Minnesota Studies in the

PHILOSOPHY OF SCIENCE

VOLUME I
The Foundations of Science and the Concefts
of Psychology and Psychoanalysis

EDITED BY

HERBERT FEIGL AND MICHAEL SCRIVEN

FOR THE MINNESOTA CENTER FOR PHILOSOPHY O F SCIENCE
 THEORETICAL CONCEPTS
RUDOLF CARNAP
bility of drawing any boundary line. It is true that empiricists today
generally agree that certain criteria previously proposed were too narrow;
for example, the requirement that all theoretical terms should be defin-
The Methodological Character of able on the basis of those of the observation language and that all theo-
I retical sentences should be translatable into the observation language.
T'hmetical Concepts 1 W e are at present aware that these requirements are too strong because
I the rules connecting the two languages (which we shall call "rules of
I correspondence") can give only a partial interpretation for the theo-
retical language. From this fact, some philosophers draw the conclusion
1 that, once the earlier criteria are liberalized, we shall find a continuous
I. O u r Problems line from terms which are closely connected with observations, e.g.,
IN DISCUSSIONS on the methodology of science, it is customary and use- 'mass' and 'temperature,' through more remote terms like 'electromag-
ful to divide the language of science into two parts, the observation I netic field' and 'psi-function' in physics, to those terms which have no
language and the theoretical language. The observation language uses I specifiable connection with observable events, e.g., terms in speculative
terms designating observable properties and relations for the descrip metaphysics; therefore, meaningfulness seems to them merely a matter
tion of observable things or events. The theoretical language, on the I of degree: This skeptical position is maintained also by some empiricists;
other hand, contains terms which may refer to unobservable events, un-
observable aspects or features of events, e.g., to micro-particles like
-Hempel for instance
A ---.Lhas _
-- given clear and --- forceful arguments for this
I view (see his articles, (14) anF^ (15) ). Although he still regards the
electrons or atoms, to the electromagnetic field or the gravitational field e d e a of the empiricist meaning criterion as sound, he believes that
in physics, to drives and potentials of'various kinds in psychology, etc. deep-going modifications are necessary. First, the question of meaning-
In this article I shall try to clarify the nature of the theoretical language fulness cannot, in his opinion, be raised for any single term or sentence
and its relation to the observation language. The observation language but only for the whole system consisting of the theory, expressed in the
will be briefly described in Section 2 of this paper. Then a more detailed theoretical language, and the correspondence rules. And secondly, even
account of the theoretical language and the connections between the for this system as a whole, he thinks that no sharp distinction between
two languages will be given in Sections 111-V. meaningful and meaningless can be drawn; we may, at best, say some-
One of the main topics will be the problem of a criterion of signifi- \ thing about its degree of confirmation on the basis of the available
cance for the theoretical language, i.e., exact conditions which terms
and sentences of the theoretical language must fulfill in order to have
a positive function for the explanation and prediction of observable
I observational evidence, or about the degree of its explanatory or pre-
dictive power for observable events.
The skeptics do not, of course, deny that we can draw an exact bound-
events and thus to be acceptable as empirically meaningful. I shall leave ary line if we want to. But they doubt whether any boundary line is an
aside the problem of a criterion of significance for the observation lan- 1 adequate explication of the distinction which empiricists had originally
guage, because there seem to be hardly any points of serious disagree- t in mind. They believe that, if any boundary line is drawn, it will be
ment among philosophers today with respect to this problem, at least '
inore or less arbitrary; and, moreover, that it will turn out to be either
if the observation language is understood in the narrow sense indicated L )too narrow or too wide. That it is too narrow means that some terms
above. On the other hand, the problem for the theoretical language is I or sentences are excluded which are accepted by scientists as meaningful;
a very serious one. There are not only disagreements with respect to that it is too wide means that some terms or sentences are included
the exact location of the boundary line between the meaningful and the which scientifically thinking men would not accept as meaningful.
meaningless, but some philosophers are doubtful about the very possi- \ M attitude
~ is more optimistic than that of the skeptics. I believe
i,
 Rudolf Carnap THEORETICAL CONCEPTS

that, also in the theoretical language, it is.possible to draw an adequate The terms of V, are predicates designating observable properties of
boundary line which separates the scientifically meaningful from the events or things (e.g., "blue," "hot," "large," etc.) or observable rela-
meaningless. I shall propose criteria of significance; the criterion for tions between them (e.g., "x is warmer than y," "x is contiguous to y,"
theoretical terms will be formulated in Section VI, and the question of etc.).
its adequacy will be examined in Section VII; the criterion for theoreti- Some philosophers have proposed certain principles which restrict
cal sentences will be given in Section VIII. either the forms of expression or the procedures of deduction in "the
Two alternative forms for the ir~troductionof scientific concepts into language," in order to make sure that everything said in the language
our two-language system will be explained and their comparative use- is completely meaningful. It seems to me that the justification of such
fulness examined (Sections IX and X). One kind is that of theoretical requirements depends upon the purpose for which the language in
concepts introduced into the theoretical language by postulates. The question is used. Since Lo is intended for the description of observable
other kind I call "disposition concepts." They may be introduced into events and therefore is meant to be completely inter reted, the require-
an extended observation language. Concepts defined by so-called opera- .$- 'I('?
ments, or at least some of them, seem to have merit. Let us consider
tional definitions and the so-called intervening variables belong to this the most important requirements that have been proposed for some
kind. I shall try to show that the introduction in the form of theoretical or any language L.
concepts is a more useful method because it allows greater freedom in
1. Requirement of observability for the primitive descriptive terms.
the choice of conceptual forms; moreover, it seems more in accord with 2. Requirements of various degrees of strictness for the nonprimitive
the way the scientists actually use their concepts. descriptive terms.
In the last section, I discuss briefly the possibilities and advantages (a) Explicit definability.
of the use of theoretical concepts in psychology. (b) Reducibility by conditional definitions (e.g., by reduction sen-
tences as proposed in (5) ).
11. The Observation Language Lo 3. Requirement of nominalism: the values of the variables must be con-
The total language of science, L, is considered as consisting of two crete, observable entities (e.g., observable events, things, or thing-
moments).
parts, the observation language Lo and the theoretical language LT. 4. Requirement of finitism, in one of three forms of increasing strict-
I shall here briefly indicate the nature of Lo; the later discussion will ness: ?

chiefly concern LT and its relations to Lo. Without actually specifying (a) The rules of the language L do not state or imply that the basic
it, we assume that the logical structure of Lo is given. This would in- domain (the range of values of the individual variables) is infi-
clude a specification of the primitive constants, divided into logical and nite. In technical terms, L has at least one finite model.
descriptive (i.e., nonlogical) constants. Let the observational vocabulary (b) L has only finite models.
(c) There is a finite number n such that no model cdntains more
Vo be the class of the descriptive constants of Lo. Further, for each than n individuals.
language part the admitted types of variables are specified. In Lo, it may 5. Requirement of constructivism: every value of any variable of L is
suffice to use only individual variables, with observable events (including designated by an expression in L.
thing-moments) taken as individuals. Then rules of formation, which 6. Requirement of extensionality. The language contains only truth-
specify the admitted forms of sentences, and rules of logical deduction functional connectives, no terms for logical or causal modalities
are given. (necessity, possibility, etc.).
Let us imagine that L o is used by a certain language community as a Any language fulfilling these requirements is more directly and more
means of communication, and that all sentences of L o are understood completely understandable than languages transgressing these limita-
by all members of the group in the same sense. Thus a complete in- tions. However, for the language as a whole, the requirements are not
terpretation of L o is given. justified; we shall reject them later for the theoretical language LT.
 Rudolf Carnap THEORETICAL CONCEPTS

Since then we have in the part LT all the freedom of expression desired, A theory is given, consisting of a finite number of postulates formu-
we may well accept some or all of these requirements for Lo. lated in L.Let T be the conjunction of these postulates. Finally, cor-
W e have already accepted requirements 1 and 3. The decision about respondence rules C are given, which connect the terms of VT with those
requirement 2 depends upon our intention concerning disposition terms of VO.These rules will be explained in Section V.
(e.g., "soluble," "fragile," "flexible"). W e shall not include them in Lo
itself; thus Lo is here taken as a restricted observation language fulfilling IV. The Problem of the Admissibility of Theoretical Entities
the stronger requirement 2(a). Later (in Section IX) the possibility It seems that the acceptance of the following three conventions CI-
of an extended observation language L'o, which allows the introduction C? is sufficient to make sure that LT includes all of mathematics that is
of disposition terms, will be explained. Another method consists in needed in science and also all kinds of entities that customarily occur
representing the disposition concepts by theoretical terms in L (Sec- in any branch of empirical science.
tion X). Conventions about the domain D of entities admitted as values of
The weakest requirement 4(a) of finitism is fulfilled in Lo. Therefore variables in LT.
it is easily possible to satisfy requirement 5. Further, we take Lo as an CI. D includes a denumerable subdomain I of entities.
extensional language; thus requirement 6 is fulfilled. I C2. Any ordered n-tuple of entities in D (for any finite n ) belongs
also to D.
C?. Any class of entities in D belongs also to D.
111. The Theoretical Language LT
The primitive constants of L are, like those of Lo, divided into logi- I shall now indicate briefly how these conventions yield all the cus-
cal and descriptive constants. Let the theoretical vocabulary VT be the tomary kinds of entities referred to in scientific theories. To facilitate
class of the descriptive primitive constants of L. W e shall often call the understanding, I shall first use the customary way of speaking and
these constants simply "theoretical terms." (They are often called "theo- the customary terms for certain kinds of entities, and only later add a
retical constructs" or "hypothetical constructs." However, since the warning against a possible misinterpretation of these formulations.
term "construct" was originally used for explicitly defined terms or con First about mathematical entities. Since the subdomain I stipulated
cepts, it might be preferable to avoid this term here and use instead in CI is denumerable, we may regard its elements as the natural num-
the neutral phrase "theoretical term" (or "theoretical primitive"). This bers 0, 1, 2, etc. If R is any relation whose members belong to D, then
use seems to be in better accord with the fact that it is, in general, not R may be construed as a class of ordered pairs of its members. There-
possible to give explicit definitions for theoretical terms on the basis fore, according to C2 and C?, R belongs also to D. Now the (positive
of Lo.) and negative) integers can, in the usual way, be constructed as relations
W e may take it for granted that LT contains the usual truth-functional of natural numbers. Thus, they belong also to D. Analogously, we pro-
connectives (e.g., for negation and conjunction). Other connectives, ceed to rational numbers as relations among integers, to real numbers
e.g., signs for logical modalities (e.g., logical necessity and strict impli- as classes of rational numbers, and to complex numbers as ordered pairs
cation) and for causal modalities (e.g., causal necessity and causal im- of real numbers. Furthermore, we obtain classes of numbers of these
plication) may be admitted if desired; but their inclusion would require kinds, relations among them, functions (as special kinds of relations)
a considerably more complicated set of rules of logical deduction (as whose arguments and values are numbers, then classes of functions,
syntactical or semantical rules). The most important remaining problem functions of functions, etc. Thus D includes all those kinds of entities
for the specification of the logical structure concerns the ranges of values needed in the purely mathematical part of LT.
for the variables to be admitted in universal and existential quantifiers, Now we proceed to physics. W e assume that LT is based upon a par-
and thereby the kinds of entities dealt with in LT. This problem will be ticular space-time coordinate system; thus the space-time points are
discussed in Section IV. ordered quadruples of real numbers and hence, according to C2, belong
 Rudolf Carnap THEORETICAL CONCEPTS

to D. A space-time region is a class of space-time points. Any particular without cognitive content. In contrast to this, there is a good sense of
physical system of which a physicist may speak, e.g., a material body the word "real," viz., that used in everyday language and in science. It
or a process of radiation, occupies a certain space-time region. When may be useful for our present discussion to distinguish two kinds of the
a physicist describes a physical system or a process occurring in it meaningful use of "real," viz., the common sense use and the scientific
or a momentary state of it, he ascribes values of physical magnitudes use. Although in actual practice there is no sharp line between these
(e.g., mass, electric charge, temperature, electromagnetic field intensity, two uses, we may, in view of our partition of the total language L into
energy, and the like) either to the space-time region as a whole or to the two parts Lo and LT, distinguish between the use of "real" in con-
its points. The values of a physical magnitude are either real numbers nection with Lo, and that in connection with LT. W e assume that Lo
or n-tuples of such. Thus a physical magnitude is a function whose contains only one kind of variable, and that the values of these variables
arguments are either space-time points or regions and whose values are are possible observable events. In this context, the question of reality
either real numbers or n-tuples of such. Thus, on the basis of our con- can be raised only with respect to possible events. The statement that
ventions, the domain D contains space-time points and regions, physical a specified possible observable event, e.g., that of this valley having been
magnitudes and their values, physical systems and their states. A physi- a lake in earlier times, is real means the same as the statement that the
I
cal system itself is nothing else than a space-time region characterized sentence of Lo which describes this event is true, and therefore means
in terms of magnitudes. In a similar way, all other entities occurring in just the same as this sentence itself: "This valley was a lake."
physical theories can be shown to belong to D. For a question of reality in connection with LT, the situation is in
Psychological concepts are properties, relations, or quantitative mag- certain respects more -,omplicated. If the question concerns the reality
nitudes ascribed to certain space-time regions (usually human organisms of an event described in theoretical terms, the situation is not much
or classes of such). Therefore they belong to the same logical types as different from the earlier one: to accept a statement of reality of this
concepts of physics, irrespective of the question of their difference in kind is the same as to accept the sentence of LT describing the event.
meaning and way of definition. Note that the logical type of a psycho- However, a question about the reality of something like electrons in
logical concept is also independent of its methodological nature, e.g, general (in contradistinction to the question about the reality of a
whether based on observation of behavior or on introspection; philoso- cloud af electrons moving here now in a specified way, which is a
phers seem sometimes not to realize this. Thus the domain D includes question of the former kind) or the electromagnetic field in general is
also all entities referred to in psychology. The same holds for all social of a different nature. A question of this kind is in itself rather ambigu-
sciences. ous. But we can give it a good scientific meaning, e.g., if we agree to
W e have considered some of the kinds of entities referred to in mathe- understand the acceptance of the reality, say, of the electromagnetic field
matics, physics, psychology, and the social sciences and have indicated in the classical sense as the acceptance of a language LT and in it a term,
that they belong to the domain D. However, I wish to emphasize here say 'E,' and a set of postulates T which includes the classical laws of the
that this talk about the admission of this or that kind of entity as values electromagnetic field (say, the Maxwell equations) as postulates for 'E.'
of variables in LT is only a way of speaking intended to make the use For an observer X to "accept" the postulates of T, means here not
of L,, and especially the use of quantified variables in LT, more easily simply to take T as an uninterpreted calculus, but to use T together
understandable. Therefore the explanations just given must not be with specified rules of correspondence C for guiding his expectations
understood as implying that those who iccept and use a language of by deriving predictions about future observable events from observed
the kind here described are thereby committed to certain "ontological" - events with the help of T and C.
doctrines in the traditional metaphysical sense. The usual ontological I said previously that the elements of the basic domain I may be
questions about the "reality7' (in an alIeged metaphysical sense) of num- regarded as natural numbers. But I warned that this remark and the
bers, classes, space-time points, bodies, minds, etc., are pseudo questions others about real nirmbers, etc., should not be taken literally but merely

44
 THEORETICAL CONCEPTS
Rudolf Carnap
as a didactic help by attaching familiar labels to certain kinds of entities certain sentences of LT or vice versa. They serve indirectly for deriva-
or, to say it in a still more cautious way, to certain kinds of expressions tions of conclusions in Lo, e.g., predictions of observable events, from
in LT. Let the expressions corresponding to the domain 1 be "0," "0," given premiies in Lo, e.g., reports of results found by observation, or
"O"," etc. T o say that " 0 designates the number zero, "0"' the num- for the determination of the probability of a conclusion in Lo on the
ber one, etc., gives merely the psychological help of connecting these basis of given premises in Lo. Since both the premises and the conclusion
expressions for the reader with useful associations and images, but should belong to Lo, which fulfills the restricting requirements, there can be
not be regarded as specifying part of the interpretation of LT. All the no objection against the use of the C-rules and of LT, as far as the mean-
interpretation (in the strict sense of this term, i.e., observational inter- ingfulness of the results of the derivation procedure is concerned.
pretation) that can be given for LT is given in the C-rules, and their
function is essentially the interpretation of certain sentences containing
V. The Correspondence Rules C
descriptive terms, and thereby indirectly the interpretation of the de- There is no independent interpretation for LT. The system T is in
scriptive terms of VT. On the other hand, the essential service that the itself an uninterpreted postulate system. The terms of VT obtain only
expressions "0"etc. give, consists in the fact that they represent a par- an indirect and incomplete interpretation by the fact that some of them
ticular kind of structure (viz., a sequence with an initial member but are connected by the rules C with observational terms, and the remain-
no terminal member). Thus the structure can be uniquely specified but ing terms of VT are connected with the first ones by the postulates of T.
the elements of the structure cannot. Not because we are ignorant of Thus it is clear that the rules C are essential; without them the terms
their nature; rather, there is no question of their nature. But then, since of VT would not have any observational significance. These rules must
the sequence of natural numbers is the most elementary and familiar be such that they connect sentences of Lo with certain sentences of LT,
example of the sequential structure here in question, no harm is done for instance, by making a derivation in the one or the other direction
in saying that those expressions designate entities and that these entities possible. The particular form chosen for the rules C is not essential.
are the natural numbers, as long as we are not misled by these formula- They might be formulated as rules of inference or as postulates. Since
tions into asking metaphysical pseudo questions. we assume that the logical structure of the language is sufficiently rich
In the earlier discussion of the observatioh language LO (Section 11), to contain all necessary connectives, we may assume that the rules C
we considered certain restrictive requirements, like those of nominalism, are formulated as postulates. Let C be the conjunction of these corre-
finitism, etc., and found them acceptable. However, the situation with spondence postulates. As an example, we may think of LT as a language
respect to the theoretical language is entirely different. For LT we do of theoretical physics, based on a space-time coordinate system. Among
not claim to have a complete interpretation, but only the indirect and the rules C there will be some basic ones, concerning space-time designa-
partial interpretation given by the correspondence rules. Therefore, we tions. They may specify a method for finding the coordinates of any
should feel free to choose the logical structure of this language as it best observationally specified location, e.g., the method used by navigators
fits our needs for the purpose for which the language is constructed. for determining the position (the spatial coordinates: longitude, lati-
Thus here in LT there is no reason against the three conventions, tude, and altitude) and time. In other words, these C-rules specify the
although their acceptance violates the first five requirements mentioned relation R which holds between any observable location u and the co-
in Section 11. First, before the C-rules are given, LT, with the postulates ordinates x, y, z, t, where x, y, z are the spatial coordinates and t is the
T and the rules of deduction, is an uninterpreted calculus. Therefore time coordinate of u. More exactly speaking, the relation R relates to
the earlier requirements cannot be applied to it. W e are free in the con- an observable space-time region u, e.g., an observable event or thing,
struction of the calculus; there is no lack of clarity, provided the rules a class u' of coordinate quadruples which may be specified by intervals
of the calculus are clearly given. Then the C-rules are added. A11 they around the coordinate values x, y, z, t.
do is, in effect, to permit the derivation of certain sentences of Lo from On the basis of these C-rules for space-time designations, other C-rules
47
 THEORETICAL CONCEPTS
Rudolf Carnap
as not containing references to any particular position in space or in
are given for terms of VT,e.g., for some simple physical magnitudes like
time.
mass, temperature, and the like. These rules are spatiotemporally gen-
eral, i.e., they hold for any space-time location. They will usually connect In the above examples, the C-rules have the form of universal postu-
lates. A more general form would be that of statistical laws involving
only very special kinds of value-distributions of the theoretical magni-
the concept of statistical probability (which means roughly, relative fre-
tude in question with an observable event. For example, a rule might
quency in the long run). A postulate of this kind might say, for example,
refer to two material bodies u and v (i.e., observable at locations u and
that, if a region has a certain state specified in theoretical terms, then
v); they must be neither too small nor too large for an cbserver to see
there is a probability of 0.8 that a certain observable event occurs (which
them and to take them in his hands. The rule may connect the theo-
means that, on the average, in 80 per cent of those cases this event
retical term "mass" with the observable predicate "heavier than" as
follows: "If u is heavier than v, the mass of u' (i.e., the mass of the occurs). Or it might, conversely, state the probability for the theoretical
property, with respect to the observable event. Statistical correspondence
coordinate region u' corresponding to u ) is greater than the mass of v'."
rules have so far been studied very little. (The probability conception
Another rule may connect the theoretical tirm "temperature" with the
observable predicate "warmer than" in this way: "If u is warmer than v, of the psi-functions in quantum mechanics might perhaps be regarded
as an example of probabilistic C-rules, as some customary formulations
then the temperature of u' is higher than that of v'."
by physicists would suggest. I think, however, that this conception con-
As these examples show, the C-rules effect a connection only between
stitutes a probability connection within LTrather than between LT and
certain sentences of a very special kind in LTand sentences in Lo.The
earlier view, that for some terms of Vr there could be definitions in Lo. What physicists often call "observable magnitudes," e.g., mass, posi-
tion, velocity, energy, frequency of waves, and the like, are not "observ-
terms of Vo,called either 'correlative definitions' (Reichenbach) or
able" in the sense customary in philosophical discussions of methodol-
'operational definitions' (Bridgman), has been abandoned by most em-
ogy, and therefore belong to the theoretical concepts in our terminology.)
piricists as an oversimplification (see Section X). The essential incom-
For the sake of simplicity, in most of my discussions here I shall think
pleteness of the interpretation of theoretical terms was pointed out in
of the C-rules as postulates of universal form.
my Foundations of Logic and Mathematics (6) and is discussed in
detail by Hempel in (15, $3) and (16, $7). Moreover, it cannot be
VI. A Criterion of Significance for Theoretical Terms
required that there is a C-rule for every term of VT.If we have C-rules
for certain terms, and these terms are connected with other terms by My task is to explicate the concept of the empirical meaningfulness
the postulates of T, then these other terms thereby also acquire observa- of theoretical terms. I shall use "empirical significance" or, for short,
tional significance. This fact shows that the specification, not only of "significance" as a technical expression for the desired explication. In
the rules C, but also of the postulates T, is essential for the problem of preparation for the task of explication, let me try to clarify the expli-
meaningfulness. The definition of meaningfulness must be relative to candum somewhat more, i.e., the concept of empirical meaningfulness
a theory T, because the same term may be meaningful with respect to in its presystematic sense. Let 'M' be a theoretical term of VT ; it may
one theory but meaningless with respect to another. designate a physical magnitude M. What does it mean for 'M' to be
In order to have a more concrete picture, we may think of the terms empirically meaningful? Ruughly speaking, it means that a certain
of VT as quantitative physical magnitudes, e.g., functions from space- assumption involving the magnitude M makes a difference for the
time-points (or finite space-time-regions) to real numbers (or n-tuples prediction of an observable event. More specifically, there must be a
of real numbers). The postulates T may be conceived of as representing certain sentence SM about M such that we can infer with its help a
the fundamental laws of physics, not other physical statements, however sentence So in Lo. (The inference may be either deductive, as I shall
well established. Let us think of the postulates T and the rules C as take it to be in the following discussion, or, more generally, probabi-
being completely general with respect to space and time-that is listic.) It is, of course, not required that Sois derivable from SM alone.
 Rudolf Carnap THEORETICAL CONCEPTS

It is clear that we may use in the deduction the postulates T and the versa. However, it should be noted first that the theory T which is here
rules C. If now Sx contains not only 'M' but also other terms of VT, presupposed in the examination of the significance of a term, contains
then the fact that So is deducible does not prove that 'M' is meaning- only the postulates, that is, the fundamental laws of science, and not
ful, because this fact may just be due to the occurrence of the other other scientifically asserted sentences, e.g., those describing single facts.
terms. Therefore I shall require that Sx contain 'M' as the only term Therefore the class of the terms of LT admitted as significant is not
of VT . Now it may be that any assumption involving only the magni- changed whenever new facts are discovered. This class will generally
tude M is in itself too weak to lead to an observational consequence, be changed only when a radical revolution in the system of science
and that we have to add a second assumption SK containing other terms is made, especiallyvby the introduction of a new primitive theoretical
of VT but not 'M'. Let K be the class of these other terms. For example, term and the addition of postulates for that term. And note further that
Sx may say that, at a certain space-time point, M has the value 5, and the criterion here proposed is such that, although the whole of the
Sg may say that, at the same space-time point or in its surroundings, theory T is presupposed in the criterion, the question of significance is
certain other magnitudes have specified values. If So can be deduced still raised for each term separately, not only for the vocabulary V T as
from the four premises SM,SK, T, and C, while it cannot be deduced a whole.
from SK, T, and C alone, then the sentence Sar makes a difference for On the basis of the preceding considerations, I shall now give defini-
the prediction of an observable event, and therefore has observational tions for the concept of significance of descriptive terms in the theoreti-
significance. Since 'M' is the only descriptive term in Sx, 'M' itself has cal language. The definition Dl will define the auxiliary concept of
observational significance. However, this result must be qualified by a relative significance, i.e., the significance of 'M' relative to a class K of
proviso. Since we have used the second assumption S K involving the other terms. Then the concept of significance itself will be defined in
terms of K, we have shown only that 'M' is meaningful provided that D2. According to our previous considerations, the concept of significance
the terms of K are meaningful. For this reason the definition of the - - LT, the observa-
must furthermore be relative to the theoretical language
significance of 'M' must be made relative not only to T and C, but also tion language Lo, the set of postulates T, and the correspondence rules
to the class K. 'M' is shown by the indicated procedure to be significant C. W e presuppose that the specifications of the languages LT and Lo
provided the terms of K have been found by a previous examination contain alsq a specification of the classes of descriptive terms, that is,
to be significant. Therefore the terms of VT must be examined in a VT and Vo, respectively.
serial order. The first terms of VT must be such that they can be shown Dl. A term 'M' is significant relative to the class K of terms, with
to be significant without presupposing the significance of other descrip respect to LT, Lo, T, and C = the terms of K belong to VT, 'M'
tive terms. This will be the case for certain terms of VT which are belongs to VT but not to K, and there are three sentences, SMand SK
directly connected by C-rules with Lo. Other terms of VT can then be in LT and So in Lo, such that the following conditions are fulfilled:
shown to be significant by using the proved significance of the first (a) SH contains 'M' as the only descriptive term.
(b) The descriptive terms in SK belong to K.
terms, and so on. The total V, can be regarded as significant only if (c) The conjunction Sr.SK.T. C is consistent (i.e., not logically false).
we can show for a certain sequence of its terms that each term is sig- (d) So is logically implied by the conjunction S,.SK.T.C.
nificant relative to the class of the terms preceding it in the sequence. (e) So is not logically implied by SK.T. C.
It is clear that the definition must be relative to T, because the ques- The condition (c) is only added to assure that the situation described
tion whether a certain term in LT is significant cannot possibly be in SMand Sg is possible, i.e., not excluded by the postulates T and the
decided without taking into consideration the postulates by which it C-rules; otherwise the condition ( d ) would be trivially fulfilled.
D2. A term 'M,'is significant with respect to LT, Lo, T and C = Dt
is introduced. Perhaps the objection might be raised that, if significance there is a sequence of terms 'M,', . . ., 'M,,' of VT, such that every term
is dependent upon T, then any observation of a new fact may compel 'Mt (i = 1, . . ., n) is significant relative to the class of those terms
us to take as nonsignificant a term so far regarded as significant or vice 1 which precede it in the sequence, with respect to LT, Lo, T, and C.
 Rudolf Carnap THEORETICAL CONCEPTS

The sequence of terms referred to in D2 must obviously be such that introduced. Hempel has given in his article (15) a clear survey of this
the first term 'MI7can be shown to be significant without the help of development. One change was the replacement of the principle of veri-
other terms of VT.In this case 'MI7 satisfies D l ; the class K is the null fiability by the weaker requirement of confirmability or testability, as
class; the sentence SK contains no descriptive terms; it is logically true formulated in my paper (5). At the time of that paper, I still believed
and can therefore be omitted. In the simplest case of this kind, 'MI7 that all scientific terms could be introduced as disposition terms on the
occurs in a C-rule, like "mass7' and "temperature" in our previous basis of observation terms either by explicit definitions or by so-called
examples. Suppose that the first three terms of our sequence are of the reduction sentences, which constitute a kind of conditional definition
kind described. Then for the fourth term, the sentence SK may contain (see Section X). Today I think, in agreement with most empiricists, that
any one or all three of these terms. In this way we may proceed, step the connection between the observation terms and the terms of theo-
by step, to other terms, which may be more and more remote from retical science is much more indirect and weak than it was conceived
direct observation. either in my earlier formulations or in those of operationism. Therefore
(A slightly stronger criterion might be taken into consideration, ob- a criterion of significance for L T must likewise be very weak.
tained by the following modification of D l . In addition to the sentence 8 In discussions of the requirement of confirmability (or, in earlier
SY7another sentence SYis used, which contains likewise 'M' as the times, verifiability) the question was sometimes raised whether the
only descriptive term. Then the analogue to condition (c) for Sr is possibility of the event which constitutes the confirming evidence was
added, and furthermore the analogue to condition (d) with S'r taking to be understood as logical possibility or,a<c'idsal possibility (i.e., com-
the place of SY and the negation of So taking the place of So. Thus here patibility with the laws of nature or the laws of a given theory). Accord-
the assumption SH leads to an observable consequence, as in D l , but ing to Schlick's conception (22, p. 153) the possibility should be under-
another assumption SYabout M, incompatible with Sr, leads to an- stood in the widest sense, as logical possibility. His main argument was
other observable consequence. However, the simpler criterion stated in the uncertainty about possibility in an empirical sense. He pointed out
D l seems sufficient as a minimum requirement for significance.) that the observer does not know whether certain operations are empiri-
In the informal discussion at the beginning of this section, I have cally possible for him or not. For example, he does not know whether
referred to the deduction of So from certain premises. Correspondingly, he is able to lift this table; he is quite certain that he cannot lift an
D l (d) requires that So is logically implied by the premises. However, this automobile; but both events are still conceivable and should therefore
simple situation holds only if the C-postulates have universal form, as I be regarded as possible evidence. Schlick's point was that a question
we mostly assume in our discussions. In the more general case that also of significance should never be dependent upon contingent facts.
statistical laws are admitted as C-postulates (see the remark at the end 8 On the other hand, Reichenbach and I (5, p. 423) maintained the
of Section V) and perhaps also as postulates of T, then the result is a view that logical possibility is not sufficient, but that physical (or, more
probability connection between SM.SK on the one hand, and So on the generally, causal) possibility is required. The question whether a given
other. In this case, the conditions (d) and (e) in Dl are to be replaced sentence of L T is confirmable must be taken as relative to a theory T.
by the condition that the probability of So relative to SH.SK, presuppos- In examining such a question, a proposed evidence or a proposed test
ing T and C, is different from the prob?bility of So relative to SK alone. procedure could certainly not be accepted if they were incompatible
with T. For example, on the basis of modern physics, which takes the
VII. T h e Adequacy of t h e Criterion of Significance velocity of light as the maximum signal velocity, any proposed test or
The criterion here proposed is admittedly very weak. But this is a evidence involving a signal with a higher velocity could not be accepted
result of the development of empiricism in these last decades. The as proof of significance. The definition Dl is based on this conception.
original formulations of the criterion were found to be too strong and The conjunction Su.SK.T.C is required to be consistent by condi-
too narrow. Therefore, step by step, more liberal formulations were tion (c). Since So is logically implied by this conjunction, SM.SK.Sois
53
 THEORETICAL CONCEPTS
Rudolf Carnap
compatible with T and C and thus causally possible. However, it is 'This criterion is, however, not generally adequate. It would, for
to be noted that causal possibility as here understood is much weaker example, not work for a theory T" logically equivalent to T' but such
than the kind of empirical possibility which Schlick had seemed to that no postulate of T" is isolated. Those who are sceptical about the
have in mind. In Schlick's example, neither the lifting of the table possibility of a criterion of significance for LT have probably a situation
nor that of the automobile is excluded by our criterion, because these of this kind in mind. (Hempel discusses a similar example.) They be-
events are not incompatible with the T (and C); T contains only the lieve that it is not possible to give a criterion for postulate systems like
fundamental laws of science, while those events are merely excluded T". However, I think that the criterion for terms proposed in Section VI
by our empirical knowledge of the observer's ability to lift things. is adequate for cases of this kind. Consider for the postulate system T"
I shall now examine the question of the adequacy of our criterion in the sequence of terms which is required in 0 2 . This sequence must
more specific terms. Let us consider the case that the vocabulary VT necessarily begin with physical terms of V l , because, according to our
consists of two parts, V, and V, ,such that the terms of V, are empiri- assumption ( I ) , there are no C-rules for any of the metaphysical terms
cally meaningful, while those of V, are entirely devoid of any empirical of V,. Then the sequence may go on to further physical terms, which
meaning. To make this presupposition about V, and V2 more specific, are connected with Lo not directly by C-rules, but indirectly by other
we assume the following: physical terms. Now we shall see that the sequence cannot reach any
(1) If S , and S 2 are any sentences of L such that all descriptive terms term of V,; thus our criterion is not too broad for systems like T".
of S1 belong to Vl or to the observational vocabulary Vo and those of W e will show this by an indirect proof. W e assume that the sequence
S, to V, ,then neither of the two sentences logically implies the other, reaches terms of V,; let 'M' be the first term of V, in the sequence;
unless the implying sentence is logically false or the implied sentence is hence the preceding terms belong to V,, and thus are meaningful. 'M'
logically true. is significant relative to the class K of the preceding terms, with respect
Now a proposed criterion for the significance of terms of VT should to LT, Lo, T", and C, in the sense of D l . Intuitively speaking, 'M7must
be regarded as too narrow if it exc1uded.a term of V,, and as too broad then be meaningful, in contradiction to our presupposition about V,.
if it admitted a term of V, . I t would be adequate only if it were neither Our task is, to derive formally a contradiction with the presupposi-
too narrow nor too broad. tion (1).
For example, we might think of V1 as containing terms of physics, According to Dl (d) :
and of V, as containing meaningless terms of speculative metaphysics
(2) SM.SK.Tm.C 3 So is logically true.
such that the supposition (1) holds.
First let us consider a postulate system T' consisting of two parts, TI Now T" is logically equivalent to T and thus to Tf1.T', . Hence we
and T',, T', containing only terms of V,, and T,only terms of V,. obtain from (2) with a simple transformation:
T', may, for example, consist of fundamental laws of physics, and T', (3) SM.T', 3 U i s logically true, where U is S,.T', .C 3 So.
of metaphysical principles. A criterion of significance which is adequate Hence :
in this special case can easily be given. W e call a postulate of a system
(4) SM.T'Z logically implies U .
T an isolated postulate if its omission from T does not diminish the
class of sentences in Lo which are deducible from T with the help of Now all descriptive terms in SM.T', belong to V,, and those in U
the C-rules. Then we take a tenn of VT as significant if it occurs in a belong to V, or Vo . Thus (4) is in contradiction to ( I ) , because
C-rule or in a non-isolated postulate of T. In the case of the above (5) SM.T', is not logically false (by Dl ( c ) ) , and
system T', according to ( l ) , all postulates of T', and no others are (6) U is not logically true (by Dl (e) ) .
isolated; therefore all terms of V, and no others fulfill the criterion
This shows that the sequence cannot reach the terms of V,.
of significance just mentioned.

--
54
--
-- ----
 Rudolf Carnap THEORETICAL CONCEPTS

W e have shown that our criterion is not too broad if the given set where a' is the set of coordinates corresponding to the location a re-
ferred to in So. Finally we take SK and Sar as follows:
of postulates T" is logically equivalent to a set T' which consists of two
(SK) M1(a') = M', (a'),
parts, one containing only meaningful terms of V,, the other only (S,) M,(a1) = 4.
meaningless terms of V,. The situation would be different for a theory Now we can derive from Sx with A,:
T that did not fulfill this condition. In this case, T must include a
(9 M',(al) = 5,
postu1ate.A such that A contains terms from both Vl and V,, but A is
hence with SK:
not logically equivalent to a conjunction AI.A8 in which Al contains (ii) Mlaf = 5,
only terms of V1, and A, only terms of V,. But such a postulate A and hence with C,:
would express a genuine connection between the occurring terms of (iii) So.
V, and those of V,. Therefore these terms of V, would not be entirely
devoid of empirical meaning, against our assumption. Thus the condition (d) in D l is fulfilled. Therefore, 'M,' is signifi-
cant relative to the class K of the terms 'MI' and 'M','.
The result that our criterion of significance is not too broad depends
W e have just seen that, in the definition of the significance of 'M'
essentially on the following feature of our definitions. W e refer in D2
relative to K, we must not admit a meaningless term in K and thereby
to a sequence of terms, and we require in effect for the significance of
in the additional assumption SK, because otherwise an observation sen-
a term 'M' of the sequence that 'M' is significant (in the sense of D l )
tence could be derived, leading to a deceptive appearance of significance.
relative to the class K of the terms which precede 'M' in the sequence
This is indeed excluded by 0 2 . However, DI allows other premises for
and which therefore have already been found to be significant. W e can the derivation which contain meaningless terms, viz., postulates of T.
easily see that the criterion would become too broad if we were to Not only the postulates which contain the meaningful terms of V1 and
change D2 so as to give up the requirement just mentioned. More the term 'M' in question are allowed but also postulates containing any

2
specifically, we can ow the following. A meaningless term ' M i of V2
can, according to I, be significant relative to a class K which contains,
in addition to terms of V,, also a meaningless term of V2 different from
terms of V,. Could this not lead to the same false appearance of sig-
nificance for an actually meaningless term 'M' as the use of meaning-
less terms in SK would do? In the above example, SK connected a mean-
'M,', say 'M',.' W e shall show this first informally. The decisive point ingless term 'M',' with a meaningful term 'MI7, and this fact led to the
is that now, in distinction to our actual definition 0 2 , we can have as undesired result. Now the use of T would lead to the same result if a
the additional assumption SK a sentence connecting the meaningless postulate of T were to make a connection between those terms. For
term 'MtS7with a meaningful (physical) term of V,, say 'Mi.' Now example, a postulate might yield as an instance the sentence "M,(al) =
there may be a (metaphysical) postulate A, of T which connects MI M',(af)" which was used as Sg in the earlier example. Thus the same
with M', . With the help of this postulate, we can derive from the observation sentence So could be derived from S, even without the use
assumption Sy about M, alone a sentence about M',;from this with of any second assumption SK.AS an alternative, a postulate might state
the sentence Sg mentioned above a physical sentence about MI, and a connection between 'M',' and 'MI7 in a conditional form, which,
from this with a suitable C-rule an observation sentence. though weaker, would likewise make possible a derivation of an observa-
The formal derivation is as follows. W e take as a postulate of T: tion sentence. Does then the fact that Dl permits the use of all postu-
(A,) For every space-time point, the value of Mf2 is higher than that lates T make this definition inadequate? It does not, because the occur-
of Ms by one. rence of a postulate making a genuine connection between a term of V,
W e take as an instance of a C-rule: and one of V, is excluded by our presupposition that the terms of
(C,) Ml(af) = 5 3 SO, Vl are meaningful and those of V, meaningless. By virtue of such a
 Rudolf Carnap
postulate, the term of V, (in the example, 'M',') would obtain some conclusion So from SM,we need, in addition to T and C and the assump
measure of empirical meaning, as we observed earlier in this section tion SK in theoretical terms, some assumption in observation terms,
with reference to the postulate A. The essential difference between the say So?This might well happen. But then the conditional sentence
two cases is the following. If a sentence connecting a meaningful term S'O3 So is derivable from the premises specified in Dl, and this is a
with another term in an inseparable way (e.g., by an equation, a con- sentence in Lo. Thus 'M' would fulfill Dl, with the conditional sen-
ditional, a disjunction or the like, in distinction to a conjunction, which tence taking the place of So.
can be separated into its components) is a postulate or provable on l c and 2. The condition (a) in D l requires that Sx contain 'M' as
the basis of postulates, then it is stated as holding with physical neces- the only descriptive term. The question might be raised whether this
sity; therefore it conveys some empirical meaning on the second term. requirement is not too strong. Could not the following situation occur?
On the other hand, if the same sentence is not provable but is merely
--
used as the additional assumption SK in Dl, t h a t has no such effect;
'M' and the terms of K are meaningful, and So can indeed be derived
with the help of T and C from an assumption S containing no other
it need Z even be true. - -
descriptive terms than 'M' and the terms of K, but S cannot be split
The Preceding considerations have shown that our criterion of signifi- up into two sentences SMand SK such that SH contains only 'M' and
b
cance, formulated in D1 and D2, is not too liberal. It does not admit SK does not contain 'M.' Let us assume that the sentence S refers to
a term entirely devoid of empirical meaning. Now we shall consider space-time points of a certain spatioternporal region a'. Then we can
the question whether the criterion might be too narrow. Suppose that form sentences SM and SK which fulfill the requirements of D l in the
the term 'M' has some empirical meaning. Then it will be possible to following way. Since S is supposed to be compatible with T and C, there
derive an observation sentence from a suitable assumption S involving must be a possible distribution of values of M for the space-time points
'M' and other terms. Could it then still happen that our criterion would -
of the region a', which is compatible with T, C, and S. Let 'F' be a
exclude 'M'? The definitions D l and D2, while permitting the inclusion logical constant, designating a mathematical function which represents
of all postulates T and C among the premises for the derivation of the such d value distribution. Then we take the following sentence as Sv:
observation sentence, allow in addition only the two sentences Sg and d' /.
For every space-t~me point in a', the value of M is equal to that of F."
SM,for which specific restrictions are stated, especially the following: This sentence SMis compatible with T.C.S. Then we take as SK the
(1) SK may contain only terms of VT which are different from 'M' and sentence formed from S by replacing the descriptive term 'M' by the
have to be significant; hence the following terms are not allowed in SK: logical constant 'F'. Then SMcontains 'M' as the only descriptive term
(a) terms of V,, and Sg contains only terms of K. Furthermore, S is logically implied by
(b) terms of Vo, Sar,and SK. SOis logically implied by S.T.C., according to our assump-
(c) The term 'M'. tion, and hence also by SBI.SK.T.C.Therefore 'M' fulfills the defini-
(2) SMcontains 'M' as the only descriptive term. tion D l .
Thus we have not found a point in which our criterion is too narrow.
W e will now examine whether these restrictions are narrower than
is necessary and thus might lead to the exclusion of a meaningful term
VIII. A Criterion of Significance for Theoretical Sentences
'M.'
The following two problems are closely connected with each other:
la. W e found earlier that it is necessary to exclude the terms of V, first, the problem of a criterion of significance for descriptive constants,
from SK, because otherwise the criterion would become too broad. and second, the problem of the logical forms to be admitted for sen-
Ib. Is it necessary to exclude the observational terms Vo from the tences. For the theoretical language, the connection between these
premises? Could it not be that, for the derivation of an observational problems is still closer than for the observation language. In the latter,
 Rudolf Carnap THEORETICAL CONCEPTS

we may decide to have primitive predicates like "blue," "cold," "warmer example, verifiability as a condition for the significance of a sentence
than," and the like, while we are still undecided as to the forms of was sometimes understood in the strict sense of the actual possibility
sentences, especially of general sentences, and the structure of the logic of carrying out a procedure which would lead either to a verification
to be built into the language. On the other hand, if we wish to have or a falsification of the sentence. According to this criterion, in con-
terms like "temperature," "electromagnetic field," etc. as primitives in trast to D3, the significance of a sentence is not only dependent upon
Lh then we need also the accepted postulates for them, and thus we its logical form and the nature of the'descriptive constants occurring
have to admit real number expressions, general sentences with real in it, but also upon the location in s s e and time referred to and the
number variables, etc. development Gfctechnology. For E m p l e , an empEcist applying this
It seems to me that the best approach to the problem of a criterion narrow criterion would regard as significant a sentence ascribing an
of significance for sentences is the following. W e look first for solutions observable property P to a body in his laboratory, while he would reject
to the two problems mentioned above; and then we take the most lib- as nonsignificant another sentence which ascribes the same property to
eral criterion of significance for sentences which is compatible with those a body not accessible to him or not accessible to any human being, e.g.,
solutions. That is to say, we then accept as a significant sentence any because of technical difficulties or remoteness in space or time.
expression that has any of the admitted logical forms and contains only However, even at the time of the Vienna Circle, we did not interpret
descriptive constants which are significant. ( I have used a similar the principle of verifiability in this narrow sense. W e emphasized that
approach for Lo in (5) .) I propose to apply this procedure now to LT. the principle required, not the actual possibility of determination as
A criterion of significance for descriptive terms was given in Section true or false, but only the possibility in principle. By this qualification
VI. Some of the questions concerning the logical forms of sentences we intended to admit cases in which the determination was prevented
were discussed in Section IV, especially the question of the kinds of only by technical limitations or by remoteness in space or time. W e
variables to be admitted in universal and existential quantifiers. W e accepted, for example, a sentence about a mountain on the other side
decided to admit at least those kinds of variables and forms of sen- of the moon as meaningful. W e stated the general rule that, if a de-
tences which are essential for classical mathematics. Without actually scription of an event in our neighborhood is regarded as meaningful,
specifying here the details of the rules, we shall now assume that the then an analogous description of an event in prehistoric times, or an
logical forms of sentences have been chosen on the basis of the con- event on the earth before there were human beings, or before there
siderations in Section IV, and that the rules of formation for LT have were any organisms, or at a future time when human beings will not
been laid down in accordance with this choice. Then, applying the exist any more, should likewise be accepted as meaningful. On the basis
procedure proposed above, we define as follows: of this conception, the space-time location referred to in a sentence
was regarded as irrelevant for the question of meaningfulness; this is
0 3 . An expression A of LT is a significant sentence of L = Df in accord with D3.
(a) A satisfies the rules of formation of LT, If D3 is accepted and, in line with our earlier considerations in Sec-
(b) every descriptive constant in A is a'significant term (in the sense tion IV, all constants, variables and forms of sentences of classical
of D2). mathematics are admitted in LT, then the class of significant sentences
The procedure used in this definition might perhaps appear as obvi- of LT is very comprehensive. W e must realize that it includes certain
ous. However, a closer examination shows that this is not the case. In sentences for which no observational evidence can ever be relevant, e.g.,
fact, this form of the definition (aside from the question of its content, the sentence: "The value of the magnitude M at a certain space-time
i.e., the choice of the particular rules of formation and of the particular point is a rational number," where 'M' is significant. But every physicist
significance criterion for terms) is not in agreement with certain very would reject a language of physics so restricted that sentences of this
narrow criteria of significance which were sometimes proposed. For and similar kinds were excluded. He would regard their inclusion as a
 Rudolf Carnap THEORETICAL CONCEPTS
neglighle price to be paid for the great convenience of using the whole discussed in detail in this article. The other kind I will call (pure) disposi-
of classical mathematics. It seems to me that no serious objections can tion terms. In my view, they occupy an intermediate position between
be raised against these sentences, since it is in any case not possible to the observational terms of Lo and the theoretical terms; they are more
give an observational interpretation for more than a small part of the closely related to the former than to the latter. The name 'observation
sentences of LT. W e should require no more than that for such a mag- language' may be understood in a narrower or in a wider sense; the
nitude there are certain sentences which have an influence on the pre- observation language in the wider sense includes the disposition terms.
diction of observable events and thus the magnitude itself has some In this article I take the observation language Lo in the narrower sense.
amount of observational meaning. All primitive predicates in this language designate directly observable
I wish to emphasize that the proposed criterion for the significance properties or relations of observable things or events; and a nonprimi-
of sentences is not meant to guarantee the fruitfulness of T. If all terms tive term is admitted in Lo only if it can be defined on the basis of
of VT fulfill 0 2 and the postulates T are in accord with the rules of the primitive terms by an explicit definition in an extensional form,
formation, then these postulates are indeed regarded as significant. But that is, not involving either logical or causal modalities. The extended
this should by no means be understood as implying that T must then observation language Eo is constructed from the original observation
be a scientifically satisfactory theory. T may still contain postulates language Lo by the addition of new terms in a way now to be described.
which are of very little use from a scientific point of view. But the Suppose that there is a general regularity in the behavior of a given
question of scientific fruitfulness of sentences and of a theory should thing of such a kind that, whenever the condition S holds for the thing
be clearly distinguished from the question of empirical significance. or its environment, the event R occurs at the thing. In this case we
There is no sharp boundary line between fruitful and useless hypotheses shall say that the thing has the disposition to react to S by R, or for
or theories: this is rather a matter of degree. It seems even doubtful short, that it has the property DaE. For example, elasticity is a disposi-
whether it is possible to formulate in a completely general way a defini- tion of this kind; a thing is called elastic if it shows the following regu-
tion of a quantitative degree of fruitfulness of a scientific theory. larity: whenever it is slightly deformed and then released (S), it resumes
It should be noted that the significance criterion for LT cannot be its original form ( R ) . Or, an animal may have the disposition to react
simply absorbed into the rules of formation. These rules determine only to a light in an otherwise dark environment (S), by appdching the
the forms of sentences, not the choice of primitive descriptive terms. light ( R ) . Thus, S is sometimes a stimulus, and R is the response
The significance of these terms depends on other rules of L , viz., the characteristic for the disposition in question (if we allow ourselves to
list of postulates T and of C-postulates and the rules of logical deduc- use the terms 'stimulus' and 'response' not only in their literal sense
tion, as a glance at the essential condition (d) in Dl shows. (The rules applied to certain-processes in organisms, as in the last example, but
of deduction may be given either in a syntactical form, as rules of deriva- in a wider sense also to processes with inorganic bodies). When both
tion in a calculus, or in a semantical form, in terms of logical implication. S and R are specified, then the disposition concept DaR is thereby com-
I have used in Dl the latter form because it is more comprehensive; pletely characterized in its meaning. If both S and R can be described
it presupposes rules specifying models and ranges, not given in this in Eo, then we admit the introduction of the disposition term 'DRR'
article.) as a new predicate in Eo.The introduction of the first disposition terms
IX. Uisposition Concepts
,
" -R_ PU-

pressible in Lo. But once some disposition terms have been introduced
Among the descriptive terms which do not belong to the observa- in this way, then further disposition terms may be introduced in such
tion language Lo there are two different kinds, which today, in distinc- a way that S and R are described by using not only the terms of Lo,
tion to my previous conception, I should like to regard as essentially but also the previously introduced disposition terms of Eo.
different. One kind is that of the theoretical terms, which we have (We will not discuss here the possible forms for the rule by which a
 Rudolf Carnap THEORETICAL CONCEPTS

disposition term is introduced on the basis of given S and R. This in- test condition S are sometimes called test operations. The introduction
volves some technicalities which are not necessary for our present discus- of DsB by a specification of the test operations and the characteristic
sions. I will only mention two different forms for such rules that have result R is therefore sometimes called an operational definition. There
been proposed. The first consists of so-called reduction sentences, which is actually no sharp line between observable properties and testable dis-
I proposed in (5). They represent a kind of conditional definition positions. p n observable property may be regarded as a simple special
which uses only truth-functional connectives, but no modalities. The case of a testable disposition; for example, the operation for finding out
other method uses an explicit definition .of a special form, involving whether a thing is blue or hissing or cold, consists simply in looking
logical and causal modalities; the exact form of definitions of this kind or listening or touching the thing, respectively. Nevertheless, in the
is at present not yet sufficiently clarified, but still under discussion.) reconstruction of the language it seems convenient to take some prop-
Sometimes multiple dispositions are used: D ~ ~ ~ ~ , a.d.a .,t ,s,R, is the erties, for which the test procedure is extremely simple (as in the three
disposition to react to S, by R1, to S, by R,, . . ., and finally to S, by examples just mentioned), as directly observable and use them as primi-
R,. (In (5) I proposed to introduce a conce t of this kind by several tives in Lo.
pairs of reduction sentences.) However, -it-seems
7 pre erable to admit The view has often been maintained, especially by empiricists, that
only simple dis ositions. Something Slmilar to a multiple disposition only terms of the kind just described, may be regarded as empirically
b-P
can still e expressed by a conjunction of simple dispositions. Bridg- meaningful. Thus testability was taken as a criterion of significance.
The principle of oper~tionismsays that a term is empirically meaning-
man has emphasized that, strictly speaking, for one concept not more
than one test procedure must be given. If we specify, say for "electric ful only if an operational definition can be given for it. The require-
charge," three test procedures, then thereby we have given operational ments of testability and of operationism as represented by various
definitions for three different concepts; they should be designated by authors are closely related to each other, differing only in minor details
three different terms, which are not logically equivalent. As far as dis- and in emphasis. (In my simplifying account above they even appear
position concepts are concerned, in distinction to theoretical terms, I as identical.) The principle of operationism, which was first proposed

-
would agree with Bridgman in this point.
Let us now consider an important special kind of disposition. Let
in physics by Bridgman and then applied also in other fields of science,
including psychology, had on the whole a healthy effect on the pro-
L"0 be that sublanguage of L'o, in which the introduction of a disposi- cedures of concept formation used by scientists. The principle has con-
tion term 'DSR' is permitted only if S and R are such that the observer tributed to the clarification of many concepts and has helped to elimi-
is able to produce the condition S at will (at least in suitable cases), nate unclear or even unscientific concepts. On the other hand, we must
and that he is able to find out by &%6le experiments whether the realize today that the principle is too narrow.
event R does or does not occur. In this case, by specifying S and R, a That the requirements of testability and of operationism exclude some
test procedure for the disposition DSR is given. This procedure consists empirically meaningful terms, can easily be seen. Suppose that 'S' and
in producing the test condition S and then finding out whether or not 'R' are both testable and hence accepted as meaningful by a scientist
the positive test result R occurs. If the observer finds for a given thing who takes testability as a criterion of significance. Since now the mean-
a sufficient number of positive instances, in which S is followed by R, ing of the term 'DBR9is given by the specification of S and R, there
.
IVC
.
1 1..
1
1L33

infer that the general regularity holds and thus that the thing possesses even if the condition S cannot be produced at will. In the latter case,
the disposition DSR. Let us call a disposition of this kind a "testable is not testable; but S may still occur spontaneously and then, by
disposition." The class of testable properties includes observable prop- finding R or non-R, the observer may determine whether or not DsR
erties and testable dispositions. All predicates in Lno designate testable holds. Thus it seems preferable not to impose the restriction as in Yo,
properties. The manipulations by which the experimenter produces the but to allow the general procedure as in Lro: we start with observable
 Rudolf Carnap THEORETICAL CONCEPTS

properties and allow the introduction of any disposition DBR, provided the condition S and the characteristic result R are formulated in LT-
that S and R are already expressible in our language Eo. rather than in Lo or L'o. (They might be called "theoretical disposition
(In ( 5 ) , I gave an example of a meaningful but not testable term (p. terms"; we shall not discuss them further.) The second characteristic
462) of the kind just described. I expressed there ($27) my preference distinguishes 'D8R' from any theoretical term because the latter is never
for the more general procedure (as in in comparison with that completely interpreted. In (5) I recognized this "open" character of
restricted by the requirement of testability (as in Lrto).Later it became scientific terms, that is, the incompleteness of their interpretation. At
clear by the consideration of theoretical concepts (see the next section that time I tried to do justice to this openness by admitting the addi-
of this paper) that a far more extensive liberalization of operationism tion of further dispositional rules (in the form of reduction sentences;
is needed; this was emphasized by Feigl in (7) and (10) and by see my remarks in Section IX above on multiple dispositions). I think
Hempel in (16) and (17).) now that the openness is more adequately represented in LT; whenever
additional C-rules or additional postulates are given, the interpretation
X. The Difference between Theoretical Terms and of the term may be strengthened without ever being completed.
Pure Disposition Terms The third characteristic leads to the following important consequence:
I think today that, for most of the terms in the theoretical part of (i) If the thing b has the disposition DsR and the condition S is fulfilled
science and especially in physics, it is more adequate and also more in for b, then it follows logically that the result R holds for b.
line with the actual usage of scientists, to reconstruct them as theoreti- Therefore :
cal terms in LT rather than as disposition terms in Eo.The choice of (ii) If S holds for b, but R does not, then b cannot have the disposition
the form of reconstruction depends to some extent upon the interpre- DSR.Thus, from a premise in L', not involving D 8 ~at , least a negative
tation which we wish to give to the term, and this interpretation is sentence a b u t DgR is derivable. For a theoretical term, say 'M,' the
not uniquely determined by the accepted formulations in science. The situation is different. Let SN be a sentence containing 'M' as the only
same term, say "temperature," may be interpreted, as I do interpret it, descriptive term. In the situation described in Dl in Section VI, So is
in such a way that it cannot be represented in Eobut only in LT; and, derivable from Sx and Sg (with the heIp of 7'and C, which may be
on the other hand, it may also be interpreted, e.g., by an operationist, regarded as belonging to the rules of LT), and therefore non-Snris de-
in such a way that it fulfills the requirement of operationism. I shall rivable from non-So and SH. Since Sg is not translatable into Lo or L'o,
now explain the reasons for my present view, which differs from that the situation is here different from that in (ii). It is true that, for a
stated in (5). term 'M' occurring in a C-rule, there are sentences SMand So such that
A disposition term like 'DSR7 introduced by the general method de- So is derivable from Saralone without the need of a second premise SK;
scribed in the last section (for E o ) may be called a "pure disposition and hence non-SM is derivable from non-So, so that the situation is
term" in order to emphasize that it has the following characteristic similar to that in (ii). However, this holds only for sentences of a very
features which distinguish it from terms in LT: special kind. Most of the sentences about M alone, even if 'M' is a
1. The term can be reached from predicates for observable properties term occurring in a C-rule, are such that no C-rule is directly applicable,
by one or more steps of the procedure described. and therefore the derivation of an observation sentence is more indirect
2. The specified relation between S and R constitutes the whole , + . . the
meaning of the term. term "mass," which is one of the physical terms most closely related to
3. The regularity involving S and R, on which the term is based, is observational terms. There may be C-rules for "mass" (see the example
meant as universal, i.e., holding without exception. in Section V). But no C-rule is directly applicable to a sentence S,
The first characteristic distinguishes a pure disposition term like *DzR9 ascribing a certain value of mass to a given body, if the value is either
from other disposition terms which are analogous to 'DBR7but such that so small that the body is not directly observable or so large that the
 Rudolf Carnap THEORETICAL CONCEPTS

observer cannot manipulate the body. ( I mentioned in Section V the spite of the negative result, viz., S and non-R, shows that he does not
possibility of probabilistic C-rules. If all C-rules have this form, then no take I, as the pure disposition DBRcharacterized by S and R, because,
theoretical sentence is deducible from sentences in Lo or L'o. Thus in according to (ii), this disposition is logically incompatible with the
a language of this kind, the difference between pure disposition terms negative result. The scientist will point out that the test procedure for
and theoretical terms becomes still more striking.) I, based on S and R should not be taken as absolutely reliable, but only
W e have seen that pure disposition terms and theoretical terms are with the tacit understanding "unless there are disturbing factors" or
quite different in their logical and methodological characteristics. To "provided the environment is in a normal state." Generally, the explicit
which of these two kinds do scientific terms belong? For the terms of or implicit hclusion of such an escape clause in the description of a
theoretical physics, both conceptions are represented among leading test procedure for a concept M in terms of a condition S and a result R
physicists. Bridgrnan interprets them in such a manner that they fulfill shows that M is not the pure disposition D9& Also, the name "opera-
the requirement of operationism and thus are essentially pure disposi- tional definition" for the description of the test procedure is in this
tions. On the other hand, Henry Margenau emphasizes the importance case misleading; a rule for the application of a term that permits pos-
of the method of introducing these terms by postulates and connecting sible exceptions should not be called a "definition" because it is obvi-
only certain statements involving them with statements about observ- ously not a complete specification of the meaning of the term.
ables; in this conception they are theoretical terms. On the other hand, if the term in question, e.g., 'I,', is a theoretical
It seems to me that the interpretation of scientific terms as pure dis- term, then the description of the test procedure involving S and R may
positions cannot easily be reconciled with certain customary ways of well admit of exceptions in case of txnnsua! disturbing facters. Fer
using them. According to (ii), the negative result of a test for a dis- example, it may be possible to derive from the postulates T, the C-rules,
position must be taken as conclusive proof that the disposition is not and factual premises about usual circumstances in a laboratory the con-
present. But a scientist, when confronted with the negative result of a clusion that, if there is no strong current, there will not be a strong
test for a certain concept, will often still maintain that it holds, pro- deflection of the needle, except in the case of unusual circumstances
vided he has sufficient positive evidence to outbalance the one negative like a magnetic field from another source, a strong current of air, or
result. For example, let I, be the property of a wire carrying at the the like.
time to no electric current of more than 0.1 ampere. There are many Thus, if a scientist has decided to use a certain term 'M' in such a
test procedures for this property, among them one in which the test gay, that for certain sentences about M, any possible observational
condition S consists in bringing a magnetic needle near to the wire, results can never be absolutely conclusive evidence but at best evidence
and the characteristic result R is the fact that the needle is not deflected yielding a high probability, then the appropriate place for 'M' in a
from its normal direction by more than a certain amount. Suppose that dual-language system like our system Lo-LTis in LT rather than in Lo
the observer assumes from the arrangement of the experiment that I, or Po.
holds, e.g., because he does not see any of the ordinary sources of a
current and he has obtained, in addition, positive results by some other XI. Psychological Concepts
tests for I, (or for a physically equivalent property). Then it may be The method of reconstructing the language of science by the dual
I.1 nf the

tioned test with S and R leads to a negative result, that is, a strong language LT and the distinction between pure dispositions and theo-
deflection of the needle. He may maintain I, because it is possible that retical concepts were so far in this article illustrated mostly by examples
the negative result is due to an unnoticed disturbing factor; e.g., the taken from physics. In the historical development of science, physics
deflection of the needle may be caused by a hidden magnet rather than was indeed the field in which the method of introducing terms by
. by a current in the wire. The fact that the scientist still assumes I, in postulates without a complete interpretation was first used systemati-
 Rudolf Carnap THEORETICAL CONCEPTS

cally. The beginning phase of this development may perhaps be seen ing, etc., must be recognized as a kind of observation, in principle not
in the classical mechanics of the eighteenth century; its character be- different from external observation, and therefore as a legitimate source
came more clearly recognizable in the nineteenth century, especially of knowledge, though limited by its subjective character. Secondly, Be-
in the Faraday-Maxwell theory of the electromagnetic field and the haviorism in combination with the philosophical tendencies mentioned
kinetic theory of gases. The greatest and most fruitful application is led often to the requirement that all psychological concepts must be
found in the theory of relativity and in quantum theory. defined in terms of behavior or behavior dispositions. A psychological
W e see at present the beginnings of similar developments in other concept ascribed to a person X by the investigator Y either as a momen-
fields of science, and there can be no doubt that here too the more com- tary state or process or as a continuing trait or ability, was thus inter-
prehensive use of this method will lead in time to theories much more preted as a pure disposition DgB of such a kind that S was a process
powerful for explanation and prediction than those theories which keep affecting a sensory organ of X but observable also by Y, and R was a
close to observables. Also in psychology, in these last decades, more and specified kind of behavior, likewise observable by Y. In contrast to this,
more concepts were used which show the essential features of theoreti- the interpretation of a psychological concept as a theoretical concept,
cal concepts. The germs of this development can sometimes be found although it may accept the same behavioristic test procedure based on
in much earlier periods and even, it seems to me, in some prescientific S and R, does not identify the concept (the state or trait) with the
concepts of everyday language, both in the physical and psychological pure disposition D S R . The decisive difference is this: on the basis of
field. the theoretical interpretation, the result of this or of any other test or,
In psychology still more than in phfsics, the warnings by empiricists generally, of any observations, external or internal, is not regarded as
and operationists against certain concepts, for which no sufficiently clear absolutely conclusive evidence for the state in question; it is accepted
rules of use were given, were necessary and useful. On the other hand, only as probabilistic evidence, hence at best as a reliable indicator, i.e.,
perhaps due to the too narrow limitations of the earlier principles of one yielding a high probability for the state.
empiricism and operationism, some psychologists became overcautious In analogy to what I said in the previous section about physical terms,
in the formation of new concepts. Others, whose methodological super- I wish to emphasize here for psychological terms that their interpreta-
ego was fortunately not strong enough to restrain them, dared to trans- tion as pure disposition terms is not in itself objectionable. The question
gress the accepted limits but felt uneasy about it. Some of my psycholo-
gist friends think that we empiricists are responsible for the too narrow
-
is only whether this interpretation is in accord with the way the psy-
chologist intends t o use the term, and whether it is the most useful
restrictions applied by psychologists. Perhaps they overestimate the in- for the purpose of the W ofE psychological theory, which is pre-
fluence that philosophers have on scientists in general; but maybe we sumably the explanation and prediction of human behavior. Suppose
should plead guilty to some extent. All the more should we now empha- that the psychologist Y declares that he understands the term "an IQ
size the changed conception which gives much more freedom to the higher than 130" in the sense of the pure disposition D g R to react to
working scientist in the choice of his conceptual tools. a specified kind of test S by a response of a specified kind R, where S
In a way similar to the philosophical tendencies of empiricism and and R are specified in terms of overt behavior. He is free to choose this
operationism, the psychological movement of Behaviorism had, on the interpretation provided he is consistent in it and willing to accept its
U11L
implications. Suppose that he assumes on the basis of ample previous
servation of behavior as an intersubjective and reliable basis for psycho- evidence that (at present) the person X has an IQ higher than 130.
logical investigations, while, on the other hand, it imposed too narrow Then, due to his interpretation, he is compelled to give up the assump-
restrictions. First, its total rejection of introspection was unwarranted. tion if today the test result is negative, i.e., X's response to the test S
Although many of the alleged results of introspection were indeed is not of the specified kind R. (This follows from (ii) in Section X.)
questionable, a person's awareness of his own state of imagining, feel- He cannot even re-accept the assumption later when he learns that
 Rudolf Carnap THEORETICAL CONCEPTS

during the test X was in a very depressed mood, which, however, he open to correction in view of later observations of their behavior. To
neither admitted on question nor showed in his behavior at the time the extent that someone is willing to change his judgments in this way,
of the test. Can the psychologist not escape from this embarrassing h ~ suse of psychological terms might be regarded as a beginning of the
consequence by saying that X's later admission of his depressed state development which leads finally to theoretical terms. By the way, it
showed that the condition S was actually not fulfilled? Not easily. There would be interesting to make an empirical investigation of the degree
would have to be a rule as part of the specification of S that would of rigidity and flexibility shown by non-psychologists (including phi-
enable him to make the exception. Let us consider three possibilities losophers) in making and changing psychological statements about other
for a rule. people and about themselves. This would give a clearer indication of
1. Let the rule merely say that, at the time to of the test, there must the nature of their concepts than any answers to direct questions about
be first a complete lack of any observable sign of a disturbed emotional the concepts.
state at time to and second a negative answer to a question about such The distinction between intervening variables and theoretical con-
a state. Here the condition S was actua!~fulfilled and thus the psy- structs, often discussed since the article by MacCorquodale and Meehl,
chologist has no way out. seems essentially the same or closely related to our distinction between
2. Let the rule add, moreover, that also at no later time must there pure dispositions and theoretical terms. "Theoretical construct" means
be a sign indicating a disturbance at time to.In this case, S was indeed certainly the same i s here "theoretical term7', viz., a term which cannot
not fulfilled. But a test procedure containing a rule of this kind would be explicitly defined even in an extended observation language, but is
be p&tically useless, because it could never be completed before the introduced by postulates and not completely interpreted. The inter-
death of the perscn. vening variables are said to serve merely for a more convenient formu-
3. Finally, let the rule refer not to behavioral signs but to the emo- lation of empirical laws and to be such that they can always be elimi-
tional state itself. Here the test procedure is not a strictly behavioristic nated. Therefore it seems that they would be definable in a language
procedure; I, is not defined as a behavior disposition. similar to our extended observation language Eo but containing also
If, on the other hand, "an IQ higher than 130" is taken as a theo- quantitative terms; thus they seem essentially similar to pure disposi-
retical term, the situation is entirely different. The same test procedure tions.
with S and R may still be accepted. But its specification is no longer Among empiricists, it was especially Feigl who early recognized and
regarded as an operational definition of the term. There cannot be a continually emphasized the importance of theoretical laws (which he
definition of the term on the basis of overt behavior. There may be called "existential hypotheses"; see his ( 8 ) ) . And he showed in par-
various test procedures for the same concept. But no result of a single ticular that in the present phase of psychology the use of theoretical
test nor of any number of tests is ever absolutely conclusive, although concepts and laws constitutes one of the most important methodological
they may, under favorable circumstances, yield a high probability. Any - problems and tasks. He made important contributions to the clarifica-
statement ascribing the term in question to a person -. on the basis of tion of this' problem, especially in his article (10); there he points out
a given test result-. may later b e cO'G2cted in view of new evidence, the close analogy with the earlier development of physics.
/--

even if there is no doubt t h a t - t E test rules S were fulfilled and that Psychological theories with theoretical terms will no doubt be further
the response R was made. If a psychologist accepts this non-conclusive, developed, probably to a much larger extent than so far. There are good
probabilistic character of a test, as, 1 suppose, practicaiiy aii wouid do, cny~ opmenr: of this kind wiii prove to be
then the concept tested cannot be a pure disposition and is best recon- very fruitful, while without it the possible forms of theory construction
structed as a theoretical term. are too limited to give a good chance for essential progress. This does
I think that, even on a prescientific level, many people would regard not imply that the so-called "molar" approach in terms of observable
their psychological judgments about other people as in principle always behavior is to be rejected; on the contrary, this approach will always be
73
I
Rudolf Carnap THEORETICAL CONCEPTS

I an essential part of psychological investigationgWhat is wrong is only from those of physics, is asserted by the thesis of physicalism (in its
the principle which demands a restriction of the psychological method strong sense). (My recent views on the question of physicalism are not
to this approach. The molar approach in psychology has a function yet represented in my publications. Feigl (11) explains them, describes
similar to that of macrophysics both in the historical development and the historical development of physicalism in our movement, and gives
in present research. In all fields, the study of macro-events is the natural an illuminating discussion of the theses of physicalism and the argu-
approach-in the beginning; it leads to the first explanations of facts by ments for them.) By far the greater part of the development of psy-
the discovery of general regularities among observable properties ("em- chology just outlined is, of course, today no more than a program for
pirical laws"); and it remains always indispensable as the source of the future. Views vary a great deal as to the probability and even the
confirming evidence for theories. possibility of such a development; and many will especially oppose,
In physics great progress was made only by the construction of theo- with either scientific or metaphysical arguments, the possibility of the
ries referring to unobservable events and micro-entities (atoms, electrons, last step, the assertion of physicalism. My personal impression, in view
etc.). Then it became possible to formulate a relatively small number of the progress made within the last decades in psj~chology,physiology,
of fundamental laws as postulates from which many empirical laws, both the chemistry of complex organic molecules, and certain parts of physics,
those already known and new ones, could be derived with the help of especially the theory of electronic computers, is that the whole develop
suitably constructed correspondence rules. In psychology analogous de- ment of psychology from the molar phase through the theoretical, the
velopments have begun from two different starting points. The one physiological, and the micro-physiological phases to the final founda-
development began with the introspective approach. It proceeded from tion in micro-physics seems today much more probable and much less
introspectively observed events (feelings, perceptions, images, beliefs, remote in time than it appeared even thirty years ago.
remembrances, etc.) to unconscious, i.e., introspectively not observable,
events. These were first conceived as analogous to the observable events, REFERENCES
e.g., unconscious feelings, beliefs, etc; Later also, new kinds of entities 1. Bridgman, P. W . The Logic of hlodern Physics. New York: Macmillan, 1927.
were introduced, e.g., drives, complexes, the id, the ego, and the like; 2. Bridgman, P. W . The Nature of Physical Theory. Princeton: Princeton Univ.
Pr., 1936.
however, the laws involving these entities are so far only stated in a 3. Bridgman, P. W . "Operational Analysis," Philosophy of Science, 5.114-3 1
qualitative form, which limits their explanatory and still more their pre- (1938).
dictive power. The other development began with the molar behavioris- 4. Bridgman, P. W . " 7 l e Nature of Some of Our Physical Concepts," British
journal for the Philosophy of Science, 1 : 257-72 (February 1951 ) ; 2: 25-44
tic approach. It started with a study of observable events of behavior, (May 1951); 2: 142-60 (August 1951). Reprinted as a separate monograph by
and then proceeded to dispositions, tendencies, abilities, potentialities Philosophical Lib., New York, 1952.
5. Carnap, Rudolf. "Testability and hieaning," Philosophy of Science, 3: 420-68
for such events, and further to more abstract entities. Here the stage (1936); 4: 1-40 (1937). Reprinted as monograph by Whitlock's, Inc., New
of the first quantitative laws has been reached. Haven, Connecticut, 1950. Parts reprinted in H. Feigl and M. Brodbeck (eds.),
Readings in the Philosophy of Science. New York: Appleton-CenturyCrofts,
Both these approaches in psychology will probably later converge 1953.
toward theories of the central nervous system formulated in physio- 6. Carnap, Rudolf. Foundations of Logic and Mathematics, Vol. I, No. 3 of the
International Encyclopedia of Unified Science. Chicago: Univ. of Chicago Pr.,
logical terms. In this physiological phase of psychology, which has 1939. A part ("The Interpretation of Physics") is reprinted in H. Feigl and
already begun, a more and more prominent role will be given to quanti- M. Brodbeck (eds.), Readings in the Philosophy of Science, pp. 309-18. New
Y
tative concepts and laws referring to micro-states described in terms of
I ~ S c i e n t i f i Method,"
c Psychological Review,
cells, molecules, atoms, fields, etc. And finally, micro-physiology may 52: 250-59 (1945). Reprinted, with some alterations, in H. Feigl and W. Sellars
be based on micro-physics. This possibility of constructing finally all (eds.), Readings in Philo~ophicalAnalysis, pp. 498-509. New York: Appleton-
Century-Crofts, 1949.
of science, including psychology, on the basis of physics, so that all 8. Feigl, Herbert. "Existential Hypotheses: Realistic Vs. Phenomenalistic Interpre-
theoretical terms are definable by those.of physics and all laws derivable tations." Philosophy of Science, 17: 35-62 (1950).

75
 Rudolf Carnap
9. Feigl, Herbert. "Confirmability and Confirmation," Revue Internationale de B. F. SKINNER
Philosophie, 5: 268-79 (195 1 ) . Reprinted in P. P. Wiener (ed.), Readings in
Philosophy of Science, pp. 522-30. New York: Scribner's, 1953.
10. Feigl, Herbert. "Principles and Problems of Theory Construction in Psychology,"
in W. Dennis (ed.), Current Trends in Psychological Theory, pp. 179-213.
Pittsburgh: Univ. of Pittsburgh Pr., 1951.
11. Feigl, Herbert. "Physicalism, Unity of Science, and the Foundations of Psy-
chologv," in P. A. Schilpp (ed.),
Critique of Psychoanalytic Concepts and Theories
A - . . The Philosophy of Rudolf Carnap. New York:
~udo;"(forthcomin~).
12. Feigl, H., and M. Brodbeck (eds.). Readings in the Philosophy of Science. New
York: Appleton-Century-Crofts, 1953.
13. Feigl, H., and W. Sellars (eds.). Readings in Philosophical Analysis. New York:
~~pleton-century-crofts, 1949.
14. Hempel, C. G. "Problems and Changes in the Empiricist Criterion of Meaning," FREUD'Sgreat contribution to Western thought has been described as
Revue Internationale de Philosophie, 4: 41-63 (1950). Reprinted in L. Linsky
(ed.), Semantics and the Philosophy of Language, pp. 163-85. Urbana: Univ. the application of the principle of cause and effect to human behavior.
of Illinois PI., 1952. Freud demonstrated that many features of behavior hitherto unex-
15. Hempel, C. G. "The Concept of Cognitive Significance: A Reconsideration,"
Proceedings of the American Academy of Arts and Sciences, 80: 61-77 ( 1951 ) . plained-and often dismissed as hopelessly complex or obscure-could
16. Hempel, C. G. Fundamentals of Concept Formation in the Empirical Sciences, be shown to be the product of circumstances in the history of the indi-
. -
Vol. 11. No. 7 of the International Encyclopedia of Unified Science. Chicago:
vidual. Many of the causal relationships he so convincingly demon-
Univ. of Chicago PI., 1952.
17. Hempel, C. G. "A Logical Appraisal of Operationism," Scientific Monthly, 79: strated had been wholly unsuspected-unsuspected, in particular, by the
21 5-20 (1954). very individuals whose behavior they controlled. Freud greatly reduced
18. Hempel, C. G. "Implications of Camap's Work for the Philosophy of Science,"
in P. A. Schilpp (ed.), The Philosophy of Rudol: Camap. New York: Tudor the sphere of accident and caprice in our considerations of human con-
(forthcoming) - duct. His achievement in this respect appears all the more impressive
MacCorquodale, Kenneth, and P. E. Meehl. "On a Distinction Between Hypo-
thetical Constructs and Intervening Variables," PsychoIogicaI Review, 55: 95-107 when we recall that he was never able to appeal to the quantitative
(1948). Reprinted in H. Feigl and M. Brodbeck (eds.), Readings in the Phi- proofs characteristic of other sciences. H e carried the day with sheer
losophy of Science, pp. 596-611. New York: Appleton-Century-Crofts, 1953.
Mareenau, Henrv. The Nature of Physical Reality. New York: McGraw-Hill, persuasion-with the massing of instances and the delineation of sur-
1956. ' prising parallels and analogies among seemingly diverse materials.
21. Schilpp, P. A. (ed.). The Philosophy of Rudolf Camap. New York: Tudor
(forthcoming). This was not, however, Freud's own view of the matter. At the age
22. Schlick, hloritz. "Meaning and Verification," Philosophical Review, 45:339-69 of seventy he summed up his achievement in this way: "My life has
(1936). Reprinted in H. Feigl and W. Sellars (eds.), Readings in Philosophical been aimed at one goal only: to infer or guess how the mental apparatus
Analysis, pp. 146-74. New York: Appleton-Century-Crofts, 1949.
is constructed and what forces interplay and counteract in it." (2) It
is difficult to describe the mental apparatus
-- he refers to in noncontro-
versial terms, partly because Freud's conception changed from time to
time and partly because its very nature encouraged misinterpretation
and misunderstanding. But it is perhaps not too wide of the mark to
indicate its principal features as follows: Freud conceived of some realm
of the mind, not necessarily havina physical extent, but nevertheless
capable of topogmphic description and of subdivision into regions of
the conscious, co-conscious, and unconscious. Within this space, various
mental events-ideas, wishes, memories, emotions, instinctive tenden-
NOTE: This paper appeared, in somewhat different form, in The Scientific Monthly,
November 1954, and is reprinted by permission of the editor and the author.