural languages’ and senses respec-
tively. We certainly have to distinguish between the ways in which the truth
values (or truth conditions) of mathematical statements and other state-
ments are defined. Nevertheless, a consideration of mathematical sentences
can help us to show analogous features of other realms of talk and to see
what it means to talk about object- and meta-levels of language-use.
According to Frege, not so much the senses of names, but concepts (pred-
icates) ‘are’ functions. They are functions with exactly two possible values,
called „(The) True” and „(The) False”. Their ‘arguments’ are objects, i.e.
possible meanings of possible names.
It is important to see the following ambiguity: When we say that the
meaning of an expression (i.e. of a saturated expression) is the object named,
then there is a conventional reading and a rather unusual one. The first as-
sumes that there are objects and that some expressions are used to name
them. In this reading it seems more natural to translate the German
154 Pirmin Stekeler-Weithofer
2. Realms of entities
meaningful names, the ‘existence’ of the abstract objects to which the names
refer cannot be questioned.
If we ask what there is and what we can and do quantify about, a classical
answer gives us the advice to believe in realms of possible ‘entities’ that can
be values of variables. Such a belief is often called „ontological commitment”.
Logicians like Quine obviously assume that there are functions that lead us
from variables or names as arguments to entities or objects as values. But
what does it mean to assume that a function like this exists? How is such
a function defined? What is it to be a possible value of a variable? The
question leads us back to the questions above.
The reason why I am not satisfied with the answers or explanations
that are usually given to these questions is this: They do not tell us what it
means to believe in certain realms of objects or entities or to embrace certain
ontological commitments. They do not tell us what functions or relations
and objects really are or how they can be defined and identified. To see the
difficulty, notice how superficial it is to assume abstract objects as given and
to look for a relation between such objects and names. What is needed here
is a constitutional explanation of particular realms of abstract objects. I even
dare to claim that neither Quine nor Tarski nor any other modern ‘formal
ontologist’ gave us a sufficient answer to this question. Instead, they made
their different appeals to ‘beliefs’ and ‘commitments’. Of course, they tried
to overcome the mere appeal to vague beliefs by a precise axiomatization
of a holistic framework like formal set and model theory. The concepts of
formal truth and interpretations are defined in such a framework.
I cannot show all the problems of such a formalist theory of meaning
and truth or of a philosophy of language following the formal lines drawn by
Davidson and Montague. I just show some problems. In short, I claim that
the procedure of formalist semantics is unsatisfying not because it is false,
but because it stops short of asking and answering important questions. The
most important is: What do we presuppose as clear when we use a certain
formalistic framework? For it is crucial to see that any formalist analysis,
e.g. any ‘axiomatic foundation’ of set and model theory or of mathematics,
always leaves certain questions of content open. When we restrict our at-
tention to what can be treated schematically, we always (and deliberately)
avoid the disturbing fact that we would have to rely on reasonable judge-
ment — a very vague concept — when we would have to answer questions
concerning our reasons for the choice of the formal framework as a whole
and questions about their right external use. This external usage takes place
in contexts and situations in which we are interested in good orientations
Lorenzen and Friedrich Kambartel seem to be underestimated until today, even if some
particular ideas of their ‘constructivism’ might be problematic.
158 Pirmin Stekeler-Weithofer
and successfull practices. The mastery of the internal rules of the formal
games, e.g. of proofs controlling the correctness of our rule following, plays
a certain limited role here. Of course, it is legitimate that in our division of
labour ‘working mathematicians’ and ‘technical logicians’ are not interested
‘professionally’ in such ‘external’ questions. But it would be (and somehow
already is) a disaster for a reasonable development of science if departmental
research forgets that not (only) the internal, regional, criteria of ‘success’ de-
termine the reasonability of the research itself, but the place of the research
and its results in a division of scientific, technical and cultural labour.
Different from a merely formalistic procedure, Frege wants to show how
(structured) realms of objects can be and are constituted, namely via a
realm of possible ‘names’ and by an explanation of truth conditions for
equalities and other sentences. Frege knows that there is no entity and no
name without reference to a whole system of names and sentences. As long
as there is no analysis of the linguistic constitution of the realms of entities,
we remain naive and do not even know what we are committed to and what
it is to believe that quantified sentences were meaningful (‘true’ or ‘false’).
The question how a realm of objects is constituted does not merely arise
in the case of abstract objects. Neither abstract, nor ‘concrete’ objects are
just (to be treated as) ‘given entities’. The question what a particular ob-
ject in such a realm of objects is cannot be answered sufficiently by nam-
ing or showing the object ‘immediately’. Names and examples give us only
(re)presentations of such objects — just like any presentation of a number
is a presentation of a name or of another ‘representation’ of the number.
This is so, even if the object is a concrete thing like a chair or a table.
In some sense, nobody can really ‘show’ us a thing like a chair as such.
What can be shown are always (re)presentations of the thing. At least for
such objects, linguistic expressions (names) are only a special kind of repre-
sentations; appearances are another, especially if they are accompanied by
situation-dependent utterances of namelike ‘descriptions’ as, for example,
„this object here”.
There are presupposed criteria telling us which sensational and symbolic
representations (would) count as representations of the same object or, what
amounts to the same, which representations are to be treated as ‘equivalent’.
Only by such (truth) conditions for equalities it is defined what the repre-
sentations are representations of. Notice that there are ‘phenomenological’
forms of talking ‘about’ sensations or gestalts or subjective representations
without already referring to physical objects. Only ‘later’ objects can be
counted as bodily causes for ‘sensations’.
What Is a Meaningful Name? 159
We speak about abstract objects when the formal truth conditions for sen-
tences containing symbolic or linguistic representations are defined by ‘rules’
that do not depend on non-symbolic appearances of the objects. Hence, it is
a conceptual truth that we cannot ‘see’ or ‘feel’ abstract objects, and that
they are no ‘impressions in our mind’ (whatever the last expression should
mean). Abstract truths as such are not tested immediately by experience.
They are defined by conventional rules we have to learn and we can learn
just by listening to texts, reading books, and by learning to rephrase and
formally deduce further sentences. To show how we can analyse abstract
objects of a certain category, we consider the example of (natural) numbers.
We have to answer the following questions:
1) What are (syntactically well defined) candidates for meaningful names
of numbers?
2) What is a meaningful name of a number?
3) How are the truth conditions for equalities between meaningful names
defined?
4) Which further properties (sentence-forms with truth conditions) are
defined for the numbers? I.e. what sentence-forms are to be taken in consid-
eration and what are their satisfaction conditions?
As we said already, a name N always presupposes a whole system of
names and a whole system of sentences S including equalities of the form
„N = M ” such that for the sentences the following conditions hold:
5) For any sentence of the system there is one and only one of two truth
values (e.g. 1 and 0, or T and F , or True and False) defined. (This principle
could be called the principle of distinction or consistency.)
6) If the value of an equality N = M is the peculiar value called „True”,
and if the value of S(N ) is „True”, then the value of S(N/M ) is „True”.
(This principle can be called the principle of Leibniz.)
We can say, roughly, that if 5.) and 6.) are fulfilled, a one place sen-
tence-form (or predicate) „S(x)” divides the names (or the objects named)
into two classes. On the other side, a name (an object) divides the sentence-
-forms (predicates) into two classes. The principle of Leibniz ‘says’ that the
distinctions defined by the predicates are not (or may not be) finer than
the equation (or equivalence) N = M . True equalities ‘say’ that in the cor-
responding semantical category that we do not (and cannot) differentiate
between possibly different but equivalent representations (names or appear-
ances). The last means that they ‘have the same meaning’ and ‘name the
same object’ with respect to CR .
164 Pirmin Stekeler-Weithofer
story talks about. We usually use rules of the following sorts: For making our
story a story ‘about’ persons or objects, we use the Leibniz-Principle (LP),
which defines equality: If A(N ) is true and if N = M , then A(M ) is true.
Some terminological rules (TR) are usually presupposed, too: If N loves M ,
then N does not hate M . If N hates M , then N does not love M . In order
not to complicate things too much here, the story does not contain logically
complex, e.g. quantified, sentences. But it is clear that such sentences are
used by applying the rules of predicate calculus (PC). (These rules often
include LP).
In our example we use LP in order to conclude that Joe commits suicide.
Because we say that to commit suicide is no murder (this is a terminological
rule belonging to type TR), we can apply a rule of PC to get the following
result: there is (possibly) no murder(er). The story does not tell us that any
person killed another person. Hence, the expression „the murderer”, if we
read it as a definite description, is not meaningful, does not have a meaning,
does not refer to anybody with respect to the story.
When we apply rules like these we often say that the consequences ‘fol-
low logically resp. conceptually’ from what was said in the story. For some
readers, even the following rule might be assumed: „N always loves N ”.
This, however, is rather a ‘synthetic assumption’ (SA) than a terminological
rule.
In the end, a reader might even assume the following rule: „If our story
(together with a use of ‘rules’ listed above) does not tell us anything about
the relation of love (resp. hatred, killing and even identity) between N and
M , then we may assume that N does not love (hate, kill, or is equal with)
M . We call an ‘assumption’ like this a ‘completeness rule’ (CR) with respect
to the set of elementary (i.e. logically non-complex) sentences defining the
possible properties S(x) and relations S(x, y, . . .) of the categorical context-
-system CR . CR presupposes that we know the whole story, especially that
we know all relevant contexts of category CR . When we ‘concluded’ that
according to our story there is no murderer, we already used CR silently.
When using CR, positive elementary sentences play a special role: If we
cannot deduce the positive sentences from the text by rules of the first three
types, CR tells us to assume the negated assertions. Things get complicated
if the text contains more logically complex statements and if the story is
(to be) treated as an uncomplete picture. Indeed, it is often unclear how
far we can rely on rule CR: We might read the story as a mere beginning
of a longer story. I claim, however, that we always use rules or principles of
the first four types, and that we often use CR, too, when we ‘understand’ a
166 Pirmin Stekeler-Weithofer
text. Even when we say what the story says ‘explicitly’, we normally do not
restrict our judgements to sentences that are listed explicitly.
This means that we do not read such texts merely as axiomatic systems
‘describing’ classes of possible structures on arbitrary sets of objects that
only have to have ‘enough’ elements in order to fulfill the formal conditions
of the story. We often assume silently that a story describes a structure that
is ‘smallest’ and ‘easiest’ in the sense that it does not contain more objects
or elements and it does not contain more true elementary statements about
them than it must in order to fulfill the conditions told in the story. Often, of
course, such a ‘smallest’ and ‘easiest’ structure is not determined uniquely.
Even if a story only describes ‘possibilities’, i.e. if it only determines a
subclass of ‘really’ possible state of affairs in a full class of ‘all’ possibilities,
any such possibility is defined by a possible ‘full description’ of a relational
structure or model.
If we use ‘full’ rule CR together with the other rules, the class of ‘really’
possible state of affairs defined by our story contains exactly one structure,
the ‘real state of affairs’ of the story — if we restrict our attention to the
objects named in the story. With respect to this state, we can answer ‘all
possible CR -questions’ concerning what is true about our persons.
If rule CR is omitted or weakened, the story only tells us, which worlds
are the possible worlds according to our story. Notice that to define one of
these structures by the truth values of the corresponding elementary sen-
tences of a categorical context-system CR is something else than to define
subclasses in a class of ‘all possible’ structures by an axiomatic theory, i.e.
by a list of logically complex formulas and the rules of PC. There is no way
to describe a particular formal structure in another way than by describing
the true and false sentences ‘of it’. This means that the usual proceeding of
set theoretic model theory already presupposes the class of possible models
or, what amounts to the same, what it means to quantify over all sets of
naive set theory.4
With respect to our (formal) real state of affairs, some definite descrip-
tions of the form „the x such that A(x)” are meaningful, others are not.
„The person loved by Joe” is a meaningful definite description (assuming
CR), „the murderer” is not. But for all such expressions, their sense is de-
fined.
In On Denoting, Russell reformulated the idea of Frege how to use defi-
nite descriptions t with respect to a realm R in sentences S(t) of a context-
4
I do not say anything, here, about a mere internal model theory that does not leave
the framework of formalistic, axiomatized, set theory, even if this would be important for
an analysis of Tarski’s work.
What Is a Meaningful Name? 167
With respect to the realm of objects given in our story, a new namelike
expression like „Peter” does neither have sense nor meaning, whereas an
expression like „the sailor” or „the murderer” seems to have sense, even if
the story does not fix their meanings. (Notice that this is not a question of
knowledge.) What do we have to do in order to make Peter into a person in
our story? Just to add the name „Peter” to the list of elementary names does
not suffice. We can turn „Peter” into a name of an object in our story only
if some predicates (properties, relations) ‘for Peter’ are defined. We might,
for example, use principle CR and assume that Peter is different from all
others and neither loves nor hates nor kills anybody else in our realm. Then,
Peter is a fifth person.
We might proceed, now, in the same way with the expression „the sailor”,
considering what the story ‘says explicitly’ about a person with that ‘nick-
name’. Then, the number of persons of our original story is not four, but
five. Indeed, if we use CR in full scale, this would be ‘the real’ reading of
our story: There is no way for reading „the sailor” as a definite description
if we did not assume that our story were incomplete.
This short and certainly not sufficient example shows why Frege said that
we cannot just ‘construct’ objects if not with reference to a whole system of
objects. Hence, we cannot construct new abstract objects by a declaration
like the following: Assume that the real equation „x2 = −1” had a solution
and name it „i” or „the positive imaginary root of −1”. Then you will arrive
at the complex numbers.
The real procedure has to tell us what possible names of complex num-
bers are and what possible properties they have. And this must be done
in a whole system. Hence, Frege does not contradict himself when he said
that ‘in some sense’ his own procedure in the Basic Laws of Arithmetic is
also constructive. The main difference to a ‘naive’ construction of objects by
adding new names to a system of names are Frege’s holistic principles that
define what it means that a name is meaningful with respect to a context
or category or ‘story’ CR .
A definition of a realm of basic R-names occurring in elementary sen-
tences and of truth values (or truth conditions) for R-sentences is the basis
for a definition of the truth value or truth conditions of logically complex
R-sentences in which sentential connectives and quantifiers occur. An im-
plicit reference to a (possible) system of sentences fulfilling the principles
listed above is the basis, too, for the interpretation of a deduction according
to the rules of classical predicate calculus as a ‘valid consequence’.
A common confusion in interpretations of Frege’s enterprise results from
the following mistake: Frege thought that we can combine any two realms
What Is a Meaningful Name? 169
R, Q of objects just by adding two new predicates R(x) and Q(x) and a
principle of the form: any object N of the first realm R is totally different
from any object M of the second realm Q. I.e., for any elementary sentence
„S(x)” of the second realm „S(N )” is wrong if „R(N )” is true; and for any
elementary sentence „S(x)” of the first realm, „S(M )” is wrong if „Q(M )”
is true. This was the (wrong) reason why Frege thought that we did not
need to bother about the local realms of the quantifiers but could combine
all realms into one realm of ‘all objects of meaningful discourse’. This is
also the reason why he says in his Principles of Arithmetics that we have
to make sure that a number, let us say the number five, is distinguished by
some property from everything else, even from the sun and the moon. In
other words, Frege did not see that the constitution of realms of entities as
realms for qantificational logic always are bound to local categories.
5. Oblique contexts
names. – Unfortunately, Frege did not tell us enough about this constitu-
tion of oblique contexts. Therefore, the following statement may lead to
some misunderstandings, if it is not read with a grain of salt: In oblique
contexts, the meaning of an expression ‘is’ its normal sense. The objects we
talk about in oblique contexts are closely related to the way the meanings of
the expressions are determined in non-oblique usages of the expressions. The
(presupposed) equivalence relation ∼ determines (implicitly) ‘how close’ this
relation is. It depends heavily on the context in which we use the expression
„the sense of X” as a singular term.
If the meanings are different, the senses should be different, too. This
principle says that any equivalence relation ∼ used for defining the sense
of an expression as an object of talk in an oblique context should be finer
than the equivalence relation defining identity of meaning in a corresponding
object-level. In short, senses ‘formally supervene’ on meanings. We should
see, however, that the differentiations between sense and meaning, between
non-oblique and oblique contexts and between an object-level and meta-
-levels are, in a large extent, context-relative.
Expressions having the same meaning may be counted as expressing (or,
in certain oblique context: naming) different senses. Which expressions, ut-
terances or (symbolic or nonsymbolic) acts ‘have the same sense’, as we say,
depends on many things, especially on the interest we take in the specific
‘meta-level’ and ‘oblique’ differentiations and identifications (‘nondifferen-
tiations’) of senses, intensions, propositions, properties, intentions and so
on.
Insofar as all texts in written books can be learned, they can in some sense
always be read as ‘abstract’ texts about ‘abstract’ objects. Thus, ‘formal
knowledge’, i.e. knowledge of literature, is largely independent of ‘empirical’
knowledge concerning ‘concrete’ things or phenomena. It often is not clear
whether we talk about a concrete or about an abstract and ideal person
or object when we talk about Moses, Homer, Socrates, or other persons of
written history, or when we talk about atoms, electrons, quantums, neurons
and other objects of scientific literature. Sometimes, we know fairly well that
we do not talk about ‘real’ persons or objects, for example when we talk
about Sherlock Holmes or about the ‘personal God’ of Jews, Christians or
Muslims. On the other hand, historical or empirical truth is not the only
‘good’ reason we use when we orient our actions on sentences we assume
as formally true. This can be seen clearly in the case of arithmetics and
What Is a Meaningful Name? 171
in its consistency, if we do not have a sufficient proof or argument for it. But
neither the axioms of Peano arithmetics, nor of geometry, nor of set theory
really can define arithmetics, geometry or set theoretical truths — whatever
Hilbert believed neglecting Frege’s (correct) criticism. We can (and must)
show that the formal axioms of such a theory articulate ‘true’ statements and
‘true’ rules with respect to ‘elementary’ arithmetics, geometry or set theory,
where the concept of truth is defined by certain non-axiomatic conditions. I
do not have the space here to say more about this rather far-reaching claim.
As an example, it suffices to show that it might be a reasonable (or even
true) claim.
If we consider the role played by true statements in wider contexts
of judgements, practice and actions, we may see that pragmatists rightly
claimed that truth, not only mathematical truth, is a formal concept. The
same holds for the concept of an object (or ‘entity’) relative to a category
R: It is, as I hope to have shown, at least at first always a totally formal
concept.
The result of this insight is, if I am right: When we want to know how
we use the forms of speech considered here and the corresponding formal
terminological and logical rules, we have to experiences, goals and inter-
ests, practical successes and certainties into our picture. Only if our interest
remains restricted to an analysis of formal features of meaning, i.e. of for-
mal truth-values and formal truth-conditions of sentences (or statements),
as Frege was, we may ‘abstract’ from possible ‘external’ roles played by
the concept of truth and from how we relate the formal realms of ‘inter-
nal’ objects we talk about to external experience and practice. Any such
‘abstraction’ refrains from an analysis of concrete and complete meaning.
Wittgenstein and before him William James noticed that a Russellian
and Fregean analysis of sense, meaning and truth remains abstract and for-
malistic in this sense. Indeed, truth conditional semantics as it was developed
by Frege, Tarski, Davidson or Montague remains ‘internalistic’, i.e. it has
only a limited interest, just like Frege’s differentiation between sense and
meaning does not suffice (at all) for a ‘whole’ analysis of meaning.
But let us come back to our title question: What is a (meaningful) name?
The result of our consideration should be this: The see mingly obvious an-
swer: „A (meaningful) name is a sign or symbol that is used to name an
object” does not suffice at all. We cannot give a sufficient answer to this
question without explaining what it is to be an object that can be named.
For this, we want to know what it is to name an object. I take it that it
was these questions Frege tried to answer. And he noticed that we cannot
answer it without reference to a whole system of sentences.
What Is a Meaningful Name? 173
Pirmin Stekeler-Weithofer
Institut für Philosophie
Universität Leipzig
Augustusplatz 9
04109 Leipzig, Germany