This dissertation has been microfilmed exactly as received 70-6878
SHANAB, Robert E lias Abu, 1939- LOGICAL POSITIVISM, OPERATIONALISM, AND BEHAVIORISM.
The Ohio State University, Ph.D., 1969 Philosophy
University Microfilms, Inc., Ann Arbor, Michigan LOGICAL POSITIVISM, OPERATIONALISM,
AND BEHAVIORISM
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University
By
Robert Elias Abu Shanab, B.A., A.M.
********
The Ohio State University 1969
Approved by
/ Adviser Department of Philosophy Dedicated to
Professor Virgil Hinshaw, ACKNOWLEDGMENTS
I am especially indebted to my adviser, Professor
Virgil Hinshaw, Jr. Several of his suggestions have been incorporated in the final manuscript. I wish also to express my thanks to Professor Charles F. Kielkopf.
Finally I wish to extend affection and gratitude to my wife for encouragement, patience and for the hours spent typing and retyping manuscripts.
ii VITA
September 29, 1939 B o m - Jerusalem, Palestine
1962 ...... B.A. , San Jose State College, San Jose, California
1964 ...... M.A., San Jose State College, San Jose, California
1965-1966 ...... Instructor, College of San Mateo, San Mateo, California
1967-1968 ...... Teaching Assistant, Department of Philosophy, The Ohio State University, Columbus, Ohio
1969 ...... Lecturer, The Ohio State University, Newark, Ohio
iii CONTENTS.
Page
ACKNOWLEDGMENTS ...... ii
V I T A ...... iii
Chapter I. INTRODUCTION ...... 1
II. THE VERIFIABILITY CRITERION ...... 7
III. THE TRAN SLA TABILITY CRITERION...... 45
IV. OPERATIONALISM ...... 92
V. WATSON AND LOGICAL POSITIVISM...... 133
VI. TOLMAN, LOGICAL POSITIVISM, AND OPERATIONALISM ...... 161
VII. CONCLUSION ...... 219
BIBLIOGRAPHY ...... 230
Iv CHAPTER I
INTRODUCTION
The primary aim of this study is to exhibit the influence of both logical positivism and operationalism on neo-behaviorism. Specifically, I shall attempt to show how logical positivists and Bridgman influenced the neo- behaviorist, E. C. Tolman. My secondary aims include critical examinations of the logical positivists' quest for an adequate meaning criterion, of Bridgman's opera tionalism, and of Watson's behaviorism.
Just as logical positivism aimed at severing philos ophy from its ontological heritage, and just as operation alism -- spearheaded by Bridgman -- strove to cleanse phy sics of its last vestiges of metaphysics, so behaviorism
-- as prophetically advocated by Watson — struggled to sunder psychology from its introspective armchair specu lations and to erect a science of man.
The avowed intention of the logical positivists in the early 1920's was to declare their independence from , traditional philosophy which they considered as being bur
dened with too much speculation. What the logical positi
vists wanted to fashion was a philosophy that was in con
formity with the results of the current formal and factual
sciences. They asked: What is the proper business of philosophy? What should a contemporary philosopher do?
They decided that the task of philosophy is the analysis of knowledge, especially of science, and that the chief method of philosophy is the logical analysis of the inter-
subjective language of science.^- Hence at the inception of the Vienna Circle the logical.positivists wanted to for mulate a meaning criterion whereby they could decide whether statements are cognitively meaningful — and there by discern meaningful science and its logical analysis from meaningless metaphysics. In their various attempts
to formulate such a criterion, the logical positivists were led to several difficulties. In Chapters II and III, I shall attempt to trace some of the changes which occurred during the formulations of the meaning criterion.
An early and important statement of the meaning
^•Carnap, "Scientific Empiricism," in D. Runes, Dictionary of Philosophy. New York: Philosophical Library, 1942, p. 285. criterion was Schlick's version in the form of the well- known slogan, "The meaning of a proposition is the method of its verification." Schlick's formulation prompted other philosophers such as Carnap, Hempel, Ayer and Popper to reformulate Schlick's version. Many objections were directed toward these various reformulations. Chapter IX will be devoted to discussing the verifiability criterion which seems to embody two main views: first, complete verifiability; second, partial verifiability. In Chapter
III I shall examine the criterion of translatability from
two main views: first, translatability into an empiricist
language; second, translatability into the ordinary idiom.
Chapter III also proposes and defends a meaning criterion which parallels Marhenke's version of translatability into
the ordinary idiom. The proposed criterion takes account of meaningful statements which are excluded both by the
verifiability criterion and the criterion of translata bility into an empiricist language.
Spearheaded by Bridgman's The Logic of Modern Physics, operationalism became another movement which focused its attention upon getting rid of the last vestiges of meta physical obscurity still residing in some parts of science.
In his book Bridgman set forth a program for judging the 4
admissibility of scientific concepts. Briefly, his view was that for a concept to be scientifically acceptable, one must be able to perform a set of relevant operations which
ensure its applicability. Undoubtedly this general view has influenced schools of thought in physics and other
sciences. However, the main purpose of Chapter IV is not so much concerned with Bridgman*s influence upon other schools of thought as it is (1) to present Bridgman's views on operationalism and (2) to compare and contrast Bridgman's operationalism with logical positivism.
Although in Chapter IV I shall attempt to show the striking resemblance between the methodological views of
Bridgman and the logical positivists, I want to make one point clear. Whereas the logical positivists had pain stakingly devoted a great deal of time to elaborating a precise formulation of a meaning criterion, Bridgman, on the other hand, did not seem to bother about such an enterprise. I shall contend rather that Bridgman's main
4 concern was to "operationize" scientific concepts without actually presenting any elaborate theory of meaning. The various quotations from Bridgman's writings will, I hope, bear me out that: he was interested in presenting a certain technique for the formation of scientific concepts. Behaviorism, as propounded by Watson early in this
century, was a movement directed against the old mentalis-
tic psychology. In Watson’s writings one comes across one of the earliest attempts to rid psychology of its specu
lative vestiges. Chapter V will be devoted to pointing out
some of the salient characteristics of Watsonian behaviorism, and then to contrasting these ideas with logical positivism.
Whereas Watson’s views were attractive to the logical positivists, the latter were interested in the logical commitment underlying Watson's behaviorism. In this con nection I shall discuss the physicalistic views of Carnap and Hempel and try to show how, according to the physi calistic thesis, psychological terms are analyzed.
Neo-behaviorism can also be viewed as an attempt to cleanse psychology of its past and to provide for psy chology a more effective scientific methodology. Neo- behaviorists who followed in the footsteps of Watson were significantly influenced by the logical positivist- operationalist views on concept formation and the integral part it plays in the formulation of laws. As basis for my analysis, I shall concentrate in Chapter VI on Tolman. It is my contention that the methodological views championed by logical positivism and by Bridgman deeply influenced 6
Tolman who was genuinely concerned with (a) finding an
adequate base to anchor securely his purposive behaviorism,
and (b) finding sound ways of introducing intervening
variables in his pursuit of psychological law. In Chapter
VI I shall also attempt to defend Tolman's claim that
behavioristic psychology need not forfeit its scientific
status despite his admission in 1959 that intervening
variables are derived from "intuition, common sense and
phenomenology." I shall contend that the chief reason for
this admission was Tolman’s desire to stress the importance of the context of discovery, about which we know so little, when psychologists are faced with the problem of speci
fying the various functions that relate intervening
variables to the independent and dependent experimental
variables. CHAPTER II
THE VERIFIABILITY CRITERION
Introduction
A brief glance at the recent literature of logical
positivism would baffle a reader who is interested in
finding a formulation for the meaning criterion of logical
positivists. Bafflement would primarily result because of
the various formulations for the meaning criterion which
have been propounded at various times by its advocates.
Indeed, one wonders whether it is even appropriate to ask
"What is the meaning criterion?" in light of the diverse
formulations and highly controversial issues centered
around it. The answer to this question varies. Some might
contend that it is; others, that it is not. But, let us
grant for our purpose that an answer to the question is
appropriate and that the highly debated issues about it
are worth considering. j
Thus, In this chapter, I shall attempt to trace some
of the important changes that occurred during the for mulations of the meaning criterion. Mainly, I shall discuss in this chapter one of their earliest versions,
i.e., the verifiability criterion, which seems to embody
two main views: first, complete verifiability; and second
partial verifiability. Let us first consider complete ver
fiability.
1. Complete verifiability as a meaning criterion.
One of the earliest versions of complete verifi
ability as a meaning criterion was formulated by Moritz
Schlick in the early 1930's. Schlick's first formulation
of the meaning criterion was stated in the following manner:
. . . it is simply impossible to give the meaning of any statement except by describing the fact which must exist if the statement is false. The meaning of a proposition consists, obviously, in this alone, that it expresses a definite state of affairs.
If we apply Schlick's first formulation of the meaning
criterion to the sentence "Caesar crossed the Rubicon,"
one then would need a description of the state of affairs which if obtained would either render the sentence true
or false. In this case, one can offer the sentence
"Caesar travelled from one bajik of .the Rubicon to the ' i • •
^Moritz Schlick, "Positivism and Realism," Erkenntnis Vol. 3, (1932-33). Reprinted in Logical Positivism. A. J. Ayer (ed.), (Illinois: The Free Press, 1960), pp. 86-87 other” as description of the state of affairs that must
have obtained if and only if Caesar crossed the Rubicon.
Marhenke, in his article "The Criterion of Significance”
suggested that Schlick's first formulation of the meaning
criterion amounts to saying that "a sentence is signi
ficant if it is possible to formulate another sentence which is synonymous with the given sentence."^
Schlick added to the first formulation of the meaning
the following:
Accordingly, in order to find the meaning of a proposition, we must transform it by successive definitions until finally only such words occur in it as can no longer be defined, but whose meanings can only be pointed out.
But this first criterion formulated by Schlick is only a necessary condition of significance unless one stipu
lates that the transformed sentence is itself significant.
Schlick does not say whether the transformed sentence must be significant. Hence Russell's meaningless sentence
"Quadruplicity drinks procrastination" can be easily • r
^Paul Marhenke, "The Criterion of Significance," in L. Linsky (ed.)> Semantics and the Philosophy of Language, (Urbana: University of Illinois Press, 1952), p. 150. 3 Schlick, "Positivism and Realism," p. 87. 10
rendered in such a manner that its "meaning can only be
directly pointed out," i.e., one can ostensively define
all the words occurring in it. Yet one hardly wants to
contend that the transformed sentence is ipso facto a
significant sentence.
Schlick’s second formulation of the meaning criterion
is stated as follows:
A sentence is significant if and only if it is possible to specify the cir cumstances under which the sentence is true.4
But this second formulation is equivalent to the first formulation. In order to specify the circumstances which are relevant to the truth of the sentence "Caesar crossed the Rubicon," one needs another sentence which is synon ymous with this sentence. But again, this criterion runs
* into the same problem as the first one. For Schlick does not specify whether the transformed sentence must be significant.
Schlick, however, offers another version of this * second criterion by saying
A sentence is significant if and only if
^Marhenke, "The Criterion of Significance," p. 151. This is Marhenke's translation of the passage from Schlick’s Erkenntnis. Vol. 3, p. 7. 11
it is verifiable, and it is verifiable if it is possible to give a description of the conditions under which it is true as - well as of those under which it is false.
It is interesting to note that this formulation makes
use of the notion of verifiability. And, to Schlick,
to say that a sentence is verifiable is simply to state
the conceivable conditions which if obtained will render
the sentence either true or false.
This notion of verifiability has led Schlick to
formulate another version of the meaning criterion whereby he again makes use of the word "verifiability."
Yet Schlick gives us a different view of what he meant
by "verifiability." Schlick’s third formulation of the meaning criterion runs as follows:
A sentence is significant if and only if it is verifiable, and to say that it is verifiable is to say that it is logically possible to verify it . ... a fact or a pro cess is logically possible if it can be described, i.e., if the sentence which is supposed to describe it obeys the rules of the grammar we have stipulated in our language.
^Marhenke, Ibid., this is Marhenke*s translation of the passage from Schlick's Gesammelte Aufsaetze 1932-33, p. 340. 6 Schlick, Gesammelte Aufsaetze. p. 348. The above passage was Marhenke*s translation. See Marhenke*s "The Criterion of Significance," p. 151. 12
In a later article (1936), Schlick offers us another
version of the third formulation of the meaning criterion which goes as follows:
Stating the meaning of a sentence amounts to stating the rules according to which the sentence is to be used, and this is the same as stating the way in which it can be verified (or falsified). The meaning of a proposition is the method of its veri fication. ?
The assumption behind this contention was not that
the meaning of a proposition is its verification --as g C, I. Lewis in his article "Experience and Meaning"
suggested. Lewis was doubtful about what "method of
verification" actually means. He thought that unless a further analysis of the phrase "method of verification"
is given, one would be tempted to interpret Schlick's
criterion as indeed implying that the meaning of a proposition is its verification. And, as a result of
this, one would be easily misled into thinking that a proposition could not have sense or meaning unless it
• * was verified. One could extend this point further by adding that no false proposition could have meaning if
^Schlick, "Meaning and Verification," (1936), in Feigl and Sellars, (eds.), p. 148.
®C. I. Lewis "Experience and Meaning," (1936), in Feigl and Sellars (eds.), pp. 128-145. 1 3 by "method of verification" one means a process of finding out or coming to know that a proposition is true.
Hence, this interpretation would indeed lead to the absurd conclusion that only true propositions are meaningful, and only when they are verified. But surely false propositions are meaningful, although they are never verified.
Actually, Schlick did not want his meaning criterion to be interpreted in the aforementioned way. When Schlick asserted that "the meaning of the proposition is the method of its verification," he did not mean that a proposition is not meaningful unless it is verified. On the contrary, what he meant was, as he puts it:
It is obvious that verifiability is used here in the sense of "verifiable in principle," for the meaning of a proposition is, of course, independent of whether the conditions under which we find ourselves at a specified time g allow or prevent the actual verification.
Hence Schlick, together with most members of the
Vienna Circle, considered, at the beginning, a definition of meaning in terms of "logical possibility of verifi cation." This does not have to be a physical or technical possibility. One finds many meaningful propositions that
^Schlick, "Positivism and Realism," p. 88. 1 4 one is not in any position to verify. The classical
example that Schlick and some other members of the Vienna
Circle cited was: There are mountains on the other side
of the moon. This example was a case in point; for in the
1920's, it was technically impossible to observe the other
side of the moon; yet, Schlick's contention was that pro
positions about the other side of the moon are meaningful,
because in principle one can verify them.^ One could
describe certain observations which if obtained would count
for or against the truth of this proposition. Thus, as long
as we can intelligibly give an account of what observations would establish the truth or falsity of propositions, we
have satisfied the requirements of complete verifiability.
Many criticisms were leveled at Schlickfs third formulation
of the meaning criterion. We shall consider a represent
ative sample of some of the difficulties.
As mentioned earlier, Schlick together with some other members of the Vienna Circle emphasized that the method
of verification is to be understood as "logical possi
bility of verification." Schlick wanted to say that a
sentence is meaningful if and only if it is verifiable and
^Schlick, "Meaning and Verification," p. 158. 15 to say it is verifiable means that it is logically possible to verify it. Now, at times, Schlick seems to be con fusing between "P is logically possible" and "it is logically possible to verify P."*’*’ Schlick's example 12 "rivers flow uphill" is a case in point; This sentence is false, though meaningful. Schlick contended in "that it is logically possible for rivers to flow uphill."
But the logical possibility of rivers flowing uphill is not the same as the logical possibility of verifying or testing that rivers flow uphill. And Schlick's version asserts the latter. If one had simply said that the sentence that rivers flow uphill is meaningful, because it is logically possible for rivers to flow uphill, then one would be using lack of contradiction as a criterion of meaningfulness.
Perhaps, lack of self-contradiction is a more useful cri terion than verifiability,, but that is a different story and the two should not be confused.
Another difficulty which arises as a result of this confusion is that before someone knows whether it is
^ Se e Marhenke*s paper, "The Criterion of Signi ficance," pp. 150-152. Also, see Professor Hinshaw1s paper "Levels of Analysis," Philosophy and Phenomeno logical Research. Vol. XI, No. 2, (1950), p. 217.
^Schlick, "Meaning and Verification," p. 157. 16 logically possible to verify p, one must know whether p describes a conceivable state-of-affairs. In other words, one has to know what a sentence means before knowing what observations would verify it; if not, how does one know what to verify, or what possible obser vations might verify p. As Marhenke puts it:
Schlick’s test comes to nothing, because in order to apply it, we must know in advance that the sentence S is signifi cant.*^
One is actually led to this conclusion if one con siders meaning as a function of verifiability. There are many sentences that one understands without having to resort to the possibility of verification. Consider the following sentence riThe earth will continue to V exist even after all living things cease to exist." W e do know what this sentence means despite the fact that no human being will be there to verify it. There does not seem to be any difficulty concerning what the sentence means. The problem for the verifiability criterion here is the lack of verifiers, i.e., sentient beings to establish the truth or falsity of this sentence. One grants what conceivable state-of-affairs would render the
*^Marhenke, "The Criterion of Significance," p. 152. 17 sentence true or false, but as pointed out earlier, that is a different criterion; it simply is not verifiability.
Aside from Schlick, Carnap in his earlier writings advocated a complete verifiability criterion. In his article "The Elimination of Metaphysics,"^ Carnap offered the following analysis for considering whether a word has meaning or not. Let "a" be any word, and "S(a)" the basic sentence in which it occurs. Then, according to Carnap, the sufficient and necessary condition for the meaning fulness of "a" is given by each of the following, ulti mately equivalent, formulations:
1. The empirical criteria for "a" are known. In other words, the empirical application of "a" must be known.
2. It has been stipulated from what observation sentences "S(a)" is deducible.
3. The truth-conditions for "S(a)" are fixed. That is to say, the conditions under which "S(a)" would be true or false are determined.
4. The method of verification of "S(a)M is known.
The above account seems to indicate that Carnap was
^R. Carnap, "The Elimination of Metaphysics," in Logical Positivism. A. J. Ayer (ed.), pp. 64-65. 18 advocating a meaning criterion in terms of complete veri fiability. Although Carnap’s aim in his article was obviously to demonstrate that metaphysics is meaningless, this enterprise could not be carried out without first formulating a meaning criterion. This, as I just said, turned out to be the complete verifiability criterion.
Hempel, in his well-known article, "Problems and
Changes in the Empiricist Criterion,generalized on these earlier versions of Schlick and Carnap in the following manner:
A sentence has empirical meaning if and only if it is not analytic and follows logically from some finite and logically consistent class of observation sentences. ^
The point behind this formulation was that any non- analytic sentence which follows logically from a finite set of observation sentences will be true if the obser vation sentences are true. Such sentences are said to be conclusively verifiable. This criterion specifies what kind of sentences are capable of being verified
^■^C. Hempel, "Problems and Changes in the Empiricist Criterion," in L. Linsky (ed.), Semantics and the Philosophy of Language, (Urbana: Univ. of Illinois Press, 1952), pp. 163-185.
■^6Ibid. 9 p . 167. 19
independently of any actual verification.
A. Other objections, raised against complete verifi
ability.
Hempel traced the various difficulties that arose
as a consequence of adopting the criterion of complete
verifiability. We shall consider a representative sample of some of the problems.
(a) One difficulty with this criterion arose on account of its "completeness" claim. The point is simply
this: how much observational data do we require for
scientific laws in order for them to be completely
verified? Regardless of how much observational data be assumed, the data always fall short of guaranteeing
the universal character of scientific laws. Consider
the following general statement:
(1) All metals expand when heated,
This universal statement ranges over an unlimited number of instances; it is not logically equivalent to a con
junction of any finite number of observational state ments. Consequently, the criterion in requiring conclu
sive, complete evidence is itself incomplete, because
that requirement excludes genuine scientific statements,
e.g., laws, from the realm of empirically significant 20
(b) It rules out all sentences whose formulations
contain the universal and existential quantifiers such as:
(2) For any electric current there exists some vol
tage source.
Sentences of type (2) are not completely verified by a
finite number of observation sentences because they
contain the universal operator.
(c) It is too inclusive in failing to eliminate non-sensical sentences formed by alternation of non
sensical sentences with significant sentences. For
example:
(3) My arms are bruised or Satan is crippled
is logically entailed by
(4) My arms are bruised.
Thus (3) leads to the untenable consequence that any
sentences can by alternation be added to (4) and, hence,
in principle all sentences become empirically significant.
Thus, this criterion fails to accomplish its purpose, namely, to distinguish those sentences which are empiri
cally significant from those which are not. In this sense,
the criterion is too inclusive. This difficulty coupled with the one mentioned in (a) leads to the untenable 21 position that, on one hand, the criterion is not inclusive enough and, on the other, is too inclusive.
(d) It leads to the strange consequence that denial of existential sentences are not verifiable; hence, they would never be meaningful* A completely verifiable exis tential sentence of the form "(Ex) P(x)," when denied, yields the universal sentence "(x)-P(x)Now consider an observation sentence which yields the information that something is red. In symbols, the information becomes:
(5) (Ex) Red(x)
The denial of (5) is a sentence to the effect that nothing is red. It is symbolized as:
(6) (x)-Red(x)
But (6) has the universal operator and thus is not com pletely verifiable. But this is, indeed, a strange conse quence. Adherence to this account either would lead us to rejecting the fundamental logical principle that if
S is true, then its negation is false, and if S is false, its negation is true, or we would have to resort to an intuitionistic interpretation of logic whereby "-(x)-(Px)" is not logically equivalent to "(Ex) P(x)." 22
B. Complete falsifiability as a meaning criterion 1 7 Popper, being aware of the difficulties that the complete verifiability ran into, proposed a criterion of meaningfulness in terms of complete falsifiability rather than complete verifiability. This view amounts to saying that S is meaningful only in case S is com pletely falsified. As Hempel puts it:
A sentence has empirical meaning if and only if its negation is not analytic and follows logically from some finite logic ally consistent class of observation sentences.18
But, if this criterion is formulated in this way some of the difficulties that occurred as a result of the complete verifiability criterion would recur. We shall v consider some of these problems.
(a) It rules out existential sentences since in seeking to falsify them one finds that their denials yield universal sentences which are not conclusively verified by
17 'Popper does not employ falsifiability as a meaning criterion, but as a criterion of demarcation between science and metaphysics. See, Popper’s book, The Logic of Scien tific Discovery. (London: Hutchinson & Co., Pub., Ltd., 1959), p. 40. See also, Logik der Forschung, (Wien, 1935), Sections 1-7, and 19-24.
■*-®Hempel, "Problems and Changes in the Empiricist Criterion of Meaning," pp. 169-170. 23 a finite set of observation sentences. But this is strange, for we ordinarily consider some existential sentences to be true and others to be false. For example,
"There is at least one university in the world" and "There are griffins" are true and false, respectively. Hence, they are significant, yet according to the falsifiability requirement, they are not meaningful because their denials are not completely verified.
(b) Again, this criterion rules out all sentences whose formulations require mixed quantifiers as lacking significance, since they are not conclusively falsified by a finite set of observation sentences.
(c) If S is meaningful by the present criterion, and
% N is any meaningless sentence such as "Satan is crippled," then their conjunction (S*N) must be meaningful. If the denial of S is deducible from a finite set of observation sentences, then denial of S*N must also be entailed by the same finite observation sentences since - S logically implies -S v-N which in turn is equivalent to the denial of S*N. As a result of this, one must consider a sentence like "My arms are bruised and Satan is crippled" signi ficant, in light of the fact that the criterion allows it to be so. (d) This criterion is also open to the special objection that a universal sentence "(x) F (x)M is mean ingful only if it is conclusively falsifiable. And, in order to falsify it, we must assert an existential sentence of the form "(Ex)-F(x)". But, this negative existential sentence is meaningful only if it is conclusively falsi- fiable. But to falsify it, one must be able to assert significantly the original universal sentence by virtue of the equivalence of "-Ex-F(x)" and "(x) F(x)11. In order to escape this circle, it seems that one must come up with a different and independent criterion of significance for either universal sentences or existential sentences.
C. Evaluations of objections
Hempel, in "Postscript (1964) on Cognitive
Significance,"^ had some second thoughts concerning the objections that he marshaled against complete verifiability as well as complete falsifiability as criteria of empirical significance. Let us consider some of these revisions.
(a) One objection marshaled earlier by Hempel against the verifiability criterion was that, if S is empirically significant, so is S v N, even when N is, as judged by the
1 9 C. Hempel, Aspects of Scientific Explanation and Other Essays in Philosophy of Science. (New York: The Free Press, 1964), pp. 120-122. 25
criterion, sheer nonsense. Since S is entailed by a
finite set of observation sentences, then whatever
completely verifies S, will also verify S v N. But, if
the complete verifiability criterion of meaning is used to
characterize all sentences which are capable of being
either true or false -- in addition to those which are
either logically true or logically false — then the
arbitrary sentence N used by Hempel to support his objection cannot be utilized if N is not capable of being
either true or false. Since N, as used in the objection,
cannot be significantly said to be either true or false,
the logical inference from S to S v N is not admissible.
Neither is the inference from S to SaN acceptable as a
% valid inference.^®
It seems to me that Hempel's analysis in his
Postscript is questionable. There is some disagreement among contemporary philosophers as to what exactly can be used in place of the variables "p", "q" etc., In actual arguments. Should one only be restricted in substituting for variables those sentences that met the logical
^ Ibid. . p. 120. Also, see Rynin's article, "Vindication of L*G*C*L p*s*T*V*SM*," Proceedings and Addresses of the American Philosophical Assoc. . (30, 1957), pp. 57-58. 26
positivists' requirements? Such a proposal seems too
narrow. Recently, for example, Castaneda^ has attempted
to demonstrate how imperative inferences may be valid.
G. H. von Wright has also tried to show the logical
relations among judgments of obligations.^ Hare's book
The Language of Morals. ^ is a prime example of how
command-sentences can play a part in moral deductions.
Certainly there are problems connected with these enter
prises, but that is no reason why one should not make use
of them. It seems to me that if validity in deduction is
dependent upon form rather than content, then many kinds of sentences can replace the variables. If this view is
correct, I thus see no force in Hempel's reconsidering
\ these objections in the Postscript. I think that his original objection, i.e., if S Is meaningful, so Is
S v N, regardless of the status of N, still holds as a
strong criticism of the complete verifiability requirement.
21 H. N. Castaneda, "Imperative Reasonings," Philosophy and Phenomenological Research. Vol. 21, (1960), pp. 21-49.
^ G . h . Von Wright, "Deontic Logic," Mind. Vol. 60, (1951), pp. 1-15.
^ R . Hare, The Language of Morals. (Oxford: The Clarendon Press, (1952), Chapter 3. 2 7 (b) An objection that retains part of its force in
the eyes of Hempel is the one against the complete falsi
fiability requirement. If S is meaningful, and N is any meaningless sentence, then S*N must be meaningful. Hempel
in the Postscript thinks that this objection holds only against the use of falsifiability, not as a criterion of significance, but as a criterion of demarcation.2^
This is, as a matter of fact, what Popper intended the use of falsifiability to be, namely, a criterion of demarcation.
As Popper tells us:
Note that 1 suggest falsifiability as a criterion of demarcation, but not of mean ing. . . Falsifiability separates two kinds of perfectly meaningful statements: the falsifiable and the non-falsifiable. It draws a line inside meaningful language, not around It.25
Yet Hempel thinks that his objection retains its
force if one considers the use of falsifiability as a
criterion of demarcation, but not as a criterion of signi
ficance. Thus, S*N would qualify as a scientific
sentence if S does, even when N Is a meaningless sentence.
24Hempel, Aispects of Scientific Explanation and Other Essays in the Philosophy of Science, p. 121.
2 ^Popper, The Logic of Scientific Discovery, p. 40. f. *3. 28
But I think Hempel's objection would equally apply to the use of falsifiability as a criterion of significance and as a criterion of demarcation, for the same reason that X suggested in (b).
(c) Another objection that Hempel reconsidered was whether the requirements of verifiability as well as falsi fiability rule out all sentences with mixed quantifiers.
He thinks, now, that this does not hold true of all hypo theses with mixed quantifiers.^ Consider the following hypothesis "All crows are black and something is white."
In symbolic notation, it is rendered as:
(1) (x) (Cx» Bx) • (Ey) Wy.
(1 ) is equivalent to:
(2) (x) (Ey) ( (CxaBx) • Wy)
From (1), we can deduce:
(3) (x) (CxoBx)
(3) satisfies the falsifiability requirement, since it is falsified by a finite set of observation sentences. In this case a non-black crow would be sufficient to falsify
26Hempel, Aspects of Scientific Explanation and Other Essays in the Philosophy of Science, p. 121. 29
(3). Also, the sentence
(4) (Ex) (y) (Cx v Wy) is verifiable, for it is entailed by "Ca." However,
Hempel still thinks that many scientific hypotheses of mixed quantifier form are neither completely falsified nor completely verified. The form of these scientific hypotheses is "(x) (Ey) (...... x ...... y ...... )." And for one to verify a hypothesis of this form, one must substitute an instance (b) for (x) in such a way that
,!(y)-(..... b ...... y)" is completely verified. But, of course, such a universal is never completely verified in the strict sense of the word. On the other hand, a hypo thesis of the following type "(x) (Ey) ( (x) v Q(x) )" is verified by an observation report "Q(a) 11 and a hypo thesis of the type "(x)(Ey) (P(x) • Q(y) ) is falsified by an observation report "-P(a) . " 2 7
Thus the failure of all these attempts to character ize meaningfulness or significance in terms of complete verifiability has led some logical positivists to preserve some parts of the meaning criterion by switching from
"complete" to "partial" verifiability -- or as some called it "confirmability."
27Ibid.. p. 122. 30
2. Partial verifiability:
This modified formulation requires of any significant sentence that it should be somehow related to a certain set of observation sentences in such a manner that obser vation sentences provide not complete verifiability, but sufficient observable data in order to confirm or discon- firm the original sentence* That is to say, the original sentence (the one that is said to be significant) should entail observation sentences in such a way that the truth of the latter guarantees the confirmation of the former.
In this case, only a finite set of observation sentences are needed in order to confirm the original sentence.
Carnap, in an important article "Testability and
Meaning,"^® distinguished the testing of a sentence from its confirmation. According to Carnap, "a sentence Is testable if we know such a method of testing for it; and we call it confirmable If we know under what conditions the 29 sentence would be confirmed." Thus, a sentence may be
Carnap, "Testability and Meaning," Philosophy of Science* 3 (1936), pp. 417-71, and 4 (1937), pp. 1-40. Reprinted in part in Feigl and Brodbeck (eds.), Readings in the Philosophy of Science. (New York: Appleton- Century-Crofts, (1953), pp. 47-92.
Ibid.* p. 47. 31 confirmable without being testable as, for example, when we know that the observations of certain events would confirm the sentence, but we are not able to carry out the required experiments for such needed observation.
After a careful analysis of the concept of confir mation, Carnap considered four different criteria which he thought to be essestial to the principle of empiricism
These four criteria were:
(a) Requirement of complete testability
(b) Requirement of complete confirmability
(c) Requirement of testability
(d) Requirement of confirmability.^®
According to Carnap, all these criteria exclude metaphysics; yet he thinks that the last-mentioned criter ion, (d), is the most liberal and allows as significant many empirical sentences of the kind that were excluded by complete verifiability as well as complete falsifi ability. Thus, Carnap thought that (d) suffices as a formulation of the principles of empiricism. As a fact.,
« in recent writings, Carnap explained the principle of empiricism in the following manner:
30Ibid. . pp. 85-86. If it is in principle impossible for any conceivable observational result to be either confirming or disconfirming evidence for a linguistic expression A, the expression A is devoid of cognitive meaning.
Although Carnap seems to be committed to a partial
verifiability criterion, this should not be taken to be his final position on the meaning criterion. Throughout his writings, Carnap continued to present other criteria hoping to find an adequate criterion. Carnap still believes that such a criterion is one of the main tools of philosophical inquiry.
Let us now turn to A. J. Ayer who in the First Edition of his book Language. Truth and Logic presented a "partial verifiability criterion" which he called the "weak" sense of verifiability. He tells us that a sentence is said to be verifiable in the weak sense, "if it is possible for experience to render it probable" or as he puts differ ently "if some possible sense experience would be relevant to the determination of its truth or falsehood."
This amounts to saying that a sentence is verifiable,
31-Schlipp, (ed.), The Philosophy of Carnap. (Illinois: Open Court, 1963), p. 874.
QO ■^A. J. Ayer, Language. Truth and Logic. (New York: Dover Pub., Inc., 2nd Ed., 1946), pp. 36-37. 33 i.e., has empirical significance if that sentence N in conjunction with at least another sentence, if N then 0, logically entails an observation sentence 0 which, if N then 0, does not entail without N. But, as Ayer recog nized in the Second Edition of Language. Truth and Logic, this meaning criterion was too wide, since it permitted empirical significance to any sentence whatsoever.
Suppose, for example, we take N as "Satan is crippled" and the additional premise, if N then 0, as "If Satan is crippled, then my arms are bruised." Then, by modus ponens. we get 0 "My arms are bruised." Hence, from the conjunction of N, and if N then 0, there validly follows a meaningful observation sentence 0 which does not follow from the conditional premise alone. Yet, who in his sane mind, let alone a scientist, would choose to verify "Satan is crippled" in this manner in order to.attribute signi ficance to it.
A. Aver1s conditions
To meet these objections, Ayer introduced a number of conditions. A statement is directly verifiable, according to Ayer, if
(a) it is either itself an observation statement, or is such that in conjunction with one or more observation-statements it entails at 34
least one observation-statement which is' not deducible from these other premises alone.33
On the other hand, Ayer proposes to say that a state ment is indirectly verifiable if it satisfies the follow ing conditions:
(b) First, that in conjunction with certain other premises it entails one or more directly verifiable statements which are not deducible from these other premises. Secondly, that these other premises don't include any statement that is not either analytic or directly verifiable or capable of being independently established as indirectly verifiable. ™
From these two conditions, he finally formulates his criterion as:
And X can now reformulate the principle of verification as requiring of a literally meaningful statement, which is not analytic, that it should be either directly or indir ectly verifiable in the foregoing sense.35
The purpose underlying these conditions is obviously to prevent every N from being verifiable, i.e., empiri cally verifiable. Now, do these conditions accomplish
% what Ayer intended them to accomplish? Let us consider the additional premise, if N then O, to see whether it is ruled out in introducing sentence N.
33 lb id. . p. 13.
34ibid.. p. 13. 35
The additional premise, if N then 0, will be ruled out by principle (a) as an observation sentence, if N is not an observation sentence. Thus, it is not permissible
to conjoin, if N then 0, to N in order to show that N is
directly verifiable. Moreover, in compliance with condi
tion (b), one cannot use, if N then 0 , in order to show
that N is indirectly verifiable, because, if N then 0, is not analytic as in this case. And neither is, if N then
0 , directly verifiable, although it yields 0 when con joined to N. For N is not an observation sentence in all cases. Also, if we assume that,.if N then 0, is indir ectly verifiable, then we have to presuppose that N is * either analytic or capable of being independently estab lished as indirectly verifiable, for which we have no proof. Thus, one notices that Ayer’s conditions appear to have alleviated the objections raised earlier, for here we cannot use, if N then 0, in all cases to show
• r that N is empirically significant. But, be that as it may, many other difficulties arose as a result of the conditions stipulated by A. J. Ayer in the Second Edition of his book Language. Truth and Logic.
B. Problems connected with Ayer’s conditions;
Many recent attempts were aimed at either criticizing 36
or defending Ayer's principle of verification even in its
amended version. As a matter of fact, many criticized the
adequacy of these conditions; others have suggested ways
to amend them. Now, let us see what some of these various
attempts accomplished.
First, A. Church,^ in a review of Ayer’s Second
Edition of Language. Truth and Logic, has presented the
following difficulty. Let 0^, O 2 , O 3 be any three obser
vation sentences, whereby no one of which implies either
of the others, and let N be any sentence whatsoever.
Now, Church's criticism is simply that on Ayer's
definition, one can construct a molecular statement from which it follows that either N or not N is always verifi
able. His argument goes as follows. Consider the fol
lowing molecular statement:
(1) ( (-0i*02) v (O3 V-N) )
Now, (1) is directly verifiable, for together with 0^ it
implies O3 , while 0^ alone does not imply O 3 . The second part of Church's argument is the following: if (-0i*02)
v (0 3 *-N) does not alone imply 02 , then N and (-O3/O 2 ) v
(0 3 *-N) imply 02. And in compliance with Ayer's
^A, Church, "Review of A. J. Ayer, Language, Truth and Logic, Second Edition," Journal of Symbolic Logic, Vol. XXV, (1949), pp. 52-53. 3 7 definition of indirectly verifiable, Church concludes that
(2 ) N is indirectly verifiable.
Alternatively, Church holds that if
(3) ("0 ^*0 2 ) v (Og'-N) alone entails O 2
then
(4) -N and O 3 alone entails 0 3 .
From this implication, Church concludes that
(5) -N is directly verifiable
But either it is the case that (1) alone does not
entail O 2 or it does. Hence, either N is indirectly
verifiable or -N is directly verifiable. Thus, one is
led to the undesirable conclusion that every sentence whatsoever or its negation is significant on the assump
tion that 0 ^, C>2 and O 3 do not imply each other.
Second, D. J. O'Connor, 3 7 offered some more damaging
criticisms which in many respects parallel Church's. Yet
O'Connor offered some amendments to Ayer's conditions.
We are told that for a statement to be directly verifiable
it must not only meet the conditions that were stipulated by Ayer, but also that "if it is a molecular statement, it
should not contain any components which are not either
3 7 D. J. O'Connor, "Some Consequences of Professor A. J. Ayer's Verification Principle," Analysis, (1950- 1951), pp. 67-72. 3 8 analytic or themselves observation statements. " 3 8 This added amendment, according to 0 *Connor, was required in order to exclude molecular statements like (6 ) as being directly verifiable for any N in light of the fact that together with 0 ^, it yields 0 ^.
0 *Connor also attempted to show that the conditions stipulated for indirect verifiability would present some problems, because he thinks that it covers direct as well as indirect verifiability. Hence, the definition for indirect verifiability is too wide . 3 9 To illustrate
0*Connor*s argument, consider the following. The sentence
(6 ) this cat is black is directly verifiable since it is an observation sentence.
Also, it could be characterized as indirectly verifiable, since it satisfies Ayer*s condition of indirect verifi ability. Thus (6 ) in conjunction with
(7) If this cat is black, then she is dangerous yields
(8 ) she is dangerous which is directly verifiable. Therefore, (1) is
3 8 Ibid.. p. 70.
3 9 Ibid. 39
indirectly verifiable. As a result of this, O ’Connor
supplemented Ayer’s conditions for indirect verifiability
in the following way: (a) that the statement should not
itself be verifiable and (b) that if it is a molecular
statement, it should not contain any components which are
not either analytic or directly verifiable or capable of
being independently established as indirectly verifi
able.^ However, at the end of his article, O’Connor
still doubted whether the principle avoids difficulties
even with the amendments that he proposed.^
Third, P. Nidditch,^ in "A Defense of Ayer’s
Verifiability Principle against Church's Criticism," has
proposed a point which somewhat parallels O'Connor’s
proposal. Nidditch's proposal was to ensure that a meaningless statement cannot occur as a component of a
40lbid.
41r . Brown and J. Watling, in their article "Amending the Verification Principle," Analysis, (1950-1951), pp. 87-89, suggested that for a statement to be verifiable, either directly or indirectly, it must contain "only com ponents whose deletion leaves a statement which entails verifiable statements not entailed by the original state ment, or does not entail verifiable statements entailed by the original statement."
^P. Nidditch, "A Defense of Ayer's Verifiability Principle Against Church's Criticism," Mind. Vol. LXX, (1961), pp. 88-89. 40 verifiable molecular statement and then logically derive verifiability from the statement in which it occurs. In other words, Nidditch's proposal was to prevent the joining of (1) with N on the assumption that some of its components, namely N, is not analytic, nor directly veri fiable or capable of being independently established as indirectly verifiable.
But, with due respect to Nidditch's proposal, I think that he missed the point. He is willing to grant the soundness of the inference that (1 ) is directly verifi able, for when conjoined with 0^, it yields 0^. But, it seems to me that here he failed to notice that (1 ) is unacceptable, because it is partially metaphysical if -N is. Consequently, v (0 3 *-S)-N cannot be used to show that N is directly verifiable. N is indeed not verifiable.
Another mistake in the Nidditch's formulation was pointed out by I. Scheffler.^ His argument rests on a molecular statement which he labelled (4*) < (S-Oi) v <-02)
His claim was that if 0^ and 0£ do not imply each
4 ^ 1 . Scheffler, The Anatomy of Inquiry. (New York: Alfred A. Knopf, Inc., 1967), p. 154. 41 other, then (4*) is directly verifiable on the assumption that it yields Oi when conjoined with O2 . Obviously, when
S is conjoined to 0^, it yields (4*). Now Scheffler's conclusion was that S is indirectly verifiable provided that 0^ does not imply S. As he puts it: "Every state ment S whatever, unless it is simply a logical consequence of an observation-statement, is indirectly verifiable, even under the present construal of Ayer's criterion."^
Fourth, 1 think that one could go beyond Scheffler's suggestion. That is to say, one can on Nidditch's pro posal show that every statement is verifiable, either directly or indirectly. The point I am trying to make has been recently suggested by Markinson^ in his article
"Nidditch's Defense of Verifiability." The argument goes as follows: Let N be any sentence, and let 0^, O2 be any observation sentences.
Ibid., Still, other philosophers have cri ticized Scheffler's conclusion as too broad. For a dis cussion of this, see Ullian's article, "A Note on Scheffler on Nidditch," Journal of Philosophy. Vol. 62, No. 10, (1965), pp. 274-75.
4^d . Markinson, "Nidditch Defense of Verifi ability," Mind, Vol. 74, No. 294, (1965), pp. 242-250. 42
Then, the molecular statement
(9) N*(0 ^ v -O2 )
is directly verifiable, because when conjoined with O2 ,
it implies the observation sentence 0-^. Again, one can
show that N is indirectly verifiable, because the con
junction of N and the observation sentence 0^ imply
the directly verifiable molecular sentence (9). This
argument works only if we assume that (a) 0 ^ does not
imply N, and (b) 0^ and C>2 are logically independent.
Now, if one is to suggest anything at this stage
of the game in order to save or render Ayer's definition
of direct or indirect verifiability plausible, one must,
it seems to me, provide a three-valued logic for the con
junction of two.sentences. Also, one should provide an
interpretation of implication (entailment) in the three
valued logic. I think the value of a three-valued logic would be to assign some truth-value status to any arbi
trary statement. Yet, X hesitate to pass on this sug
gestion, for I am not sure about the adequacy of a three
valued logic. It is true that some recent attempts have been made to explicate it, but there are still many
unsolved problems. 43 C. Final note on Chapter XI
So far, we have noticed that the meaning criterion of the logical positivists along the lines of complete verifi ability and partial verifiability has not fared well.
Great obstacles stood in the way of finding an adequate formulation of the meaning criterion. Another obstacle that has not been mentioned in this chapter is the problem of verifying the formulation of the criterion itself, even if we find an acceptable criterion. What X am re ferring to is the difficulty which stems from the fact that the meaning criterion, even when adequately formu lated, is itself not verifiable even though it is apparently a synthetic statement. Thus, the criticism alleges that, if judged by its own standards, the meaning criterion itself is meaningless, since it does not meet its own requirements.
From my own standpoint, this criticism is not so damaging. X believe, along with logical positivists, that the criterion of factual meaningfulness is to be construed as a linguistic proposal which itself is neither true nor false. Thus, by considering the meaning criterion as a
4^To my knowledge, A. C. Ewing was among the first philosopher who put forth this criticism in his article "Meaninglessness," Mind, (1937), p. 347. 44 proposal rather than a proposition, it becomes impossible to subject it to its own standard by asking whether it is verifiable. Nevertheless the meaning criterion has not fared too well either as complete verifiability or par tial verifiability. However, some logical positivists still continued to believe that such a criterion is fund amental to the methodology of science. Consequently,
Carnap and Hempel are still attempting to provide us with better formulations. Instead of formulating a meaning criterion in terras of verifiability, they adopted the view of formulating one in terms of an "artificial language.”
This, ultimately led to the view of empirical meaningful ness in terms of translatability into an empiricist language. How this view fared is the task of the following chapter. CHAPTER III
THE TRANSLATABILITY CRITERION
Introduction
If anything has been demonstrated in the last chapter, it was the difficulty of trying to specify empirical significance of sentences by means of certain logical relationships to some observation sentences. The result, as noted, was either too wide or too narrow.
Consequently, some logical positivists shifted their emphasis to constructing an ideal or improved language where all synthetic sentences that belong to this language are empirically significant. This suggestion led to the view of considering the meaning criterion in terms of translatability. Hence, in this chapter, I shall consider the criterion of translatability from two main views: first, translatability into an empiricist language; second, translatability into the ordinary idiom.
45 46
1. Translatability into an empiricist language as a meaning criterion.X
Hempel's version of this view reads as follows: "A sentence has cognitive meaning if and only if it is trans- latable into an empiricist language."
The assumption underlying this criterion is that there should be a well-construeted empiricist language, whereby if a given sentence, S, meets the requirements of this language, then S is empirically significant.
Now the question that naturally arises is: what are the constituents of an empiricist language? Roughly speaking, an empiricist's language would consist of a vocabulary and transformation rules. The vocabulary would consist of:
(a) logical expressions: Some of the important ones are "not," "or," "if-then," "equivalence," "identity,"
"all," and "there is."
(b) observation predicates: Roughly, observation
^Hempel remarks that the translatability view is due to Carnap's famous essay "Testability and Meaning," (esp. part XV). ,
Hempel, "Problems and Changes in the Empiricist Criterion of Significance," p. 173. 47 predicates designate observable characteristics. Terms such as "green," "blue," "black," etc., would qualify as observation predicates.^ The formation rules define
"sentence in the constructed language," and the trans formation rules or rules of logical deduction are simply
those which were introduced by Whitehead and Russell in
Principia Mathematica or those employed in other contemp orary logical systems. Consequently anything that can be said in the language of Principia Mathematica combined with sets of observation predicates and sets of individual constants would constitute the backbone of an empiricist language.
Carnap, in "Testability and Meaning," in Readings in the Philosophy of Science. Feigl and Brodbeck (eds.), p. 63. explained observation predicates in the following way. "A predicate ' P* of a language L is called observable for an organism (e.g. a person) N, if, for suitable arguments, e.g., 'b', N is able under suitable circumstances to come to a decision with the help of few observations about a full sentence, say 'P(b)1, i.e. to a confirmation of either 'P(b) 1 or '-P(b)" of such a high degree that he will either accept or reject ,P(b)1. Perhaps, it is interesting to note that recently J. W. Comman, in his article "Mental terms, Theoretical Terms, and Materialism," Philosophy of Science. Vol. 35, No. 1, (1968), p. 5, pain stakingly tried to analyze observation predicates with the end-result of a position he adopts which parallels Carnap's position. For our purpose, however, I shall not attempt to elaborate on this, for I am simply taking Carnap's explanation of observation predicates as an illustration. 48
While it is true that this translatability version would avoid the shortcomings of the verifiability as well as the falsifiability criteria, still it is too restrictive.
The only sentences that are considered significant are those which contain strictly logical terms or those which contain observational terms or terms explicitly defined by means of observation predicates. But, now, how about scientific terms such as "curved space," "solubility,"
"magnetism," "ion," "gene," etc., which do not refer to observables objects or events? Are we to consider any sen tence which contain these concepts to be doomed to the status of the empirically non-significant? I think not, for no scientists would want to arrive at such an absurd conclusion. Thus the crucial question that faced the logi cal positivists was to indicate the logical relationship between theoretical terms and observational terms. In other words, what is the criterion of significance for theoretical terms? No unanimous agreement is to be found in answering . this question. Several attempts have been suggested.
Some of the important ones were:
(1) The Method of Explicit Definition
(2) The Method of Reduction Sentences
(3) The Method of Postulates. 49
A . The method of explicit definition.
This was one of the earliest attempts provided by logical positivists. This method construed the relation ship which holds between the theoretical terms and the things which they refer to by means of an explicit definition. An explicit definition would have the following form: " Q ( x ) 3 . .. . x . . . ," where 'x' is the only free variable in the definiens.^ This view leads us immediately into problems when dealing with disposition terms. Disposition terms are these which express the disposition of one or more objects to react in a certain way under specified conditions. Consider then the con cept of solubility as an illustration. According to this view, object x is soluble if and only if, at any time t, x is placed in water, x dissolves at that time. Let "S" stand for "soluble in water;" "P" stand for "placed in water;" "D" stand for "dissolves in water." In symbolic notation, this is rendered as
(1) S(x)H(t) (P(x,t)3D(x,t) )
The difficulty which we immediately encounter is the one with material implication. Clearly, on this explicit
^See Carnap's "Testability and Meaning," p. 52. 50
definition, any object that has never been put in water would have to be considered soluble. Carnap's famous
example was of a burnt match that has never been put in water.^ Let "Sb" stand for "match b is soluble in water."
In symbols, this is phrased as
(2) S(b)==(t) (P(b,t) oD(b,t))
Now, due to the truth-functional connective "if-then" which is equivalent to "not p...or...", the falsity of the
antecedent ensures the truth of the implication. Hence, when "P(b,t)fl is false for any value of "t", then
"P(b,t)» (Dx,t)" is true for all.values of "t". Conse
quently, "Sb" is true even if b is not soluble.
The underlying assumption behind this attempt to
define dispositional terms by observation terms was that
since the definiendum and the definiens are mutually sub
stitutable, all dispositional concepts would be in prin
ciple eliminable; for the only condition a nominal defini
tion must fulfill Is that the term it defines be elimin able in all contexts in favor of the definiens. But as we have seen, the attempt fails.
An alternative to this analysis of dispositional
^Carnap, "Testability and Meaning," p. 53. 51
terms is to require that the logical constant 11 if-then"
be construed counterfactually. In this way, definition
of disposition terms would be phrased as "if.„.were...,
then.. .would." Consequently we eliminate the paradox with material implication which we ran into earlier when we said that "if a match is put in water, it dissolves"
by substituting for it "if the match were to possess the
property of being soluble, then it would dissolve."' But
it is quite evident that the acceptance of such a view
depends upon a satisfactory interpretation of the sub
junctive connective "if...were..., then...would..."
Philosophers, for the most part, have rejected it because
the problem of construing this connective is no easier
than that of explicating disposition terms.6 Another
alternative way of introducing disposition terms was
suggested by Carnap.^ This is the method of reduction
sentences.
B. The Method of reduction sentences
According to this method, reduction sentences would
^For a discussion of the problems of counter factuals, see Chisholm1s article "The Contrary-to-Fact-Conditional," Vol. LV, Mind. (1946), pp. 289-307. Also, see Goodman's article "The Problem of Counterfactual Conditions," Vol. XXXXIV, Journal of Philosophy. (1947), pp. 113-128.
7Carnap, "Testability and Meaning,” pp. 53-56. 52
introduce theoretical terms without explicitly defining
them or introducing a subjunctive construction. Let us
see how reduction sentences would avoid the difficulties
of the method of explicit definition. Consider, for
example, how the concept of solubility is to be rendered
via a reduction sentence, "x is soluble" would be intro
duced by the following sentence: "if any object x is
placed in water at any time t, then x is soluble in water,
if and only if x dissolves in water at time t". In sym bolic notation, we have the following:
(3) (x) (t) (P(xJt)3{S(x)HD(x,t»
The predicate "S" in (3) is "conditionally defined"; its meaning is given by virtue of its occurrence in (3).
Thus, whereas the predicates introduced by the method of
explicit definition are assigned meaning by straight
forward definition, concepts according to the method of
reduction sentences are assigned meaning by so-called bilateral reduction sentences.
Now, how does the method of reduction sentences avoid
the earlier difficulty that we encountered in explaining disposition terms? According to the proponents of this view, reduction sentences provide us with observational criteria, for the use of the concept "solubility". That 53 is to say, instead of providing us with an explicit definition, the reduction sentence method states that if one performs such-and-such an experiment, one can then meaningfully assert that this particular object is soluble if and only if one observes such-and-such results. Conse quently, (3) specifies the application of "S" in case "P" holds, but it says nothing as to whether the object does or does not have the property "S" in the event "P" does not hold. While (1), above in Section A, offers an ex plicit definition of "S", (3) specifies the meaning of "S" partially and does not eliminate.the term from all con texts in which it may occur. In other words, (3) speci fies the meaning of "S" only for those objects where "P" holds, and where "P" does not hold, the meaning of "S" is left open. According to the method of reduction sentences, however, there need not be just one criterion for the use of an unobservable concept. The more science advances, the more criteria we find for the application of concepts.
Scientific concepts have, in the words of Hempel, an
"openness of meaning" which enables them to be used as 54
"crystallization points"** for the advance of scientific knowledge. Thus the method of reduction sentences reveals how it is both possible and natural for these concepts to be only partially defined; for when the test-conditions are not met, the meaning of introduced predicates like
"S" (in our example) is left "undetermined." Hence there is "a region of indeterminateness" in the meaning which the method of reduction sentences assign to descriptive predicates. The "region of indeterminateness" is dimin ished by the addition of further observational criteria.
Each additional criterion that scientists discover would be put forth in the form of a reduction sentence. Thus the more our scientific knowledge increases by producing laws cast in the form of reduction sentences, the less the
"region of indeterminateness" will be. Should one succeed in totally diminishing the area of indeterminateness, one could then replace "reduced predicates" by an explicit definition. As Carnap puts it:
A set of reduction pairs is a partial
Q C. Hempel, "Fundamentals of Concepts Formation in Empirical Science," Vol. XI, No. 7 of the International Encyclopedia of Unified Science. (Chicago, University of Chicago Press, 1952), pp. 28-29. 55
determination of meaning only and can therefore not be replaced by a defini tion. Only if we reach, by adding more and more reduction pairs, a stage in which all cases are determined, mav we go over to the form of a definition.^
However, the attempt to replace "reduction sentences" by an explicit definition does not seem plausible until
further scientific knowledge is obtained. And the truth
of such a claim is not established by philosophic argu ments; rather, it is established by the results of
further scientific research. Thus advocates of reduc
tion sentences have argued that their way of introducing
theoretical terms is more in keeping with the actual
growth of science; in other words, reduction sentences
describe accurately what scientists are doing. On the
other hand, even if it were feasible, explicit definition
of theoretical terms (according to the advocates of re
duction sentences) would not accurately mirror what
scientists are doing. The explicit definition view would % lead us to consider the meaning of theoretical terms
settled once and for all. Put in this manner, theoretical
terms that are explicitly defined are "mere abbreviations;"
hence, they are in principle eliminable. The method of
^Carnap, "Testability and Meaning," p. 60-61. 56
reduction sentences, however, only gives us a "partial
definition" of theoretical terms. Hence the meaning of
introduced theoretical terms is not settled once for all;
they are non-eliminable due to their "openness." It is
thus the contention of advocates of reduction sentences
that, since scientific theoretical concepts acquire further meaning as more reduction sentences are formulated, one
cannot even explicitly define, for example, "electric
current" solely in terms of magnetic effects. Rather the
term "electric current" is partially defined by magnetic
effects, thermal effects and chemical effects. On this
view, the concept "electric current" is "open" and allows
room for expanding the meaning, step by step, on the basis
of accumulated scientific knowledge.
There are many criticisms leveled at the method of
reduction sentences; some are decisive, others are not.
(a) One critic of reduction sentences objected by
saying that "to introduce a term by means of reduction
postulates is to introduce it as an ineliminable primi
tive."^ What the critic seems to imply is that since
lOcf. N. Goodman, Fact. Fiction, and Forecast. (Cambridge: Harvard University Press, 1955), p. 60, (note 11). 57 the descriptive predicates introduced by reduction sentences are left explicitly undefined, they are thus devoid of any meaning.^ No rule of interpretation is assigned concerning how these descriptive predicates are to be used. Goodman deems it important to define dispos itional predicates in order to clarify the language of science, and he is completely unconcerned with the growth of scientific concepts. Goodman thinks that the method of reduction sentences is misleading, because it leaves the descriptive predicates undefined and uninterpreted.
On the other hand, descriptive predicates that are defined are not left as primitives and hence one can assign rules of interpretation enabling us to tell how these concepts are to be used.
(b) Other critics of reduction sentences rejected the claim that scientific concepts grow in meaning as science advances. As mentioned earlier, reduction sentences partially define scientific concepts and hence allow for future criteria which might render them more feasible. The contention was that scientific concepts are "open." But
^ S e e F. Wilson*s paper ‘'Definition and Discovery," British Journal of the Philosophy of Science. Vol. 18, (1967), pp. 291-292. 58 the critic argues that this view confuses the psychological problem of the advance in scientific concepts with the 12 logical structure of these scientific concepts. The critic continues.that on this view of the "openness" of scientific concepts, one could never explain them adequately due to their ever-changing status.
(c) Another criticism also comes from Bergmann.^
In referring to Carnap’s example about the burnt match, he asks us why one should worry about a burnt match when he knows from experience that wooden objects are not sol uble? Bergmann deduces "a burnt.match is not soluble" from "no wooden object is soluble" and "this burnt match is not soluble" is to say that "(x) (Wxa-Sx)" is a law
(where "W" stands for "wooden" and "S" is defined as
12cf. G. Bergmann, Philosophy of Science, (Madison, Wisconsin: University of Wisconsin Press, 1957), p. 51. Bergmann suggests that both Carnap and Hempel confuse two kinds of meaning, namely, reference and significance. Concepts have reference if they directly or indirectly denote. Concepts acquire significance in science as they appear in laws with other concepts. No matter how much a given concept may grow in significance, it still must have had a reference since, otherwise, we could make no sense of attributing to it growth in significance.
^G. Bergmann, "Comments on Professor Hempel's ’The Concept of Cognitive Significance,'" Proceedings of the American Academy of Arts and Sciences. (July, 1951), pp. 78-86. 59
indicated in Section I-A) . However, for Bergmann, "S" •
is not to be confused with the English word "soluble."
Furthermore a law, according to Bergmann, is not ade quately analyzed by transcribing it into an ideal scheme by a formal implication such as "f^(x)^xf2 (x)." Such a
transcription does not yield the counterfactual required of a law, namely, "if x were f^, it would be
Bergmann, aware of such a difficulty, stipulates that a
law is adequately analyzed not simply by asserting
"f-^(x)S> xf2 (x) ," but also by asserting that there are
true pairs of statements such as "f^(a)," "f2 (a),"
fl(b) "f^b)"... Bergmann considered such instances
as "non-trivial positive" instances of the formal
implication.
*1 / Hempel, in a personal communication to Bergmann, has pointed out that true statements such as "w(c)
"-f^(c) ," "-f2 (c)" would jointly imply the negation of
"w(x)^ -f3 (x)." In answering Hempelfs objections,
Bergmann stipulated that the conjunctions Hempel presented are "trivial negative" instances. That is to say,
statements such as "w(c)," "-f^(c), "-f^Cc)" are
14Ibld-, p. 83. 60 irrelevant and should not be included in the analysis of "soluble." Hence, conjunctions such as the ones
Hempel presented do not constitute negative instances for rejecting the law "(x) (Wx^-Sx)." On the other hand, positive instances such as "f-^(a) "f2 (a)" constitute positive instances for asserting a law like "(x)
(WxO-Sx)."
Thus Bergmann*s answer to Carnap's example that a burnt match is not soluble consists in showing that
"(x) (Wx d -Sx ) " is a law which is confirmed by conjunc tions as proposed by Hempel. Also, Bergmann's proposal that "S" is not to be transcribed in the ordinary English is important in his analysis that a burnt match is not soluble. As he puts it:
Yet I know as well as the critics of explicit definition that ' fq* is not a synonym of 'soluble1, if Tor no other reason than that it could be truly pre dicated of a certain match which has been burned before it was ever put into water and of which nevertheless we wish to say that it is (was) not soluble.^
Thus, the above discussion indicates how disposition concepts can be explicitly defined rather than intro duced by the method of reduction sentences. Bergmann
15lbid.. p. 81 61
summarizes his position by saying "...reduction chains
serve no useful purpose, ...the principle of acquaintance
is adequate, and ... the problem of counterfactuals which
now causes so much discussion is a pseudoproblem.
(d) One encounters many theoretical concepts that
do not lend themselves to the. method of reduction because
they are not reducible to observable predicates. Science,
especially in its highly developed stages, makes use of
terms that are highly abstract in character, such as mole
cule, positron, ion, etc. These concepts are neither part
of our observational vocabulary nor part of the hierarchy
of empirical concepts. Yet. advanced science is replete
with such terms.' Now the question arises as to how we
shall introduce these terms into the empiricist language.
The suggestion has been made recently that concepts are
assigned meaning via postulates. This takes us to the method of postulates.
^ Ibid. Perhaps one should add that what Bergmann calls "the principle of acquaintance" is to Hempel the "requirement of definability" which is stated in the following manner: "...any term with empirical signifi cance must be explicitly definable by means of obser vation terms." See Hempel, "The Concept of Cognitive Significance: A Reconsideration," Proceedings of the American Academy of Arts and Sciences. (July 1951) , p. 65. 62
C. The method of postulates
Carnap, in his article "The Methodological Character of Theoretical Concepts,"^ extended his views concerning empirical significance. He specifically rejected the views that he held at the time he wrote his essay "Testa bility and Meaning." The view that theoretical terms can be either explicitly defined or reduced to observable pre dicates by bilateral reduction sentences was rejected in favor of a tentative minimum criterion which came to be known as the "postulational view of the meaning of theor etical terms." Roughly speaking, this view goes as follows: A theoretical term M is significant in Language
L if there is a sentence S in Language L using M in such a way that from S and the remainder of the theory (i.e., postulates, rules and other significant terms), it is possible to deduce an observation sentence that could not be deduced without S. This is Carnap's postulational 18 criterion for the significance of the theoretical terms.
Now his criterion for the significance of sentences goes
^Carnap, "The Methodological Character of Theoret ical Concepts," in Minnesota Studies in the Philosophy of Science, Vol. I, Feigl and Scriven (eds.) , pp. 38-76.
■*-®Ibid., pp. 49-50. 63 as follows: Sentence S is a significant expression in
Language L, if it satisfies the following conditions:
first, S satisfies the rule of formation of L; second, every theoretical term in S is significant.*-9
A brief description of how Carnap employs the method of postulates is in order. To explain this view, one needs the following Camapian terminology: Vt, theoret
ical vocabulary; VQ , observation vocabulary; T, set of postulates or theory; C, correspondence rules. C-rules connect the terms of Vt with VQ . Also, the set of postu
lates (T) connect these terms Vt with other theoretical
terms of V t. Thus postulates provide the meaning of some
theoretical terms. The way it is done is as follows:
According to Carnap, a theoretical term M, relative
to the class K, i.e., relative to some Vt, VQ , T, and C,
is significant if it meets the following conditions:
1. A sentence ^ contains M as the only descriptive
term.
2. The descriptive term in belongs to K.
3. The conjunction of S^S^'T'C is not logically false.
19Ibid., p. 60 64
4. SQ is logically implied by the conjunction of SnT Sk*T*C.
5. SQ is not logically implied by S^'T'C.^
Prior, to these conditions, Carnap summarized his view by saying
If SQ can be deduced from the four premises Sm, Sfc, T, and C, while it cannot be deduced from S^, T, and C alone, then the sentence Sm makes a difference for the prediction of an observable event, and therefore,' has observational significance. Since 'M* is the only descriptive term in Sm. 'M' itself has observational significance.2^
Such was Carnap’s formulation of the postulational view of empirical meaning. Such a criterion appears to be more liberal than any other formulation that Carnap has proposed. But still the postulational view constitutes a recognition of the open-character of scientific terms, for these theoretical terms are only partially inter- 22 preted. In the words of Carnap, "the terms of Vt obtain only an indirect and incomplete interpretation," and this
"incompleteness" of theoretical terms is reflected by the
20Ibid., pp. 52-56. .
21Ibid., p. 50.
22Ibid., p. 47. 65
addition of postulates rather than by further reduction
sentences as was proposed in "Testability and Meaning."
Before considering some of the shortcomings of this
view, it is interesting to note that Carnap in his article
"The Methodological Character of Scientific Concepts"
seems to be moving closer to some of his pragmatist's
critics, especially Quine. But he still has not com
pletely joined this pragmatic position, for he maintains
that one can still distinguish between sentences that are
cognitively meaningful and those that are not cognitively meaningful despite the "openness" of certain scientific
concepts. Now, the question is: how adequate is this
criterion?
First, we are told by some c r i t i c s 2 ^ that this new criterion is either "too catholic or it is meaningless and the criterion is inadequately formulated."24 They arrived at this conclusion by considering a meaningless
term "E(pode)" and showed that it is empirically signifi- *
cant if judged by Carnap's postulational view. And if
2^E. Madden and M. Kiteley, "Postulates and Meaning," Philosophy of Science. Vol. 29, No. 1, (January 1962), p p . 66- 78 .
24Ibid.. p. 76, 66
this is admitted, then this criterion will not rule out
any concept as empirically meaningless.
To illustrate their point, Kiteley and Madden offered
the following: the vocabulary for their theory (Vt) con
sists of the ordinary logical constants, variables, and
the following non-logical terms: E, SG, SA, N and M
standing respectively for "Epode" (E), "structure of a
G-particle" (SG), "structure of an A-particle" (SA),
"node" (N), and "mode" (M). The observation vocabulary
(VQ) has just two terms "W" and "D", standing for the two predicates "put in water" and "dissolves in water." "E"
is the theoretical term whose empirical significance is at
stake. The sentence containing "E" is "Eui." "E" and
"Se" are equivalent to Carnap*s "M" and "Sjq." Carnap's
r sentence "S^" is "Mu^." The observation sentence "S0" is
"WupDu."
The set of postulates (T) consists of the following:
1. (x) (EX » ( S G v SAX)) 2. (x) (SAx » N xy 3. (x) (SGx oMx)
The correspondence rule (C-postulate) used was: 4. (x) (WX»(SGXS M X)
Now putting these sentences together as stipulated by
Carnap's criterion of empirical significance, we notice 67 that the conjunction of T, C, and logically implies
SQ. The conjunction of T, C and does not logically imply SQ . Thus "E(pode)" satisfies Carnap’s criterion of empirical significance. Should this be allowed, then the criterion would fail to rule out any concept as empiri cs cally meaningless. J
Second, another critic2^ suggested that a criterion of significance can only be given for a whole interpreted system, rather than for isolated sentences as portrayed by
Carnap's analysis in "The Methodological Character of
Theoretical Terms." Significant:interpreted systems con sist of statements using both theoretical and observa tional terms. We are told that significant interpreted systems extend even to "systems with hardly any bearing on potential empirical findings."2^ Thus, to Hempel, no scientific sentence can be tested by itself, but each is an integral part of a whole and must be tested in terms of its place in a whole.
2^For a more detailed discussion of this, see Ibid., especially pp. 72-76.
^Hempel, Aspects of Scientific Explanation and Other Essays in the Philosophy of Science, pp. 113-118.'
2?Ibid.. p. 117. 68
Hempel, moreover, even avoids dichotomizing inter preted systems into significant and non-significant, for he tends to think that the confirmation and meaningfulness of such systems is a matter of degree, depending upon such features as the formal simplicity of a system, the clarity and precision with which the theories are formulated, the' explanatory and predictive power of systems, and the degree to which these theories are confirmed with respect 28 to available evidence.
D. What criterion do we need to accept theoretical terms?
It seems to me, following Hempel*s suggestion, that theoretical terms are significant if they are a part of accepted and well-established scientific theories. That is to say, these scientific theories should be interpreted axiom systems which are connected and correlated at vari ous points with observational data in varying degrees. At this juncture one might worry about the need for any theoretical terms In a scientific theory. Why, it was asked, cannot observational terms suffice for scientific theory? Thus judged by the "theoretician's dilemma," that part of the theory which is couched in theoretical terms
2 8 Ibid. 69 is unnecessary. One can formulate the theoretician's dilemma as:
(a) If the theoretical terms in a scientific theory serve their purpose, they are dispensable; and if they don't serve their purpose, they are dispensable.
(b) For any given theory, theoretical terms either serve or they don't serve their purpose.
Thus (c) theoretical terras are dispensable.
If this dilemma is sound, it would indeed lead one to ask, why do we need theoretical terms? Is it not possible to dispense with them and use only observation terms?
Hempel^ has offered some rather convincing reasons that this is not possible; theoretical terms, he argues, are indispensable for the fabric of science. Two main reasons support this contention.
First, while it is true that many general sentences have been formulated In terms of observable terms, they still are not adequate because they have a narrow range of application and, even in that range, there are
^See Hempel's articles on "The Theoretician's Dilemma" in Minnesota Studies in the Philosophy of Science. Vol. 111. Reprinted in Hempel*s Aspects of Scientific Explanation, pp. 173-226; "Implications of Carnap's work for the Philosophy of Science," in P. A. Schilpp (ed.), pp. 685-709. 70 exceptions. To use Hempel's example^®
(4) Wood floats on water; iron sinks in it.
(4) is a general sentence which refers to specific kinds of objects, and it concerns their floating behavior on water. But this general sentence has exceptions; certain kinds of wood sink in water and a hollow sphere of iron with specific dimensions floats on water. Such discrep ancies could be rectified by introducing the theoretical term "density of an object" which is defined as the mass per unit of volume. Thus, instead of (4), we will have the following general sentence
(5) Any solid object whose density is less than that of water floats.
With the introduction of the theoretical term (density), we avoided the problems of the narrow range and exceptions connected with the general sentences like (4).
Hempel's second argument rests on the assumption that a scientific theory, among other things, has the task of establishing not just deductive relations but also induc tive relations in the explanation and prediction of empirical findings. Hempel illustrates his point by means
^®Heinpel, "The .Theoretician's Dilemma: A study in the Logic of Theory Construction," Ibid., p. 180. 71 of the following four sentences, where "magnet" is- con sidered as a theoretical term. Consider the hypothesis
(5.1) the parts obtained by breaking a rod-shaped magnet into two are again magnets (5.2) If an object x is a magnet, then whenever a small piece y of iron filing is brought into contact with x, then y clings to x.
In symbolic notation, we have
Mxs (y) (FxyoCxy)
(5.3) Objects b and c were obtained by breaking object a in two, and a was a magnet and rod-shaped. (5.4) If d is a piece of iron filing that is brought into contact with b, then d will cling to b.
Now according to Hempel, if (5.3) is given and
(5.2) is assumed, one is able to deduce sentences like
(5.4) via the hypothesis (5.1). But (5.3) is not an observational sentence, since it uses the sentence "a is a magnet." Also (5.3) is not deducible from other ob servational sentences, for (5.2) specifies only a necessary, but not a sufficient condition for "M" in terms of the observational terms. Thus if the transition from (5.4) to
(5.3) is to be accepted, one must take an "inductive step" leading to (5.3) which is based on a set of obser vation sentences. In this case, "Ma" might be inductively supported by a number of instances of "Fay»Cay" provided that no negative instances, "Fay.-Cay," are reported. 72
These instances tend to inductively support "(y)
(Fay3 Cay)" which by virtue of (5.2) support "Ma."^
^See Hempel, "Implications of Carnap’s work for the Philosophy of Science," p. 700.
For reasons stated in the paragraph below, I have omitted reference to two other proposals for dispensing with theoretical terms which have been labeled the "func-' tional replaceability of theoretical terms" by Hempel ("The Theoretician's Dilemma"). The first proposal, the Craig-theorem method ("On Axiomatizability within a Sys tem," Jour. Symbolic Logic 18: 3-32 (1953), shows in Hempel's words "that for any theory T* using both theoret ical terms and nontheoretical, previously understood ones, there exists,, under certain very widely satisfied condi tions, an axiomatized theoretical system T'g which uses only the nontheoretical terms of T* and yet is function ally equivalent with T* in the sense of effecting, among the sentences expressible in the*non-theoretical vocabu lary, exactly the same deductive connections*as T'" (op. cit., pp. 212-213). The second proposal, the Ramsey- sentence method (The Foundations of Science. N. Y. & London, 1931, pp. 212-215, 231), is in effect, according to Hempel, a way of treating "all theoretical terms as existentially quantified variables, so that all the extra- logical constants that occur in Ramsey's manner of formu lating a theory belong to the observational vocabulary" (pp. 215-216).
Hempel rejects Craig's method (1) because, as Craig himself puts it, the postulates of T'g "fail to simplify or to provide genuine insight" ("Replacement of Auxiliary Expressions," Philos. Review 65: 38-55 (1956) ) and (2) because the method does not lend itself to Inductive use since all the sentences of the system T'g obtained by the method are logical truths and hence T'B makes no empi rical assertions whatsoever (p. 215). Hempel likewise rejects Ramsey's method because a Ramsey-sentence, in which all theoretical terms of T* are existentially quan tified, ", . . still asserts the existence of certain entities of the kind postulated by T', without guaranteeing any more than does T' that those entities are observables 7 3
It is pointless, then, to say that science can dispense with theoretical terms. This is scientific suicide. A cluster of scientific concepts, as indicated earlier, do not lend themselves to observable predicates.
Consequently, if we do not want to dispense with theore tical terras, what is the alternative? The alternative, it seems to me, is to claim that a scientist may use whatever concepts he can use; the only restriction to which he is subject is that his terms belong to well- established scientific theories.
E. Concluding note on translatabilitv into the empiricist language as a meaning criterion.
What, now, remains of the meaning criterion when formulated in the aforementioned fashion? Should one completely abandon it? We mentioned that science is replete with significant terms that are not a part of the restricted empiricist language sketched in section I.
We discussed the several methods that attempted to define or at least characterizable in terms of observables. Hence," he concludes, "Ramsey-sentences provide no sat isfactory way of avoiding theoretical concepts" (p. 216). I thus believe that neither method offers a genuine pos sibility of dispensing with theoretical terms in the empirical sciences. theoretical terms in order to render them accessible
for the empiricist language. Needless to say, many
difficulties stood in the way. We suggested that theo
retical terms acquire empirical meaning if they are a
part of established and accepted scientific theories.
Now, what about the other attempts, i.e., the method of
explicit definition, reduction sentences, and postulates?
Should one abandon them? I think not, because in est
ablished and scientific theories one can make use of the
three methods suggested in this chapter as well as other methods that have been discussed by philosophers of science
in introducing theoretical terms. The unfortunate thing,
however, that one finds is the belief among some philo
sophers of science to make one method supreme and try to
force other cases to conform to it.
From the aforementioned discussion, one might draw
the conclusion that perhaps I am advocating the criterion
of translatability into the empiricist language.
This is a mistake. All that I tried to show was that
a strict adherence to this version would lead us to the
absured consequence of ruling out theoretical terms as
empirically significant, since they are not included in
empiricist language even if one could expand it, in order to include theoretical terms via the suggested methods.
This, I hope, should not imply that such a criterion is adequate, let alone that I advocate it. On the contrary,
I think that this criterion is far from being adequate, even with the proviso that the empiricist language could be extended to cover theoretical terms.
The major difficulty, as I tried to indicate, was that of finding any satisfactory formulation of a cri terion of significance for theoretical terms. I suggested that theoretical terms acquire empirical meaning if they are part of established and accepted scientific theories.
And scientific theories could be evaluated in terms of such characteristics as the formal simplicity of a system, the clarity and precision with which the theories are formulated, the explanatory and predictive power of systems and the degree to which these theories are confirmed with respect to available evidence.32
But be that as it may, what we suggested should not be construed as a defense of the adequacy of the trans latability criterion into an empiricist language. For one
32C . Hempel, Aspects of Scientific Explanation and Other Essays in the Philosophy of Science, p. 117. could contend that while it is true that every sentence
that is translatable into the empiricist language is
empirically significant, yet it is false to say that
those sentences that are not translatable into the empir
icist language are ipso facto meaningless. Thus, construed
in this way, translatability into an empiricist language
is not a necessary condition, but a sufficient condition
0 0 for empirical significance. As far as the significance of the remaining sentences that are not translatable
into the empiricist language, one should resort to other grounds. What other grounds should one consider? An answer to this question will take us to a consideration of the other version of the translatability thesis, i.e., translatability .into the ordinary idiom as a meaning criterion.
2. Translatability into ordinary idiom as a meaning criterion.
o M P. Marhenke, in an article entitled "The Criterion of Significance," wanted to abandon the verifiability
^ C f . 1 . Scheffler, "Prospects of a Modest Empiricism, The Review of Metaphysics. Vol. X, (1957), pp. 383-400, 602-625.
3^P. Marhenke, "The Criterion of Significance," pp. 139-159. 77
criterion as a meaning criterion. In lieu of this, he proposed to use the criterion of translatability into the
"ordinary idiom" as a criterion of significance. The ordinary idiom is characterized by Marhenke rather vaguely as the idiom or discourse which we all use in communica
ting with each other. I gather that what Marhenke actual
ly wants us to do with the meaning criterion is to say
that the "ordinary language" of everyday conversation which contains no technical terras is the criterion of
significance. Also, I think Marhenke would want to claim
that any version of the meaning criterion would be trans
latability into the ordinary idiom which if accomplished,
ensures significant discourse. As he puts it:
Translatability into this idiom is a necessary condition of. significance, because we have but one recourse when we are asked to clarify the meaning of a sentence that is not in this idiom. It would not be to the purpose to answer the question by translating the sentence into another sentence of the same idiom, for its meaning, if it has one, would not thereby become any clearer. The problem can be met only, if at all, by translating the sentence into the ordinary i d i o m . ^5
Still, the question to be adequately answered is:
35ibid., p. 142 78 what exactly is the ordinary idiom? How is one to tell, in doubtful cases, whether a string of words belongs to the ordinary idiom? Consider the following sentences.
(1) Cats drink milk.
(2) Quadruplicity drinks procrastination.
(1) is a sentence in the ordinary idiom, but how about
(2)? One is inclined to consider it meaningless; yet, if (1) is in the ordinary idiom, why is not (2)? It certainly is like (1). From a grammatical point of view, it is well-formed, and it is made up of words none of which is itself nonsensical. One can easily translate
(2) into Arabic, French, or German for that matter. But, it still seems to be meaningless. Now, is it or is it not in the ordinary idiom? If it is not, why not? And if it is and nevertheless still considered meaningless, then what becomes of the translatability into the ordinary idiom as a criterion of significance?
Realizing the difficulties that translatability into the ordinary idiom runs into, Marhenke proposed that this requirement should be replaced by a weaker requirement, i.e., translatability into some other idiom.^ •
36Ibid., pp. 142-143. 79
A. Translatability into some other idiom as a meaning criterion.
This weaker requirement runs as follows: A sentence is significant if it is translatable into some other idiom.
This liberal version seems to allow significance even to metaphysical discourse. Let us examine a passage from the writings of Heidegger to illustrate how this weaker requirement allows it to be significant. Heidegger writes:
Why are we concerned about the nothing? -The nothing is rejected by science and sacrificed as the unreal; science wants to have nothing to do with the nothing. What is the nothing? Does the nothing exist only because the not, i.e., negation, exists? Or do negation and the not exist only because the nothing exists. We main tain: the nothing is the simple negation of the totality of being. Anxiety reveals the nothing. The nothing itself n o t s . 37
Carnaps who at one time analyzed the above quotation came to the conclusion that none of these sentences can be translated into significant sentences of the ordinary idiom. First, the word ’’nothing11 is not used as a name in the ordinary idiom. Sometimes we do use it as a name
37qUoted in Marhenke, ibid.. p. 158.
38Carnap, ’’The Elimination of Metaphysics," in Logical Positivism. A. J. Ayer (ed.), pp. 69-73. 80 as in the following sentence.
(3) nothing is duller than listening to a baseball game. ■
Thus, in (3), the word "nothing" functions as the subject, but we could easily dispense with it by saying
(4) listening to baseball is the dullest thing one could do.
Second, in the last sentence of the quotation, the word
"not" is used as a verb by Heidegger. Regardless of what our ordinary idiom consists of, it certainly does not
include the verb "to not." Hence we are faced with trying to understand what the noun "the nothing" as well as the verb "to not" mean. Presumably they have none and until one translates them into the ordinary idiom, they must be considered meaningless.
Thus, if translatability into the ordinary idiom is to count as a criterion of significance, some sentences which occur in the quotation are not amenable to such a translation. But, suppose, we consider translatability into some other idiom as a criterion of significance.
What, then, are we to say of such sentences as they occur in the quotation: 81
(5) The nothing is rejected by science and sacrificed as the unreal.
(6) Science wants to have nothing to do with the nothing.
Since no restrictions are posited on what the phrase
"other idiom" must include, a metaphysician would then claim that sentences (5) and (6) are meaningful, because they are a part of his idiom. We grant him this provided that if he wants to make himself understood, he must then translate (5) and (6) into the ordinary idiom. Suppose, further, that a metaphysician like Heidegger provides us with the appropriate translation for sentences (5) and
(6) in the following manner:
(7) Feeling is rejected by science and sacrificed as the unreal.
(8) Science wants to have nothing to do with feeling.
Thus, the appropriate rendition for (5) and (6) was simply to replace the word "the nothing" by "feeling."
I am not sure whether Heidegger would agree to this. But the point I want to make is simply that if a criterion of significance is construed as translatability into some other idiom, then metaphysical sentences as well as other sentences which the logical positivists declared as 82 cognitively meaningless, i.e., unintelligible, wouid be saved from this onslaught. In this way, while one would admit that metaphysical assertions are exempt from the criteria of verifiability as well as translatability into an empiricist language, still they belong to the domain of significance provided one adheres to the criterion of translatability into some other idiom. It should be noted, in this connection, that the "other idiom" is in turn translated into the ordinary idiom— should this be called for. A simplified version of this view would be that a sentence is significant provided that it is understood. And to say that one understood it is to say that one has learned how to use it via the idiom that he is familiar with.
Undoubtedly this is a broad criterion. I admit this.
I am sure it makes many empiricist philosophers unhappy.
But the hope on the part of logical positivists that the verifiability requirement would provide us with an adequate criterion of meaning is far from being fully realized. It is, I take it, generally acknowledged that a meaning criterion solely in terms of verification re quirements or, for that matter, in terms of translability into an empiricist language is not sufficient to deal with 83 the multiplicity of meanings actually employed. And, it seems to me, if one has full respect for the concept of meaning, other forms of meaning must be included. The thesis that X am proposing does just that. It does not rule out the verifiability requirement, for we do admit that there is a minimal verifiability requirement, irres pective of the difficulties connected with it, for any scientific hypothesis. Neither does it rule out trans latability into an empiricist language, for most scien tific theories are formulated in terms of a certain speci alized language and one could certainly consider an ex- OQ tended version of the empiricist language*” as extremely important for scientific theories. Consequently a senten ce that is not verifiable nor translatable into an empiri cist language is certainly not important for scientific purposes. But this should not mean that it is unintel ligible, or just a string of meaningless noises. This view, I take to be the logical positivist's cardinal sin.
Recently some philosophers have advocated a
39An extended version of an empiricist language would include the meaningfulness of sentences encompassing theo retical terms. This was discussed earlier. See Section 1.
■ 84
philosophic method in terms of "levels of analysis."^
One philosopher tells us that by "a level of analysis" is
meant ■
a given problem area wherein questions of a certain similar sort are to be dis cussed. They are similar in the sense that they all involve similar presup positions. Issues that are significant at a given level all presuppose a common set of assumptions.^1
The thesis that I am advocating, i.e., translata
bility into some other idiom, fits well with the method
propounded by the author of this quotation. While I do
not restrict the levels of analysis, the tacit assumption
underlying levels of analysis is to attribute significance
to each level. Thus, in terms of the thesis I am advoca
ting, one arrives at the following. In each level of
analysis, the significant problems and issues are the ones
that are translatable into the idiom peculiar to the level under discussion.
Let us consider "the speculative level." According
^®In particular, see G. Bergmann's article "Frequen cies, Probabilities, and Positivism," Philosophy and Phen. Research. Vol. VI, (1945), pp. 26-41. Also, V. HInshaw's article "Levels of Analysis," Phil, and Phen. Research, Vol. XI, No. 2, (1950), pp. 213-220.
^V. Hinshaw, "Levels of Analysis," p. 214. 85 to its author, "this level comprises problems along the lines of traditional metaphysics."42 To illustrate this, allow me to use the sentence "Qualities inhere in a sub stance." We grant that this sentence is not fit for scientific purposes since the minimum requirement for something's being worthy of scientific consideration is for it to be in principle verifiable. But the thesis we are proposing would attribute some meaning or significance to the sentence "Qualities inhere in a substance" since it is translatable into some other idiom. Now since the speculative level specializes in problems concerning meta physical issues, the idiom that fits it is, of course, the one espoused by metaphysicians. Thus it fulfills the requirement of translatability into some idiom. But, if the metaphysician is to make sense to us, we must also require that he provide us with a translation of his own idiom into the ordinary idiom. For example, he might tell us that to say "Qualities inhere in a substance" means that there are certain qualities that are truly predicated of a substance. Still the word "substance" needs expli cation. Suppose we substitute "bearer or owner of certain
42Ibid., p. 214. 86 qualities" for "substance;" then let "apple" stand for
"bearer of certain qualities." Thus to say that certain qualities inhere in an apple is to say that, for example,
"Qualities are truly predicated of their bearers."
What we have above according to Carnap is that
"quality terms are predicable of subject t e r m s . Hence a translation of a metaphysical sentence like "Qualities inhere in a substance" seems to accord well with Carnap's distinction between the formal and material modes. The sentence "Qualities inhere in a substance" belongs to the material mode and "'Quality terms' are predicable of'sub ject terms'" belongs to the formal mode of speech.
The point underlying the distinction between the material mode and formal mode was that characteristic of the former mode is talk about things, objects, facts, while characteristic of the latter mode is talk about language. Hence, put in this manner, what a metaphysician should do according to Carnap is to concern himself with
the formal mode of speech in order to render his discourse
intelligible. Lest it be misunderstood, I am not sug gesting that philosophy is just doing."logical syntax."
^ C f . R. Carnap, Logical Syntax of Language. (New York: The Humanities Press, 1937), Sections, (74-81). 87
The point I am trying to make is that the thesis we sug gested could be utilized by a logical positivist to indi cate that a sentence like "Qualities inhere in a sub stance" takes on a meaning once it is translated into the formal mode of speech.
The question which arises is whether Carnap's pro posal does any justice to the metaphysician's idiom.
Carnap's proposal -- that metaphysical sentences mislead us into thinking that a factual claim is made, when very often a linguistic proposal is being suggested — is entirely unsatisfactory to a metaphysician. For a meta physician is not simply making a linguistic proposal when he uses a sentence such as "Qualities inhere in a sub stance." Rather a metaphysician wants to claim that his sentences refer to objects of some kind. The most that
Carnap has shown through his proposal is that metaphysical sentences should not be viewed as if they refer to objects of some kind, but as being logically equivalent to corres ponding syntactical sentences. Underlying this assumption is Carnap's belief that in order for a sentence to be sig nificant, it must be transformed into a significant sen tence of standard logical form. But why should a meta physician accept Carnap's view that whatever he says must 88 be translatable into the "formal mode" in order to be significant? Clearly a metaphysician would reject such a proposal; he would simply suggest another proposal whereby his own sentences would acquire significance in his own language.
Thus one could then contend that., while it is true that every metaphysical sentence which is translatable into the formal mode would become, according to Carnap, significant, yet it is false to say, as Carnap proposes, that these sentences that are in the material mode are devoid of significance. But this leads us to what I tried to suggest, i.e., in order to decide whether a sentence is significant, one should only be able to translate it into some other idiom which, if called for, must in turn be translated into the ordinary idiom. Such a view deviates from Carnap's in that it does not assume that certain forms of language are particularly privileged, i.e., mean ingful in some unique, fundamental way. For the thesis I suggested does not restrict or posit conditions on what the "other idiom" is. We allow for the "other idiom," if called for, to be translatable into the ordinary idiom.
But Carnap's proposal does just the opposite. He only allows those statements which are significant to belong to an improved language. And the improved language, according to Carnap, remains at the level where it cannot be translated with its original meaning into the ordinary idiom. For the ordinary idiom is the hallmark of the material mode. Had Carnap allowed the "formal mode," if called for, to be translatable into the material mode,
then Carnap's proposal would be in accord with what I suggested.
Thus I think a metaphysician could avail himself of the thesis we suggested to show that his sentences are significant and need not be translatable into the formal mode of speech in order for him to make sense. For the
thesis that I am proposing requires that significant sen tences be translatable into some other idiom. The "other
idiom" must in turn be translated into the ordinary langu age, should this be called for. To look for a criterion of significance along the lines the logical positivists did is a hopeless endeavor. If it is put forth as a stipulative definition, there is no reason to go along with the stipulation; as a general criterion of signifi cance, i.e., generalization about meaningful sentences, it is deficient in light of the many counter-examples that have been given. If it is offered as a proposal, then 90 again, there is no reason to accept it as covering all cases of significant discourse. The thesis we suggested could be construed as a proposal. If so, then the logical positivists might reject our proposal. But, I believe this could only be done at the expense of sacrificing much intelligible discourse.
C. Final note on chapter III.
By way of summary, I discussed in this chapter the criterion of translatability. Two main views were pre sented:
1. Translatability into the empiricist language.
2. Translatability into the ordinary idiom.
In section (1), the central issue was focused upon the introduction of theoretical terms. Three methods were discussed in connection with theoretical terms: (a) ex plicit definition, (b) reduction sentences and (c) postu lates. One other method that we suggested considered that theoretical terms are meaningful if they are a part of accepted and well-established scientific theories.
In section (2), I explained somewhat briefly the adequacy of the translatability into the ordinary idiom.
This criterion was partially accepted, because I tried to suggest that in order, to decide whether a sentence 91 is significant, one should be able to translate it into some other idiom which, if called for, must in turn be translated into the ordinary idiom. This view would do justice not only to scientific and logico-mathematical sentences, as the logical positivists wanted, but also to other sentences which were excluded by the various criteria expounded by the logical positivists..
In the following chapter, I shall present operation- alism, especially as propounded by Bridgman. The emphasis will be placed not just on establishing the relationship between the logical positivists' version of the meaning criterion and Bridgman's operationalism, but also on pointing out that if operationalism is to be of any value, it must include some of the ideas that logical positivists introduced. I shall contend that operationalism should not be considered as providing us with a meaning criterion, but with a goal for scientists to strive towards in their efforts at concept formation. T
CHAPTER IV
OPERATIONALISM
Introduc t ion
The Idea and the word "operationalism" had its origin in physics. Its originator, P. W. Bridgman, elaborated vigorously on an operational point of view in his book
The Logic of Modem Physics. Just as some of Einstein's conclusions in the theory of relativity provided the groundwork for the meaning criterion of logical positi vism, so, with Bridgman, it was the same Einsteinian views on relativity which brought him to consider operational analysis in physics.
Einstein had considered as meaningless such terms as
"absolute length," "absolute simultaneity," "absolute duration" from Newtonian physics. The scientist's In ability to state any observational or experimental pro cedures for the application of these terms led Einstein
4 to the view that they were devoid of empirical meaning.
Bridgman, having embraced this background as a stimulus,
■" 9 2 93 set forth in his book The Logic of Modem Physics a program for the legitimate admissibility of scientific concepts. Briefly his view was that for a concept to be scientifically acceptable, one must be able to perform a set of operations. Undoubtedly such a view has influ enced schools of thought in physics and other sciences.
However, the main purpose of this chapter is not so much concerned with Bridgman’s influence upon other schools of thought as it is (1) to present Bridgman's views on operationalism and (2) to compare and contrast Bridgman's operationalism with logical positivism. Hence, in what follows, I shall consider Bridgman's operationalism from two main views: first, his early and late views on oper ationalism; second, operationalism and logical positivism
1. Bridgman's early views on operationalism.
Bridgman's formulation of operationalism in his book
The Logic of Modern Physics reads as follows:
In general, we mean by any concept nothing more than a set of operations; the concept is synonymous with the corresponding set of operations.*■
. ^W. P. Bridgman, The Logic of M o d e m Physics. (New York: Macmillan Co., 1927), p. 5. 94
A page later, he adds:
It is evident that if we adopt this point of view toward concepts, namely that the proper definition of a concept is not in terms of its properties but in terms of actual operations, we need run no danger of having to revise our attitude toward nature.
It seems at least clear from the above quotations that Bridgman*s early view as depicted in his book The
Logic of Modern Physics was that the meaningfulness of scientific concepts is determined by performing certain physical operations. Thus, in accordance with his early views, Bridgman maintained that the meaningfulness of scientific concepts is a function of unambiguous sets of operations and that, when the operations are different, the concepts are different. As Bridgman puts it:
In principle the operation by which length is measured should be uniquely specified. If we have more than one set of operations we have more than one concept, and strictly there should be a separate name to correspond to each dif ferent set of operations.^
Thus two salient characteristics permeate Bridgman*s earlier views on operational analysis: (a) physical operations determine the meaning of scientific concepts
2Ibid., p. 6.
3Ibid., p. 5; 95
(b) different operations lead to different concepts.
Bridgman illustrates these two characteristics in the fol
lowing manner. In order to find the length of an object, one has to perform a set of certain physical operations; for the concept of length, according to Bridgman, involves nothing*more than the specific set of operations by which length is determined. This conviction prompted Bridgman to say "In general, we mean by a concept nothing more than a set of operations.” Bridgman goes further to point out that the word "length” means something different when it is determined by the use of a measuring rod than when it is determined by optical methods. One could also contend
that there are different concepts of optical length, depending upon the instruments, calculations and theory used. Hence if the operations are different, the con cepts corresponding to them should be different. Nothing but confusion, according to Bridgman, results when some one neglects to take into consideration that the word
"length” means different things in different operational contexts.
The question that arises is, how adequate are Bridgman's early views of operatiohalism? Various criticisms 96 were leveled against these early views. ^ We shall con
sider a representative sample of some of the problems.
(a) It is interesting to inquire how, in accordance with Bridgman*s views, one is to define "the same set of operations?" In other words, If one were to say that two concepts mean the same, the two concepts must then be
determined by "the same set of operations;" for different operations imply different concepts. One could contend
that by "the same set of operations" is meant a unique
set of physical operations which can be specified, such
as using a particular kind of a measuring rod in carrying out the experiment. But what criterion does one use in
finding out what is to count as the same set of operations
should another occasion of measuring arise? Would we want
to count a different but similar measuring rod as a member of the same set of operations? If we agree that
It would, how similar should the measuring rods be in the
event the same operations are to be repeated? Thus It appears that in certain particular cases, "the same set of operations" has to be defined In terms of operational
^I am omitting the obvious "pie-recipe" confusion often attributed to Bridgman. What he obviously meant, but did not say, was that the application of the concept is uniquely specified by a corresponding .set of operations. 97
terms if one is to preserve Bridgman's basic contention
that the application of the concept is uniquely specified by a corresponding set of operations. But this rendition would force one to argue in a circle. Hence "the same
set of operations" must be explained in non-operational'
terms in order to escape the charge of circularity.
(b) L. Russell^ in his review of Bridgman's The
Logic of Modern Physics seems to have put his finger on
the difficulty just mentioned in (a). L. Russell suggested
that in science one often makes mention of certain oper ations as being better than others and that one cannot do that except in referring to something existing over and above the set of operations. That is to say, if one re
stricts himself to physical operations, one would be
forced to appeal to something over and beyond the operations
in deciding which of the two concepts were better.. As a matter of fact, Bridgman introduced "good" and "bad" oper- £ ations in his later writings” in order to avoid the prob lem raised by L..Russell. Unfortunately, Bridgman has not
"*L. J. Russell, "Review of Bridgman's The Logic of M o d e m Physics." Mind. Vol. 47, (1928), pp. 355-361.
6Cf. W. P. Bridgman's, "Some Implications of Recent Points of View in Physics," Rev. Int. de Phil., No. 10, (1949), p. 484. Also, see Bridgman's "Operational Analysis." Phil, of Science. No. 5, (1938), p. 126. 98 provided us with a criterion for the distinction between
"good" and "bad" operations aside perhaps from adopting a pragmatic approach, i.e., a "good" operation is one that achieves the desired result; a "bad operation" is one that does not achieve the desired result. But be that as it may, we are still not told how to define "the same set of operations" without becoming victims of circularity.
(c) Bridgman's operational principle, that different operations lead to different concepts, is likewise open to objection. His basic contention that different oper ations define different concepts has not changed as wit nessed by his later writings.7 Consequently X take it that Bridgman was convinced that no more than one set of operations can be given for one concept. For example, if one were to specify what "electric current" means via three sets of operations, one would have in accordance * with Bridgman's view operational definitions for three different concepts. The three concepts in turn should be designated by three different terms. To illustrate
Bridgman's point, allow me to cite an example. Let "X is electric" stand for "there is an electric current flowing
7Cf. Ibid., p. 481. 99 through X." Furthermore, let us suppose that the concept "X is electric" is defined by the following three sets of operations.
1. "X is electric^" stand for "if x is put close to a magnetic needle, the needle deflects."
2. "X is electric2 n stands for "if a resistance coil is connected into X, the coil heats up."
3. "X is electric^" stands for "if X is in such Q and such a solution, so and so then precipitates."0
We have three different concepts (electric^, electric2 i electric^) corresponding to three different sets of operations. The question is, in what sense are these concepts different? We have given, as the example illustrates, a separate indexical name to correspond to each set of operations. If this is Bridgman's contention,
I have no qualms about the principle that different operations lead to different concepts. However, the problem arises if Bridgman wants to claim that the three concepts (electric^, electricg, electric^) are different in the sense that "X is electric^" is not synonymous with
"X is electric2 ," "X is electric2 " is not synonymous with
Q I borrowed these examples from F. Wilson's article "Definition and Discovery," p. 300. 100
"X is electric^" and "X is electric^" is not synonymous with "X is electric^: X am not sure whether
Bridgman would agree to what I am contending. Yet certain passages from his writings appear to support the view that different operations lead to a difference in meaning. For example, in his book The Logic of Modem
Physics. Bridgman tells us:
Since Einstein's operations are different from our operations, above, his "length" does not mean the same as our "length." We must accord ingly be prepared to find that length of a moving body measured by the procedure of Einstein is not the same as that above; this of course is the fact, and the transformation formulas of relativity give the precise connec tion between the two lengths.9
In his article "Some Implications of Recent Points of View In Physics," Bridgman re-emphasized his basic contention by saying that "concepts defined by two sets of operations lead to different concepts."^ Hence it seems to me that Bridgman wants to claim that two sets of operations would lead to different concepts in the sense that there Is a difference in meaning. What I am
9 ^Bridgman, The Logic of Modern Physics, p. 9.
^Bridgman, "Some Implications of Recent Points of View in Physics," p. 481. 101 objecting to is that the differences should not be con strued as though "X is electric^" has a different meaning than "X is electric2 »" nX is electric2 " has a different meaning than "X is electric^," ... Rather I contend that the differences simply represent the growth in signifi cance of the concept of electricity. ^
Hence it is my contention that scientists are refer ring to the same concept when they talk about "X is electric^," "X is electric^11 and "X is electric^." One could justify this claim to the extent that the use of the term "X is electric" is regulated by the same set of postulates in the given diverse operational contexts.
For example, a scientist might introduce certain general laws among his postulates in order to justify his claim.
In this case, he might appeal to a certain law which stipulates that electric^, electric2 » electric^ are co extensive concepts. Thus when a scientist makes the claim that "X is electric" is defined in terms of "X is electric^," or "X is electric2 ," or "X is electric^," it makes no difference which of the three concepts one adopts; it should, however, be kept in mind that the 102 scientist justifies his claim by virtue of certain laws.^
Thus the foregoing criticisms clearly indicate that
Bridgman's early views on operationalism are too narrow in demanding that the application of a concept is uniquely specified by a single set of physical operations. This view is too restrictive as a description of concept formation. However Bridgman In his later writings admitted that the views advocated in The Logic of Modem
Physics are too restrictive if one deals strictly with a set of physical operations. One of his writings would bear him out on this point.
The mistaken idea that the operations have to be physical or instrumental combined with the dictum, "the concept is synonymous with the corresponding set of operations" has in some
1 1 ■“Bergmann, in his Philosophy of Science, p. 63, has aptly shown this point when he said: ... "one could, if one wanted to, define three different terms 'current^,' with definiens R* , 'currentg' with definiens R 2 , 'current^' with definiens R3 . The point is that although they name three different notions, It makes no difference which of the three we adopt as the definition of 'cur rent*. The controlling fact is the following tripartite law L: (1) Every current^ is also a current2 , and con versely; (2 ) Every current2 Is also a currentg, and con versely; (3) Every current3 is also a current^, and con versely... Thus, whichever of the three "indexed" notions one chooses as one's notion of a "nonindexed" current, one still has to state L separately." 103
cases led to disastrous misunderstanding. If I were writing again I would try to emphasize more the importance of the mental or "paper-and-pencil" operations.
Thus one detects a less restrictive view of operationalisra subsequent to the publication of The
Logic of Modem Physics. We shall now briefly discuss these later views of Bridgman. *
2. Later views of Bridgman on operationalism
In order to understand the later views of Bridgman,
I shall attempt to list the various kinds of operations mentioned by Bridgman. The list is not by any means intended to be exhaustive. As a matter of fact, Bridgman does not present us with any list, let alone an exhaus tive one. Indeed Bridgman casts doubt on the possibi lity of exhaustively classifying operations. ^ Conse quently when one does attempt to list Bridgman's operations, one should at no time be under the impression that such a list is exhaustive. What such a list of operations might do is help to cast some light on his views.
Bridgman distinguished several kinds of operations
■^P. W. Bridgman, "The Logic of Modern Physics after 30 years," Daedalus, No. 3, Vol. 88, (Summer 1959), p. 522.
^Bridgman, "Some Implications of Recent Points of View in Physics," pp. 479-480. 104 which he thought were important in specifying the meaning of scientific terms. Two'main kinds of operations stand out: 1) physical operations 2) non-physical operations.
Under physical operations are included
(a) Manual operations. ^
(b) Instrumental and laboratory operations. ^
It is clear that what Bridgman intended to include within the physical operation category were the various devices of observation and measurement utilized by empiri cal sciences. As pointed out earlier, Bridgman used physical operations in explicating scientific concepts.
Under non-physical operations, one can subsume the fol lowing operations:
(c) Mental operations: Although Bridgman uses the term "mental" operation, he still hesitates to explain what is meant by it. Bridgman cites "counting" as an example of mental operations, but still wonders whether the operation of counting is a "mental" or a "physical" operation when; as for example, the scientist attempts to
l^See Bridgman's article "Operational Analysis," p. 128.
See Bridgman's article "The Nature of Some of our Physical Concepts," Brit. J. of the Phil, of Science. Vol. 1, (1951), p. 258. 105 measure the length of an object by counting the number of
times that a meter stick can be brought into coincidence
in a special way with the object. As he puts it
My own use of language would make the counting operation "mental,’1 yet this is an integral part of many "physical" procedures.16
Thus in spite of Bridgman's apparent lack of enthu
siasm in providing clear-cut criterion to distinguish between "mental" and "physical" operations, he still sub divided mental operations into:
(d) Verbal operations.^7 for example, verbal defin
ition and substitution, and 1 8 (e) Paper and pencil operations.x all the procedures of logic and mathematics, all formal manipulation of symbols.
The only restriction that Bridgman stipulates for the use of mental operations is that they must be capable directly or indirectly of "making connections with
l^See Bridgman's article "Operational Analysis," p. 123.
17Ibid., p. 121.
18See Ibid., p. 123. 106 19 instrumental operations." In other words, the opera
tional principle as viewed in terms of the various opera
tions that Bridgman permitted in his later writings
amounts to the following. No term qualifies for admit
tance into the vocabulary of science unless it is con
nected in some way with physical operations. Yet a
scientific concept derives its significance not just by
being introduced via physical operations but also in terms
of non-physical operations. The only restriction is that
the non-physical operations should be connectible or cor-
relatable with physical operations.
Thus in Bridgman*s writings, one detects two strands
of operationism: narrow operationism and broad opera-
tionism. The former corresponds to Bridgman's early views
on operationism, the latter to Bridgman's later views.
These two strands of operationism seem to parallel the
logical positivist's change of view concerning the veri
fication principle, i.e., the shift from complete verifi ability to partial verifiability. The remainder of this
chapter will be an attempt to establish certain relation
ships between operationism and logical positivism.
^See Bridgman's article "The Nature of Some of our Physical Concepts," p. 260. 107
3. Bridgman's operationism and logical positivism
We now turn to a consideration o£ the connections
between Bridgman's operationalism and logical positivism.
Operationalism and logical positivism seem to be conduc
ting essentially similar methodological inquiries. Let us
consider some of the positivist inquiries.
(a) Hempel in his article "A Logical Appraisal of
Operationism" pointed out that both operationalism and
logical positivism have strongly emphasized the importance
of experiential meaning as a necessary prerequisite for
significant discourse. On one hand, logical positivism
stressed the importance of experiential meaning as a
salient characteristic of statements which lend themselves
to observational criteria; on the other hand, operation
alism, according to Hempel, considered experiential meaning as it relates to concepts which in the long run 20 are susceptible to operational definitions. Hempel's
suggestion seems to be tenable especially when one com pares certain aspects of the meaning criterion of logical positivism with Bridgman's classic statement of
^Cf. Hempel's article "A Logical Appraisal of Operationism," in P. Frank (ed.). The Validation of Scientific Theories, (N. Y.: Collier Books, 1961), p. 56. operationalism. With respect to the meaning criterion of the logical positivists, we discussed in Chapter II the various phases of the development of the verifiability criterion. Though various formulations of the meaning criterion were put forth, one central point prevailed throughout the literature of logical positivism. It was the contention that a criterion of significance applies only to statements which have empirical import. It should be kept in mind that tautologous statements are devoid of experiential meaning since they are true by virtue of their form. Thus experiential meaning as a characteristic of statements ranges over statements that are at least in principle empirically testable. So far Hempel's sugges tion is tenable even when one considers the various cri teria that were offered for the significance of theoreti cal terms where, as we saw in Chapter III, according to the method of postulates the empirical significance of sentences depended on syntactical structure and possession of significant theoretical terms. Carnap was a prime example of this method when in his article "The Method ological Character of Theoretical Concepts," he first formulated a criterion of significance for sentences in terms of his criterion for complex theoretical 109
terms. 21 A
As regards Bridgman’s brand of operationalism,
Hempel*s suggestion seems to be correct when he said that
experiential meaning was a characteristic only of admis
sible concepts. This was explicitly stated in Bridgman’s
famous statement, "In general we mean^by any concept
nothing more than a set of operations; the concept is
synonymous with the corresponding set of operations.”
However, certain questions arise in connection with
Hempel's contention that operationalism considered experi
ential meaning as a characteristic of concepts. Earlier I
said that Hempel's view "seems to be correct," because I
think that one can attribute the same concern to Bridgman which Hempel attributed to logical positivism. That is to
say, experiential meaning, which logical positivism and
Bridgman stressed, is a characteristic of statements for both movements and not just for one movement (logical
positivism). I should perhaps admit that Bridgman on the whole does talk about the admissibility of concepts in
science. Yet there is the tacit, if not explicit, assump
tion on his part to extend experiential meaning to
2lCf. Chapter III. 110 statements.
Just as logical positivists pigeoned-holed meaningful statements into analytic (formal) and synthetic (empiri cal) statements, so apparently did Bridgman when he sug gested in his writings that one include not just physical operations, but also mental operations. Bridgman, like % the logical positivists, put emphasis on the distinction between formal and empirical statements as is testified by his interest in bifurcating operations into physical and mental. One could replace, as suggested by B u r e s s , ^
"physical" and "mental" operations with "empirical" and
"logical" operations, respectively. Put differently, it seems to me that while Bridgman does construe experiential meaning as a characteristic of concepts or of terms repre senting them, one can still press the point that Bridgman does also construe experiential meaning as a character istic of statements provided that one makes the distinc tion between formal and empirical sentences based on physical and mental operations, as suggested by Buress.
Thus formal sentences derive their sanction from
E. Buress, "Operationism, Construction, and Inference," Jour, of Phil.. Vol. 37, No. 10, (July 1940), p. 396. 1X1 convention; they are true by virtue of their form. Such sentences are common in mathematics and logic. On the other hand, the meaning of empirical sentences is deter mined by the operations involved in verifying them.
(b) Another puzzling question is whether Bridgman, like the logical positivists, propounded a theory of meaning. Again and again Bridgman has told us that the operational point of view is not intended to be a theory of meaning. As a matter of fact, Bridgman tried to avoid using the word "operationalism.11 In a symposium on oper ationalism, Bridgman expressed astonishment because much of the concern of the participants was focused upon erecting a certain "ism.” In this connection he said
In short, I feel that I have created a Frankenstein, which has certainly got away from me. I abhor the word operationalism or operationism, which seems to imply a dogma, or at least a thesis of some kind.^3
If Bridgman's contention was not to advocate a theory of meaning, how is one to construe the operational point of view? To answer this question, we can do no better than to cite from his works which adequately describe what Is intended by the operational point of view. First, in his
^Bridgman, "The Present State Of Operationalism," p. 76. 112 article "Operational Analysis," Bridgman tells us
It would be difficult to exhaustively charac terize the operational technique of analysis; it is more of the nature of an art to be learned by the nature of its end, by observation of its practice by others. It certainly does not attempt to set up a theory of meaning or to be an epistemology.24
Second, at the conclusion of the article "The Present
State of Operationalism," Bridgman again emphasized that
"it must be remembered the operational point of view suggested itself from observation of physicists in action."^
At least it is evident from these quotations that
Bridgman had no intention of propounding a theory of meaning. Then, why is it that his view of operationism is mistaken for a theory of meaning? It could be that
Bridgman’s famous statement "In general we mean by any concept..." lends itself to an interpretation along the lines of the verification theory of meaning. Schlick's contention that "the meaning of a proposition is the method of its verification" coupled with his remark
24sridgmani "Operational Analysis," p. 130.
^Bridgman, "The Present State of Operationalism," p. 80. 113
that Bridgman's book The Logic of Modern Physics "is an admirable attempt to carry out the program for all con cepts of p h y s i c s " ^ might suggest that perhaps Schlick at one time equated the meaning of a proposition with the
set of operations involved in verifying it. But Schlick, as indicated in Chapter II, emphasized that by "method of verification" he meant the logical possibility of verification, not any actual procedure of verification.
Thus instead of "actual set of operations," one could stipulate "logically consistent set of operations" and, hence, hold that concepts are defined in terms of logi cally possible operations. Bridgman's brand of operation alism would then read: a concept is meaningful if the
sentences in which it occurs are verifiable or falsifiable by a finite set of logically consistent operations. Put
in this manner, operationalism and logical positivism would essentially be the same and thus the shortcomings of complete verifiability and falsifiability would recur as
indicated in Chapter II. Again, one could modify it by
introducing "partial verifiability" instead of "complete
verifiability." In this way, Bridgman's principle as !
2^Schlick, "Meaning and Verification," p. 149. 114 stated In his famous statement would read: a concept is meaningful if the sentences in which it occurs are par tially, verifiable by a set of logically consistent oper ations. But thus formulated, operationism would still have to answer adequately the various objections raised against partial verifiability.
The above account was just an attempt to show how one could construe Bridgman’s famous statement as embodying a verification theory of meaning. If one contends that
Bridgman’s operationalism is nothing but a variant of the verification theory of meaning, one might formulate this point as I did above. Lest there would be any confusion,
I should perhaps mention that I am not advocating the view that Bridgman’s operationalism is nothing but a variant of the verification theory of meaning. Undoubtedly such a view may appear implicit in Bridgman’s writings.
But I still think that such an interpretation would betray
Bridgman’s main concern. It is my impression that oper ationalism is not a theory of meaning at all— let alone a verification theory of meaning. Rather it is a demand for a certain kind of scientific practice, namely, that one may only allow concepts that meet the operational standards into the scientific universe of discourse. Yet 115
concepts lacking the operational standards are still
intelligible or, if you please, meaningful. Judged from
this vantage point, though operational concepts are
essential in science, other kinds of concepts are meaning
ful. As a matter of fact, Bridgman in reconsidering his book The Logic of Modem Physics makes this point clear by saying
"Meaning" as I used it was tied up with operational analysis. My position was that, if a term could not be accounted for in terms of the operations contained in the particular universe of operations for the purpose on hand, such a term was to be adjudged "meaningless." This position is still tenable, but it may lead to misunderstanding in a wider context, because in everyday usage "meaning" has much wider connotation. For some sort of "meaning" can be ascribed to any expression whatever.27
In another article, Bridgman made it clear that
"the operational aspect is not by any means the only aspect of meaning."2® Thus I think one can safely conclude
that Bridgman did not want to restrict the word "meaning" only to operationally defined concepts, for, as the above quotations reveal, he was willing to allow meaning to
terms which are not operationally definable. This change
^Bridgman, "The Logic of Modern Physics After 30 Years," pp. 522-523.
^Bridgman, "The Nature of Some of Our Physical Objects," p. 257. 116 in Bridgman's attitude is certainly reminiscent of the developing views of the Vienna Circle wherein it was first declared that traditional metaphysics was meaningless.
(To say the least, this dogmatic attitude caused much uneasiness.) Later, the logical positivists found it necessary to distinguish various components of meanings.
Ultimately they were led to the view that traditional metaphysics is devoid of any cognitive meaning. Meta physical sentences might have other kinds of meaning, for example, emotive or pictorial, which though lacking any cognitive content, might still have strong psychological overtones. Thus corresponding to the logical positivists' classification of cognitive and non-cognitive meaning is the classification that is implicit in Bridgman's writings between operational and non-operational meaning.
In this way, operational meaning would be included within the cognitive realm and non-operational meaning would be 29 included within the non-cognitive realm.
(c) If I am correct in claiming that Bridgman's operationalism lends itself to the distinction between
The word "operational!' in "operational meaning" should be understood as ranging over both physical and non-physical operations. 117 cognitive and non-cognitive meaning along the lines suggested, one wonders whether those terms or sentences, according to Bridgman, that lack operational meaning are pseudo-terms or pseudo-sentences. We know that the log ical positivists considered pseudo-sentences to be devoid of any cognitive meaning. Carnap explains pseudo-sentences as follows:
We say that an expression A is a pseudo- statement if A is devoid of cognitive mean ing, but has the grammatical form of a declarative sentence and may therefore lead to psychological effects which are similar to those of a genuine statement.30
It seems to me that Bridgman would not object to say ing that those sentences which lack operational meaning are pseudo-sentences for, in his book The Mature of Physical
Theory. Bridgman tells us that
Not only are there meaningless questions, but many of the problems with which the human intellect has tortured itself turn out to be only "pseudo-problems," because they can be formulated only in terms of questions which are meaningless. Many of the tradi tional problems of philosophy, of religion, or of ethics, are of this character.
30 P. A. Schilpp, The Philosophy of Carnap, p. 874.
3*-P. W. Bridgman, The Nature of Physical Theory. (Princeton Univ. Press, Princeton, N. J., 1936), p. 12. 118
Thus while it is true that Bridgman does not use the terms "cognitive" and "non-cognitive," it still appears that he would not object to claiming that sen tences which lack operational meaning are pseudo sentences. But to say that they are pseudo-sentences does not suggest the dogmatic position that one might take -- namely, that they are devoid of any meaning.
Let me conclude this section by saying that Bridgman*s operationalism is not to be considered as providing us with a meaning criterion. Rather, as indicated earlier, it is a demand for a certain kind of scientific practice.
In other words, operationalism, as I think it should be understood, is a technique for the formation of scien tific concepts by means of operational definitions. In this sense, operationalism would be a genuine part' of the movement called logical empiricism. Judged from this viewpoint, operationism as a technique in concept for mation manifests itself in the practice of scientist.
Considered in this manner, operationalism would help r • scientists avoid introducing concepts into the universe of scientific discourse which do not meet operational standards. The question that naturally arises is whether operationalism is an adequate technique in the formation 119 of scientific concepts. To this task, we now direct our attention.
4. Adequacy of operationism as a technique in concept formation.
If operationism is to be considered as a technique in concept formation, one would have to^inquire about its adequacy as such a technique. We saw in previous chapters how keenly interested the logical positivists were in stating the precise link between theoretical and observa tional terms in order to construct a criterion of meaning fulness. By the same token, operationalism, if it is to be seriously considered as a valuable technique in concept formation in science, must seek to explicate the logical relationships between theoretical and observational terms.
Bridgman did not bother to articulate the logical link between theoretical and observational terms. The logical positivists, on the other hand, were quick in noticing some of the logical committments underlying Bridgman*s operationism. For instance, Hempel in his article "The
Present State of Operationalism" suggested that in order for operationism to be considered as a technique for the formation of scientific concepts, one would have to deal with two important issues: first, to provide an adequate explication of the concept of operational definition; second, to show that operational definitions are ade quate for introducing all the theoretical terms in science.3^
A. Explication of operational definition
The term "operational definition" has been used by
Bridgman in at least two ways. First, it has been used to refer to a sentence connecting a term to be "defined" with other terms describing laboratory procedures and manipulations. Second, the term "operational definition has been used to refer to mental operations such as pencil-and-paper operations, verbal operations. Hempel referred to such operations as "symbolic operations.1,33
Symbolic operations, according to Hempel, enable us to trace the connections among theoretical terms, or those between theoretical and observational terms.
With respect to the first use of "operational defi nition," one could generalize its use by saying that an operational definition is construed as introducing a scientific term by stipulating that its application
32cf. Hempel, "A Logical Appraisal of Operationism, p. 58-60.
33Ibid.» p. 57. ranges over all those cases which reveal a certain res ponse if the test conditions are met. In this connection,
Carnap as well as Hempel explained the use of operational definition in terms of stimulus and response. Their point was made in connection with the introduction of disposi tional terms which, according to Hempel and Carnap, express the disposition of one or more objects to react in a certain way (response) under specified conditions
(stimulus). As Carnap puts it:
The introduction of D^r by a specification of the list of operations and the charac teristic result R is therefore sometimes called an operational definition.34
Hempel*s way of explaining an operational definition of a concept was
Hence, an operational definition of a con- cept--such as a property or a relationship, for example--will have to be understood as ascribing the concept to all those cases that would exhibit the characteristic- response if the test conditions should be realized.35
Concerning the second use of "operational definition" for the introduction of scientific concepts, one would
^Carnap, "The Methodological Character of Theoretical Concepts," p. 65.
35Hempel, "A Logical Appraisal of Appraisal," p. 59. 122 have to depend not on Bridgman but on logical positivists when the reference is made to "mental operation;" "pencil- and-paper operation" was replaced by them by specifying
the logical form of sentences, such as Carnap*s reduction
sentences, and the logical principles governing the use of these sentences. As Hempel tells us
... every meaningful scientific term must therefore permit of an operational definition. Such definition may refer to certain symbolic operations and it always must ultimately make reference to some instrumental operation.^6
The above sketch is sufficient to explicate the concept of operational definition for our purposes. We thus interpret operational definitions in terms of test procedures or symbolic operations required to determine whether the defined term applies in a context. We should, then, inquire, about the logical connection between theo retical and observational terms if operationalism is to be considered as a technique for the formation of scien tific concepts. Let us take another glance at some methods which attempted to cast light on the logical link between theoretical and observational terms.
36Ibid., p. 57. 123
B. Ways of introducing theoretical terms.
One of the earliest attempts provided by logical positivists to introduce theoretical terms was by means of an explicit definition. This view amounts to saying that all significant terras in science are explicitly defined by means of observation terms. Operationalism ac cords well with the explicit definition method in its claim that all significant terms in sciences are intro duced by operational definitions.^ However, the explicit definition view is too restrictive. Carnap showed that any attempt to introduce disposition terms via explicit definition is doomed to failure once one formalizes the definition. Consider the term "elastic." According to the explicit definition view, one might define it by saying that an object x is elastic if and only if it meets the following condition: if x is stretched and then released at any time t , x contracts at time t. In symbolic notation, it is rendered as:
(1) E(x)S(t) ( (S(x,t) • R(x,t) )»C(x,t) )
But since the definiens of (1) is construed in terms
37Cf. C. Hempel, Aspects of Scientific Explanation and Other Essays in the Philosophy of Science, p. 109. 1 2 4
of a truth-functional "if-then," any object that is
never stretched would have to be considered elastic. This
difficulty would be avoided should one use Carnap's method
of reduction sentences. Thus, instead of (1), we have
(2) (x) (t) ( (S(x,t) ) 3 (E(x) 35 C (x,t) )
Even though the "if-then" in (2) is construed truth-
functionally, still the method of reduction sentences does
not permit one to say that an object which is never sub
jected to the test conditions is elastic. In order to
say "x is elastic," x must meet the test-conditions, i.e.,
be stretched and then released at time t. If the test-
conditions are not fulfilled, the meaning of the term
"elastic" is left undetermined. In a nutshell, this is the
advantage of reduction sentences. They permit us to ascribe
a disposition to any object only if the relevant test-
conditions are fulfilled and found to have a positive re-
sponse. On the other hand, according to the explicit def
inition view, any object has any disposition if it does not meet the relevant test conditions. Thus, according to
the method of reduction sentences, there need not be just
one criterion for the use of an unobservable term; the more science advances, the more observational criteria we have for the application of scientific concepts. Reduction 125 sentences guarantee us that certain scientific concepts are only partially defined. Their full meaning is left
"open.*1 Hence reduction sentences, according to logical positivists, provide a better interpretation of dis position terms and also allow a more adequate formulation of operational definition, which in general should not be construed as complete definition but as partial definition.
In this connection, Hempel sums up Carnap's use of re duction sentences in the following way:
... reduction sentences offer a precise formulation of the intent of operational definitions. By expressing the latter as merely partial specifications of meaning, they treat theoretical concepts as "open."™
Despite the obvious advantages of the use of reduc tion sentences,.Carnap in his later writings rejected the view that reduction sentences adequately introduce theo retical terms. In lieu of. that view, he proposed a postu- lational view of significance for theoretical terms.
Briefly, this view states that a theoretical term (M) is significant if there is a sentence S using M in such a way that from S and the remainder of the theory, it is logi cally possible to derive an observation sentence that
38Ibid., p. 189. 126 could not be deduced without S.
Carnap’s postulational view seems to have been prompted by certain difficulties that one encounters in attempting to introduce scientific theoretical concepts that are highly abstract in character, such as molecule, electron energy level, and psi-function. These concepts appear neither in our observational vocabulary nor in the hierarchy of empirical concepts; as we know, science is replete with such terms which are not introduced by ex plicit definition or reduction sentences based on obser vational terms. The way to introduce such abstract theo retical terms in the advanced stages of science is to set up a theoretical system that is formulated in terms of them.
Then we give the system an experiential interpretation which, in turn, confers empirical meaning on the theoretical terms.
In this way, theoretical terms are construed to possess experiential meaning only within the theory under consider ation. The theory, moreover, can guarantee only partial on interpretation of some of its theoretical terms. 7
Since we are again considering the introduction of theoretical terms, I should point out that in Chapter III
qq J?We discussed this method in Chapter III, Section I. 1 2 7 we suggested that theoretical terms are significant if they are a part of accepted and well-established scientific theories. These theories are interpreted axiom systems which are connected and correlated at various points with observational data in varying degrees. In other words, it is my contention that theoretical terms are V introduced in various ways including the three methods suggested, i.e., by explicit definition, by bilateral re duction sentences, and by postulates. The unfortunate thing that one finds, however, is the stubbornness of some philosophers of science to make one method supreme and try to force others to conform to the prescribed method. In this connection, Feyerabend’s suggestion-- that one is a good empiricist only if he is prepared to work with many alternative theories rather than with a single point^— best describes my view of how theoreti cal terms are introduced. Hence one might find certain theoretical terms in scientific theories that are expli citly defined, others that are introduced by reduction sentences, and still others introduced by a set of pos tulates.
^Cf. P. K. Feyerabend, "Explanations, Predictions, Theories," in Delaware Seminar on the Phil, of Sc.. (B. Baumin (ed.) , (N.Y.: Wiley Interscience), 1963) , p.37. 128
Thus I think that strict operationalism--a concept is meaningful if it is defined in terms of a set of physical operations— is too narrow. However, such a view might be tenable with respect to certain concepts. For example, I can define "length" in terms of a set of strictly physical operations such as the use of a * measuring rod. Hempel1s example concerning "weight”
(3) Def. M lx ^ (Ey) Rxy41 was an attempt to show that one could define certain low-level theoretical terms of empirical science by means of observational terms. Yet Hempel admits that his survey ”. . . has yielded no conclusive argument for or against the possibility of explicitly defining all theo- / o retical terms. . . . But surely it is evident by now that many scientific terms are not explicitly defined.
If I were to introduce the term "elastic” in my scientific theory, I would most likely use Carnap’s method of reduction; on the other hand, if I were to introduce the
41c. Hempel, "The Theoretician's Dilemma," p. 196. "M^x" stands for "an object x having the property of being exceeded in weight by at least one object." "Rxy" stands for "x is less heavy than y."
42Ibid.. p. 205. 129 concept of atom, I would use the postulational method since "atom" appears neither in the plane of observables nor in .the hierarchy of empirical concepts, but is anchored
in the latter by coordinating (correspondence) postulates or definitions.
Recall that the methods of reduction and postulates are construed so that the terms they introduce are left
incomplete or partially defined. In the reduction sen
tence method, a predicate is assigned meaning only if it meets the relevant test conditions. Likewise the method of postulates only allows us to say that theoretical terms are partially and indirectly interpreted with respect to
the theory being employed. It would seem that Bridgman has ignored striking differences in the manner in which various theoretical terms are introduced or anchored. As a technique for concept formation, operationalism merely stressed one of the relationships between theoretical and observational terms--the operational definition for concepts appearing in the empirical hierarchy of terms.
In this connection, credit is thus due the logical posi tivists who were quick in noticing inadequacies of Bridgman*s operationalism. 130
5. Summary and concluding note
I discussed in this chapter the operational point of
view as propounded by Bridgman. We focused on the early
and the late views of Bridgman’s operationalism. Then we
compared operationalism with logical positivism. We
found that they both shared similar methodological views
aricl that they differed only in emphasis. Logical posi
tivists were extremely sensitive to the logical commit
ments underlying the introduction of scientific concepts.
Bridgman, as we noticed, did not elaborate on a precise
explication of the logical relationship between theo
retical and observational terms. This task, for better
or worse, was carried out to a great extent by the
logical positivists. Undoubtedly they encountered many
difficulties. But the work the logical positivists
carried out, especially in attempting to elaborate on a
precise logical relationship between theoretical and
observational terms, greatly stimulated inquiry into the methodological problems of the empirical and social
sciences.
One should emphasize that whereas the logical posi
tivists had painstakingly devoted a great deal of time
to elaborating a precise formulation of a meaning 131 criterion, Bridgman, on the other hand, did not seem to bother about such an enterprise. His concern was pri marily. to "operationize" scientific concepts without actually presenting any elaborate theory of meaning.
The various quotations cited earlier clearly indicate that Bridgman was not interested in propounding a theory of meaning. Rather, he was interested in presenting a certain technique for the formation of scientific con cepts. According to Bridgman, this "operational" tech nique is best revealed by the practice of 20th century scientists themselves.
One should also note that operationalism, even if viewed as a criterion of significance, fits well with the thesis that was advocated in the last section of
Chapter III. It was our contention that in order to decide whether a sentence Is significant, one should be able to translate It into some other idiom, which if called for, must in turn be translated into the ordinary idiom. For example, suppose I were to inquire whether the sentence "the ring is electrically charged" Is signi ficant. Our answer would be In the affirmative since the sentence is a part of the scientist*s idiom. But let us suppose further that one is not conversant with the 132
scientist's idiom. Then, what we do is to provide a
translation for the sentence "the ring is electrically
* charged" into the ordinary idiom. One way of doing it is
to translate it into the sentence "The leaves of this
electroscope diverge." Thus the thesis we suggested in
Chapter III would attribute significance to sentences which V include operationally defined terras.
In concluding, I would like to point out that in
spite of the fact that the logical positivists encountered
a number of difficulties in their formulation of an ade
quate meaning criterion and despite the difficulties
concerning Bridgman's operationalism, both schools of
thought have far reaching implications for psychology.
As a matter of fact, this seems to be the point that
logical positivism and Bridgman have most in common,
namely, their influence upon contemporary schools of
psychology, especially behaviorism. The following chapters
will be concerned with showing this dual influence on
behaviorism. CHAPTER V
WATSON AND LOGICAL POSITIVISM
Introduction
Like logical positivism, which was a movement directed against speculative philosophy, behaviorism as initiated by Watson early in this century was a movement directed against the old mentalistic psychology, i.e. Introspec-
tionist schools of thought. Likewise one can say that operationism as initiated by Bridgman in 1927 in his
Logic of Modern Physics was another movement which focused
its attention upon getting rid of the last vestiges of metaphysical obscurity still remaining in some parts of science.
It Is clear that logical positivism had no direct
influence on Watson inasmuch as he published his first behavioristic views in 1913, in an article entitled
"Psychology as the Behaviorist Views it." This is almost
two decades before any official publication by the logi cal positivists appeared. This is not to say that
Watson could not have been influenced indirectly by the
133 134 new positivistic movement in central Europe, which was already taking shape as early as 1907.^ It is not at all unlikely that Watson was influenced by William James' pragmatism which flourished in his early years. On the other hand, certain writers have tried to establish a definite link between pragmatism and logical positivism. 2 P. Frank, for instance, writes that upon reading Carnap's exposition of the meaning of a statement in science, he was as surprised as Carnap was later to find the closeness of Carnap's views to William James' pragmatic requirement that the meaning of any statement is given by its "cash value," or by what it means as a direction for human be havior. In the United States, we find C. W. Morris^ using the name "logical empiricism" in an effort to strengthen the tie between logical positivism and pragmatism.
Be all this as it may, our task in this chapter is primarily focused upon (1) a brief account of Watson's
^See P. Frank, Modern Science and its Philosophy. Chapter 1, (passim).
2Ibid., p. 33.
^C. W. Morris, "Pragmatism and Logical Empiricism" (i954) in The Philosophy of Rudolf Carnap. Paul Schilpp (ed.), pp. 87-98. 135
behaviorism, (2) Watson’s influence upon some logical
positivists, such as Carnap and Hempel, and (3) an
evaluation of logical behaviorism.
1. Watson's behaviorism
Watson's declaration of independence from the intro-
spectionist-dominated schools of thought was originally
introduced to scholars in his famous article of 1913
entitled "Psychology as the Behaviorist Views it."^ A
salient characteristic of Watson's article was his op
position to all psychology dealing with consciousness.
Watson declared that psychology must break with the past,
discard the concept of consciousness altogether, and begin
at the beginning by constructing a new science built on objective findings. In other words, if psychology is to become a science, it must follow the example of the
physical sciences. A psychology that is based upon cons
ciousness has subject matter which cannot be treated
scientifically. As Watson puts it:
Psychology as the behaviorist views it is a purely objective, experimental branch of natural science which needs introspection as little as do the sciences of chemistry
^J. Watson, "Psychology as the Behaviorist Views it," Psvch. Rev. (1913), pp. 158-177. 136
and physics... It can dispense with con sciousness in a psychological sense. The separate observation of "states of conscious ness" is on this assumption, no more a part of the task of the psychologist than of the physicist. We might call this the return to a non-reflective and naive use of conscious ness. In this sense consciousness may be said to be the instrument or tool with which all scientists work.^
An assumption underlying this quotation was not that '
Watson denied the existence of consciousness or mental g events, as some contended. It seems to me that what
Watson recommended was that if psychology is to achieve the status of a science, it should not consider conscious ness as its primary subject matter. Rather, the primary subject matter of psychology, according to Watson, is exactly the same in kind as that of all other sciences.
In this connection, Watson said:
You will find, then, the behaviorist working like any other scientist. His sole object is to gather facts about behavior— verify
5Ibid., p. 176.
^Bergmann, in his article, "The Contribution of John B. Watson," Psychological Rev.. Vol. 63, (1956), pp. 265- 276, tells us that the contribution of Watson was method ological behaviorism rather than metaphysical behaviorism. However, Bergmann also contends that Watson was a social philosopher as well as a metaphysician. 137
his data--subject them both to logic and to mathematics (the tools of every scientist).^
In a nutshell, what I am suggesting is that Watson should not be interpreted as implying that consciousness does not exist. Rather, one should interpret Watson as claiming that one cannot prove the existence of conscious- ness or any theory based on it by scientific tests, for consciousness lacks the characteristic of being publicly observable or apprehended. Thus classical psychology which depended so much upon the concept of consciousness is unfit to be included with the physical sciences. Put differently, classical psychology is declared unfit for scientific purposes in the eyes of Watson because there is no way of verifying either directly or indirectly the mentalistic propositions within its realm.
As the above account reveals, Watson's behaviorism insisted on objective techniques for securing data for the psychologist. Again, to use Watson's words, be haviorism "attempted to make a fresh, clean start in psychology, breaking both with current theories and with
7j. Watson, Behaviorism. (Chicago: University of Chicago Press, 1920), p. 6. 138 traditional concepts and terminology. *' Thus Watson proposed to include only such observations as can be made by independent observers, exactly as in chemistry or physics. Watson also emphasized that psychology should limit itself to the study of responses made by the organism to certain stimuli. That is to say, all meaningful psy chological sentences are to be interpreted in terms of stimulus and response. In his article "Psychology as the
Behaviorist Views it," Watson made his point clear about how psychology is to be understood by saying:
I believe we can write a psychology, define it as Pillsbury, and never go back upon our definition, never use the terms con sciousness, mental states, mind, content, introspectively verifiable, images, and the like.9
Later iti the article, Watson suggested how one might define psychology. He adds that "...it can be done in terms of stimulus and response, in terms of habit formation, habit integration and the like."^ Watson1 s emphasis on stimulus-response theory (or S-R theory, as it is commonly
o J. Watson, Psychology from the Standpoint of a Behaviorist (3rd.ed.), (Philadelphia:Lippincott), 1929, p.4.
9Watson, "Psychology as the Behaviorist Views it," p. 166.
10Ibid., p. 167. 139 called) simply asserts that all behavior is in response to stimuli, whether the behavior is explicit (as in running) or implicit as (in day-dreaming).^ It should, however, be kept in mind that contemporary S-R psychologists go be yond Watson1s behaviorism in their willingness to infer between the stimulus and response all^sorts of states or processes which they call "intervening variables." In the
following chapter, I shall have more to say about "inter vening variables."
It is not surprising, then, even from this short ac count that the logical positivists should accept such a behavioristic account of psychology. Their main objection was that Watsonian behaviorism limited itself to the study of responses made by the organism to certain stimuli. 12 Carnap, for example, admitted the epistemological simi larity of positivism's physicalism with behaviorism; while
U-For a more detailed account, of stimulus-response theory, see Watson's Behaviorism, op. cit., Chapters I, II. Also, see Watson's The Ways of Behaviorism. (New York: Harper and Brothers Publishers, 1928), pp. 5-21.
^R. Carnap, "Psychology in Physical Language," (1932-33) translated and reprinted in logical Positivism. A. J. Ayer, (ed.), pp. 165-198. 140 13 Otto Neurath with his implicit approval of behaviorism
suggested the term "behavioristics" as a replacement for
"psychology." It is also interesting to note, in this
connection, that at one time Hempel called himself a logi
cal behaviorist^ in an attempt to distinguish his broader
logical theory about psychological propositions from the
restricted area of stimulus-response, which was Watson*s main concern. What Hempel actually proposed was a logical,
analysis of psychological propositions in order to dis
cover whether they display any relationship to the ones in
empirical sciences. He was not at all concerned with de
veloping a psychological theory like behaviorism.
On the one hand, Watson was primarily interested in
propounding methological behaviorism. As a methodological
thesis, behaviorism involves no commitment whatever con
cerning the logical analysis of mental words, nor does it
imply the denial of mental states. What methodological behaviorism asserts could be simply stated as follows:
The subject matter of scientific psychology is nothing but
^0. Neurath, "Protocol Sentences*' (1932-33), trans lated and reprinted in Logical Positivism, pp. 199-208.
^C. Hempel, "The Logical Analysis of Psychology,*1 (1935) translated and reprinted in Feigl and Sellars, Philosophical Analysis, pp. 373-384. 141
behavior, and as a science, psychology should study human
and animal behavior without introducing "private" mental
states over and beyond the publicly observable stimuli and
responses. Thus, as a methodological thesis, behaviorism
leaves open the question whether there are mental events;
they are of no relevance to science. On the other hand, M r logical behaviorism or (as some call it "analytical behav
iorism") claimed that all sentences containing psychologi
cal terms are transformable by analysis Into other senten
ces containing no psychological terms. In this way, the
behavioristic requirement of logical positivists such as
Hempel and Carnap would amount to a logical analysis of
psychological terms or sentences containing psychological 1 e terms. It was under the influence of 0. Neurath J which
led Carnap to formulate his physicalism. To this task, we shall now direct our attention by considering the views
l-*0n one hand Feigl, in "The Mind-Body Problem In the Development of Logical Empiricism," in Feigl and Brodbeck, Readings in the Philosophy of Science, p. 615, reports that it was Neurath's influence and Popper's critical suggestions which led Carnap to formulate physicalism. On the other hand, K. Popper in Conjectures and Reputa tions: The Growth of Scientific Knowledge. (New York: Harper and Row, Publishers, 1968), p. 265, reports that as a result of Neurath's Influence Carnap gave up the phenomenalistic thesis In favor of physicalism. 142 1 fi advocated by Carnap and Hempel on logical behaviorism,
2. Carnap and Hempel on logical behaviorism.
Carnap; Initially, Carnap adopted the "phenoroen- alistic thesis" which, roughly speaking, based the struc ture of the language of science upon sentences expressing one's immediate sensory experience. In other words these advocates of the phenomenalistic thesis believed that all concepts of social and natural sciences may be defined by elementary experiences. As Carnap later stated it in his article "Testability and Meaning."
One of the fundamental theses of positivism may perhaps be formulated in this way: every term of the whole language L of science is reducible to what we may call sense-data terms or perception terms... Positivists therefore believed that every descriptive term of science could be defined by percep tion terms, and hence, that every sentence- of the language of science could be trans lated into a sentence about perceptions.^7
Such a phenomenalistic approach was later abandoned by Carnap, as well as other logical positivists, in favor of what came to be known "physicalism" wherein the basic
^ I t should be kept in mind that behaviorism as a philosophy of psychology which Hempel and Carnap advocated is continuous with the doctrine called physicalism.
^Carnap, "Testability and Meaning," p. 67. The "Phenomenalistic Thesis" was presented by Carnap in Per Logische: Aufbau der Welt: Berlin, 1938. 143 empirical sentences are those reporting observations of physical objects, including people’s bodies. Briefly, then, the thesis of physicalism as originally accepted by the members of the Vienna Circle amounted to the following:
First, terms occuring in the language of science are to be explicitly defined in terms of publicly observables; second, every sentence in the language of science is to be translated into sentences which contain only obser vable predicates.
Having roughly outlined Carnap’s physicalism, let us see how Carnap extended it to psychology. His basic aim was to establish the possibility of translating all psychological sentences into a physical language which would, in turn,.serve as a universal language for all sciences. Carnap’s contention concerning this program is formulated in his article "Psychology in Physical
Language" in the following manner:
Every sentence of psychology may be formulated in physical language. To express this in the material mode of speech: all sentences of psychology describe physical occurrences, namely, the physical behavior of humans and other animals. This is a sub-thesis of the general thesis of physicalism to the effect that physical language is a 144
universal language, that is, a language into which every sentence may be translated.18
Later in the article, Carnap asserted that
Our thesis states that a definition may be constructed for every psychological concept (i.e. expression) which directly or indirectly derives that concept from physical concepts...All that we are deman ding is the production of the definitions through which psychological language is linked with physical language. We main tain that these definitions can be pro duced, since, implicitly, they already underlie psychological practice. ^-9
From these quotations, Carnap's aim is, I think
quite clear. He wanted to "physicalize" psychological
concepts. He wanted to provide explicit behavioristic
definitions of psychological concepts. The philosophical
principle underlying this thesis is the verifiability
criterion of meaning. As Carnap puts it: "A sentence 20 says no more than what is testable about it." For
instance, to say "A is excited" means nothing more than
that he is behaving in such and such a way, that given
certain stimuli, he responds in such and such a manner.
Consider another example from Carnap's Philosophy and
^■^Carnap, "Psychology in Physical Language," p. 165.
19Ibid., p. 167.
20Ibid., p. 174. 145
Logical Syntax showing what Carnap meant by the physicalistic thesis when extended to psychology.
The sentence "At ten o'clock Mr. A was angry" is translated into the physical istic language as: At ten o'clock Mr. A was in certain bodily conditions which are characterized by the acceleration of breathing, pulsating, by the tension of certain muscles, by the tendency to certain violent behavior.21
If one wants to claim that over and beyond these phy sical manifestations, there is an inner state of anger which is completely different from the actual or po tential behavioral manifestations, then, according to
Carnap's thesis, one is claiming something that we simply do not know how to verify. Hence judged by the verifi ability criterion, explication of mental words in terms of an inner state is devoid of any cognitive meaning.
Thus if one insists on asserting that so-and-so is angry in some mentalistic sense of the term, such assertion 99 turns out to be "a metaphysical pseudo-sentence."
Although the above account of Carnap's physicalism roughly parallels Watson's psychological behaviorism, one
21r , Carnap, Philosophy and Logical Syntax. (London: Kegan Paul, Trench, Trubner and Co., Ltd., 1936), pp. 89-90.
^Carnap, "Psychology in Physical Language," p. 174. 146 should keep in mind that Carnap's interest in behaviorism was strictly focused upon the logical and epistemological commitments of behaviorism rather than its specialized methods. This point of view was aptly put by Carnap by saying
We have not linked our exposition with a statement of behaviorism since our only concern is with epistemological foundations while behaviorism is above all else interested in a specific method of research and in specific concept formations.23
Hempel, following on the heels of Carnap, took up the physicalistic thesis in the 1930's and, like Carnap, attempted to apply it to psychology. Thus, in what follows, I will set forth some of Hempel's views on physicalism.
B. Hempel: His position on physicalism as it re lates to psychology is best summed up at the end of his article "The Logical Analysis of Psychology."
All psychological statements which are meaningful, that is to say which are in principle verifiable, are translatable into propositions which do not involve psychological concepts, but only the con cepts of physics. The propositions of psychology are consequently physicalistic propositions. Psychology is an integral
23Ibid.. p. 181 part of physics... This logical analysis of which the result shows a certain affin ity with the fundamental ideas of behav iorism, constitutes the physicalistic conception of psychology. 24
The quotation clearly indicates that Hempel, like Carnap,
attempted to "physicalize" psychological sentences. And
the test whether a particular psychological sentence is meaningful is whether it passes the requirements of the
verifiability criterion. An example of a psychological
sentence which Hempel claims to be verifiable and hence meaningful is the following: "Mr. Jones suffers from
intense inferiority feelings of such and such kinds..."2^
This sentence, according to Hempel can only be confirmed or falsified by observing various aspects of Mr. Jones' behavior, such as his gestures, flushing of his skin, blood pressure, the events that occur in his nervous sys
tem, etc. Hence the sentence concerning Mr. Jones' in
feriority feelings means only this: "Such and such hap penings take place in Mr. Jones body in such and such 26 circumstances." Another example of a psychological
^Hempel, "The Logical Analysis of Psychology" (1935), translated and reprinted in Feigl and Sellars, Readings in Philosophical Analysis, p. 378.
25Ibid.
26Ibid. 148 sentence which Hempel considers in discussing physi- calism is "Paul has a toothache." Hempel asks, "What is the specific content of this proposition, that is to say, what are the circumstances in which it would be verified?" He goes on to list some of the circum stances which might verify the proposition "Paul has a toothache." a) "Paul weeps and makes gestures of such and such kinds." b) "At the question, ’What is the matter?', Paul utters the words 'I have a toothache.'" c) "Closer examination reveals a decayed tooth with exposed pulp." d) "Paul's blood pressure, digestive process, the speed of his reactions, show such and such changes." e) "Such and such processes occur in Paul's central nervous system."
Hempel asserts that the sentence "Paul has a toothache" is "nothing but an abbreviated formulation of all these test sentences. And, of course, all of these test sentences are verifiable.
27Ibid., p. 377. 149
X think that, by now, one can see why behaviorism was attractive to some logical positivists. Certainly one can observe certain similarities between the two movements. While Watson did not formulate a "meaning
criterion" like the logical positivists, still 1 think
such a criterion is implicit in his writings. It is true
that Watson did not bother to work out a meaning criterion
in order to ascertain which part of psychology is meaning
ful. Perhaps Watson was wise in not formulating a meaning
criterion for his behavioristic psychology and simply
followed, instead, his instinct for objective observation,
since we are now fully aware of the difficulties that were
encountered in the development of the meaning criterion.
X shall now devote the remainder of this chapter to an
evaluation of logical behaviorism.
3. Evaluation of logical behaviorism:
Despite the fact that the thesis of logical behavi orism advocated by Hempel and Carnap is interesting and
attractive, there are still many difficulties that one
encounters in connection with such a thesis. In* what
follows, we shall consider a representative sample of
some of the problems. 150
(a) Recall that the thesis of physicalism, as orig
inally stated, roughly asserted (1) that every term of the
language of science, including those of psychology, can be
explicitly defined in terms of observation terms, and (2)
that every sentence of the language of science is trans
latable into sentences containing observation terms. In
"Testability and Meaning," Carnap replaced "definability" by "reducibility" as an adequate way of introducing dis positional theoretical terms. As Carnap puts it:
We only assert reducibility of the terms, but not--as we have done in our former publication--definability of the terms and hence translatability of all sentence.28
Hempel also modified his early views on physicalism. In a footnote to his early article "The Logical Analysis of
Psychology," Hempel modified his early views by contending
The thesis that all statements of empirical science are translatable, without loss of theoretical content, Into the language of physics should be replaced by a weaker assertion that all statements of empirical science are reducible to sentences in the language of physics...29
But Carnap as well as Hempel have since abandoned the
^Carnap, "Testability and Meaning," p.-70.
2^Hempel, "The Logical Analysis of Psychology," (1935), p. 373, footnote (1). 151 position of introducing theoretical terms by means of reduction sentences. Carnap, as we saw earlier (Chapter
III-C.) in "The Methodological Character of Theoretical
Concepts," chose to introduce theoretical terms (including psychological concepts) by means of postulates and corres pondence rules which relate theoretical terms to the term of the observation language. As indicated earlier, the
introduction of theoretical terms by means of postulates will, like their introduction by bilateral reduction sen
tences, allow the possibility of the "openness" of scien
tific terms. Addition of further postulates and corres pondence rules may enrich the meaning of theoretical
terms.
Granting the logical positivists' claim that scien
tific terms (including psychological terms) have their meanings specified by correspondence rules and postulates, the question that has to be answered is, how adequate is the assumption underlying the physicalistic thesis that the meaning of a sentence is the condition of its veri fication?
(b) As indicated in Section 2-B, above, the meaning of "Jones is angry" is supplied by the behavioral mani festations of Jones, namely, that Mr. Jones will respond 152
In such and such a way to certain stimuli, that various happenings are occurring in Mr. Jones* central nervous system, etc. Thus the philosophical justification of this thesis is the logical positivist's verifiability criterion of meaning. We have indicated in Chapters II and III the numerous difficulties encountered with such a criterion.
As a matter of fact, we have found good reasons to reject the verifiability criterion. But does a rejection of the verifiability criterion imply a rejection of the physical istic thesis? It would seem that if physicalism is taken as a variant of the verifiability criterion, then all the inadequacies of the verifiability criterion would equally apply to the physicalistic thesis. Hence a rejection of the verifiability criterion would perhaps lead one to reject physicalism. On the other hand, if physicalism like operationalism, is considered as a technique in concept formation, then it seems to me that one could re ject the verifiability criterion of meaning without giving up the physicalistic thesis. Naturally, the question that now arises is, how adequate is physicalism as a technique in the formation of psychological terms? In considering this question, I would like to mention briefly Chisholm's attack on physicalism. • 153
(c) Chisholm*s primary attack was centered upon
showing that all attempts made so far have failed to
provide an adequate analysis of "belief sentences."
Since, according to Chisholm "belief" is a psychological
term, the logical behaviorist*s claim must be rejected.
Let us illustrate Chisholm’s reasoning by considering what he calls the "specific response" analysis. In this case,
let us consider an analysis of the belief sentence.
(1) Smith believes that there is a storm.
Chisholm tells us that according to the "specific « response" analysis, (1) is analyzed as
(2) Smith exhibits storm-responses to his
immediate environment.
But how is one to analyze the term "storm response" in
(2). A physicalist, for instance, might offer the following as a proposed translation. To say "Smith exhibits storm-responses" means
(3) Smith exhibits that specific behavior
which is exhibited only when there is
a storm.
In other words, "storm response" is defined in this exam ple in terms of what Smith does when and only when he is in a storm. Put differently, there is some response that 154
is manifested by Smith when and only when he is in the
presence of a storm. But Chisholm tells us that (3) will
not do, because it implies that Smith believes there is a
storm only when there really is a storm. However, Smith,
like most human beings, is inclined to believe in things
that are simply false. Of course, one cannot evade the
difficulty by saying "when and only when Smith thinks
that there is a storm" on account of the psychological
term "thinks."30
Another criticism of Chisholm centered upon the
thesis which claimed that we can dispense with psycho
logical language by coining certain technical terms. For
instance, some psychologists prefer to say
(4) The rat is food-expectant or has a food-
expectancy instead of
(5) The rat expects food.
But Chisholm tells us that such analysis will not do, because "if we wish to tell anyone what our technical
______t
30For a detailed explanation of this kind of analysis as well as "appropriate behavior" analysis and "verbal response analysis," see Perceiving. (Ithaca: Cornell Press, 1957), pp. 168-73. Also, "Sentences about Believing" in Minnesota Studies in the Philosophy_o£ Science. Vol. II, pp. 513-519. 155 terms mean, we must use intentional language again.
In other words, if we ask our psychologist what he means by "food-expectancy," he might reply: an animal is said to have "food-expectancy" if and only if the animal ex pects food. Yet if the explanation of certain technical terms make use of psychological terms, then analyses of psychological sentences using such technical terms should not be taken as cases to buttress physicalism.
In a nutshell, what Chisholm attempted to do is to show that behavioral analysis of psychological sentences is inadequate. Although we have not examined all possible behavioral analyses of psychological sentences, it seems very doubtful that any of the other proposed behavioral analyses will fare better. Now, I shall assess how ade quate physicalism is as a technique in concept formation in light of the criticisms leveled against it.
(d) As indicated earlier, Carnap abandoned his earlier physicalistic views. Recall that one of the earliest attempts provided by logical positivists to introduce psychological terms was by means of an explicit definition. This view, as we know by now, amounts to
Chisholm, "Sentences about Believing," in Minn. Studies in the Philosophy of Science. Vol. II, p. 513. 156 saying that all psychological terms are empirical terms, since they are explicitly defined by observation terms.
Consider the case of defining the concept of being hungry as treated in behavior theory. Let MHxn stand for "x is hungry"; flPx" stand for "x is presented food"; nEx" stand for "x eats the food." In symbolic notation, we have
(6) H(x) =2 (t) (P(x,t) => E(x,t) )
We are familiar with the difficulty encountered in
(1), if the definiens is construed in terms of a truth- functional "if-then." The next suggested move was to use the method of reduction sentences: Thus instead of (1), we have
(7) (x) (t) (P(x,t) => (H(x,t) = E(x,t) )
In (7), we are told that reduction sentence avoids the problem encountered in (6). In order to say "x is hun gry," the test-conditions must be fulfilled, i.e., x must be presented with food. If the test-conditions are not fulfilled, the meaning of the term "hungry" is left un specified. Yet in his later writings Carnap abandoned the view that psychological terms are introduced by reduction sentences. We are told that psychological concepts are introduced by the postulates of a theory and are connected with observation terms by means of correspondence rules. 157
In "On Some Concepts of Pragmatics,"^ Carnap hints at how one can introduce psychological terms by means of postu
lates and correspondence rules. He illustrates in terms of the following:
Let "A(x,t,s,L)" means "x at t wills deliberately to utter a token of s as a sentence of the language L in the
sense of an assertion." The concept A, according to
Carnap, involves purpose or intention; hence it is a
theoretical construct. On the other hand, the following
concept belongs to the observation language. Let
"U(x,t,r)" means "x at t produces with his speaking organs a series of audible sounds r." Furthermore, suppose that
r is a token of s. Then, according to Carnap, the senten ce "U(x,t,s)" contains no reference to L. Carnap goes on
to say that "the fact that the sounds s are meant by x as a sentence of L is not directly observed, but can at best be inferred inductively.. The rules of correspondence may
Carnap, "On Some Concepts of Pragmatics," Philosophical Studies. (Oct., 1955), pp. 89-91. 158
supply a connection between A and U."33
4. Summary and concluding note
I discussed in this chapter some of the salient
characteristics of Watsonian behaviorism and then con
trasted these ideas with logical positivism. It was
pointed out that whereas Watson*s views were attractive
to the logical positivists, logical positivists were
interested in the logical commitments underlying
Watsonian behaviorism. In this connection, we discussed
the physicalistic views of Carnap and Hempel and tried
to show how, according to the physicalistic thesis, psy
chological terms are analyzed. Recall that the early
methods of introducing psychological terms were abandoned
in favor of the.method of postulates. The assumption
underlying the method of postulates is that psychological
terms are partially and indirectly interpreted with
33camap has not elaborated much on the introduction of psychological terms by means of postulates and corres pondence rules. Yet his point was that just as in the physical sciences terms like "mass," "gravitational force," etc., are introduced by postulates without com plete interpretation, so are psychological terms like "anger," "anxiety." See Kyburg*s Phil, of Science (New York: The MacMillan Co. 1968), esp. Chapter II, for more detailed account of introducing psychological terms by means of postulates. 159
respect to the theory being employed. And whether one
accepts physicalism as an adequate technique in concept
formation depends on one's willingness to accept the
general physicalistic thesis. It was our contention that
physicalism is adequate only provided one explicates the
link between theoretical terms (psychological terms
included) and observational terms.
We have not said much about Bridgman in this chapter.
Yet an operationist like Bridgman could certainly agree with the logical positivists on many issues discussed in
this chapter. For instance, Bridgman could point out
that there is essentially little difference between the
physicalistic thesis and his own operationalism when
applied to psychology. Put in terms of Bridgman*s langu
age, the physicalistic thesis might read: The meaning
of any psychological sentence is only specified by means
of public, concrete operations. As Bridgman himself might say to Watson’s followers: You may study anything
you want in psychology, but any meaningful psychological
concept you incorporate in your science is specified only
by a set of operations.
While neither logical positivism nor.Bridgman had
any direct influence on Watson, Bridgman*s operationalism as well as logical positivism exerted a great deal of influence on the neo-behaviorists who followed Watson.
To this theme, we shall address ourselves in the following chapter. CHAPTER VI
TOLMAN, LOGICAL POSITIVISM, AND OPERATIONALISM
Introduction
At the conclusion of Chapter V, It was suggested
that it was very unlikely that logical positivism or
Bridgman had any direct influence on Watson. On the other hand, one can easily detect the logical positi
vists' influence or at least their spirit in the writ
ings of some of the so-called neo-behaviorists who
followed in the footsteps of Watson. Also one might
say that Bridgman's influence on neo-behaviorism has been substantial. In this connection, Spence and
Bergmann wrote an article entitled "Operationlsm and
Theory in Psychology" in which they claimed that in psychology, as in other sciences, psychologists have
stressed two different aspects of the scientific enter prise. As Spence and Bergmann put it in emphasizing
the two aspects of the scientific enterprise:
161 162
A number of psychologists such as Tolman, Skinner, Stevens stimulated by the writings of Bridgman have centered their efforts largely on the empirical component of scientific method. Under the watchword of operationism, they have carefully consid ered and laid down the requirements that scientific concepts meet in order to insure testability and thus empirical meaning. The second aspect, the formal (theoretical) component of scientific endeavor has been brought to the forefront in psychology principally through the writings of Hull and Lewin.l
Thus it would seem from this quotation that the main source which psychology in the 1930*s took as its model was the logical positivist-operationist view of science. Undoubtedly neo-behaviorism attempted to stabilize itself within the surroundings of the posi tivist-operationist model of science. One of the basic issues in the literature of neo-behaviorism centered around the introduction of terms for unobservables, such as "expectations," "beliefs," "desires," etc. Recall that Watson's main argument in rejecting traditional psychology stemmed from its preoccupation with conscious ness. And if psychology was to achieve the status of a
*G. Bergmann and K. Spence, "Operationism and Theory in Psychology," The Psychological Review. Vol. 48, (Jan. 1941), p. 1. 163 science, he urged, it must base its data upon findings like those of physical science rather than upon intro spective data. Thus, in Watson's writings, one comes across one of the earliest attempts to rid psychology of its speculative vestiges. Neo-behaviorism could also be viewed as a second attempt to cleanse psychology from its past and provide us with a more effective scientific methodology. One should also keep in mind Watson's empha sis on stimulus-response theory which asserted that all behavior is in response to stimuli, whether the behavior is explicit or implicit. Neo-behaviorists, however, go beyond Watson's behaviorism in their willingness to assume between the stimulus and response all sorts of states and processes which they called "intervening variables." Of special significance to our concern is the interpretation of intervening variables, for it seems evident that neo- behaviorism has here been methodologically influenced to a great extent by the logical positivist-operationist model of scientific explanation. As basis for my analy sis, I will concentrate on Tolman. Thus in what follows,
I shall (1) present an account of Tolman*s use of inter vening variables. (2) discuss the controversy of inter vening variables versus hypothetical constructs, and 164
(3) show the logical-positivist and Bridgmanian influ ences on Tolman.
1. Tolman on intervening variables.
Watson’s call for the extension of scientific methods into the field of behavior was welcomed by Tolman. In
"Psychology versus Immediate Experience," Tolman writes
I am a behaviorist. I hold that psychology does not seek descriptions and intercommuni cations concerning immediate experience per se. Psychology seeks, rather, the objectively statable laws and processes governing behavior.^
Thus Tolman, like Watson, emphasized that psychology should be treated as one among other scientific inquiries.
But, unlike Watson, Tolman thought that the processes associated with pure behavior are such that there are two levels of behaviorism. The level at which Tolman builds his theory has come to be. known as "the molar level of behaviorism," as when we observe a rat running in a maze.
The other level of behaviorism which Tolman attributed to
Watson is "the molecular or micro level of behaviorism."
According to Tolman, the molecular level is essentially
^E. C. Tolman, "Psychology Versus Immediate Experi ence," Philosophy of Science. (July, 1935). Reprinted in Behavior and Psychological Man. Essays in Motivation and Learning by E. C. Tolman (Berkeley: University of California Press, 1958), p. 94. 165
physiological. As he writes
Our task, as psychologists, is the col lecting and ordering of the molar behavior facts per se. And this task can, in large, be performed in relative ignorance of both physiology and neurology.3
Tolman1s emphasis on molar behaviorism has led him
to assert other principles in his theory. In addition to
the data supplied by observations of stimuli and respon
ses, Tolman asserted that there are certain "behavior
determinants" which enter as intervening variables between
five independent variables (environmental stimuli, physio
logical drive, heredity, past training, maturity or age)
and dependent variable behavior. In other words, Tolman*s
proposal roughly amounts to saying that between S, which
is the independent variable, and R, the dependent variable,
there are intervening variables. Throughout his career,
Tolman has emphasized the importance of the role of
intervening variables in psychological theory. But un
fortunately much of the underlying confusion in contem
porary behaviorism stemmed from a lack of clarity concer ning the interpretation of intervening variables. To this
^E. C. Tolman, Purposive Behavior in Animals and Men. (New York: Century, 1932), p. 416. 166 task, I shall now address myself by showing how the con ception of intervening variables was developed in Tolman' theory.
In his article "Psychology versus Immediate Experi ence," Tolman presented us with an early account of inter vening variables. Briefly, his account of intervening V variables emphasized the functional relation between the independent variables (which represent the observable antecedents of behavior) and the dependent variables
(which represent observable aspects of consequent behav ior) . In order to explicate the functional relation between the dependent and independent variables, Tolman presented the following equation scheme
B B fi (S,H,T,P)
"S" stands for "the environmental-stimuli;" "H" stands for "heredity," "T" stands for "past training," "P" stands for "physiological drive.According to Tolman, the independent variables, S, H, T, P are the initial causes for the final observable behavior (B). After realizing the complexity of the function f^ relating the
^Tolman, "Psychology versus Immediate Experience," p. 102. 167 independent and dependent variables, Tolman proceeded to break it down into simpler component functions of the forms:
B = f2 (Ia»Ib» — • V and
Ia = f§ (S,H,T,P)
Ib = (S.H.T.P)
Ic = f| (S,H,T,P)5
Ia , 1^,...In represent what Tolman called "intervening variables." The component functions begin by introducing new intervening variables which are defined in terms of the independent variables. Once the functions £2 and f^ have been determined and the proper intervening variables have been chosen, the resulting observable behavior (B), according to Tolman, is predictable.
So far, Tolman has not told us what he meant by
"intervening variables." The emphasis was centered upon exhibiting the functional relationship between the inde pendent variables and dependent variables. Later in his article "Psychology versus Immediate Experience," we are told that intervening variables comprise a set which he calls "behavior readinesses." He further illustrates
5Ibid. 168
his concept of intervening variables by explaining it in
terms of (1) the demand for a positive goal object (2)
the demand against a negative goal object and (3) the
demand for "short" distances to a positive or from a negative goal.** Underlying this technicality, I think what Tolman had in mind when he introduced intervening
variables was to consider them as dispositional properties of organisms. As Tolman explains it:
Under the heading demands for positive goal objects I conceive a set of vari ables (i.e., behavior readinesses) such that when anyone of them is operative it will act to produce a specific con- sumatory response, if the appropriate goal object be present. It will also operate, if this goal object be absent, to produce a persistent exploratory activity until an instance of this appropriate type of goal object has been found.7
In "Operational Behaviorism and Current Trends in
Psychology," Tolman was further concerned with the func tional dependencies of the dependent variables upon inde pendent variables and with operational ways of measuring
intervening variables. According to Tolman, "operational psychology will be one which seeks to define its concepts
6Ibid., p. 103
7Ibid. 169
In such a manner that they can be stated and tested in
terras of concrete repeatable operations by independent Q observers,1 In this article, Tolman viewed behavior as being adequately represented by the following equation
scheme
Included as intervening variables in "Operational
Behaviorism and Current Trends in Psychology" were terms
like "demands," "discriminanda," "manipulanda," "traits,"
"capacities," etc.-^ "Discriminanda," for example, stands
for "sensations" and "images," Thus although intervening
variables are still not directly observable but function
ally related to the dependent and independent variables,
they are still to be operationally defined in terms of
the actual operations where their relations to the inde
pendent variables and to the final dependent behavior are
determined. Hence Tolman's account of intervening
®E. C, Tolman, "Operational Behaviorism and Current Trends in Psychology," Proceedings of the 25th Anniversary Celebrating Inaug. Grad. Stud., (Los Angeles: University of Southern California Press). (1936). Reprinted in Behavior and Psychological M an. pp. 115-129.
9Ibid., p. 117. The new term "A" stands for "maturity" (or "age.").
10Ibid., p. 118. 170 variables in this article stressed that they must be securely anchored in the plane of observables before one can admit them in scientific psychology. As Tolman tells us in talking about intervening variables:
They are objective entities defined in terms of the f2 functions which connect them to the S's, P's, H's, T's and A*s on the one hand, and to the final B, on the other. 11 ■L
So far, we have noticed that the only meanings that
Tolman attributed to these intervening variables is pro vided by their functional relationships to both the inde pendent and dependent variables. In this way, they are
"shorthand" or "summary" devices which help the psycholo gist in his search for the empirical laws which relate behavior to the variables determining it. Of special significance also is the demand that these intervening variables be explicitly linked with observables, i.e., dependent and independent variables. Moreover, it should be kept in mind that, to Tolman, intervening variables are not to be interpreted in terms of any physiological explanations, but in terms of molar behavior. Tolman*s insistence on this functionally derived concept led to
11lbid. 171 the controversy surrounding the distinction between intervening variables and hypothetical constructs. The question was raised: is the intervening variable purely functional or is it capable of being defined in physio logical terms? To this question, X shall now address myself,
2. Intervening variables versus hypothetical constructs
As indicated earlier, for Tolman intervening vari ables serve as economical devices for relating the inde pendent variables to the dependent variables; they are
"short hand" descriptions and the only meaning they possess is their relationship to both the independent and dependent variables. Following Meehl and MacCorquodale's article "On a Distinction Between Hypothetical Constructs and Intervening Variables," a controversy has centered around the distinction between intervening variables and hypothetical constructs. On the one hand, we have those who considered the distinction fruitful and important in psychology; on the other, we find those who considered 172 the distinction to be trivial.^ In 1948, Meehl and
MacCorquodale distinguished between "intervening variables" and "hypothetical constructs" by defining
the former as "construct which merely abstract the empirical relationships" (this is Tolman1s original meaning of "intervening variables") and the latter as
"those constructs which are hypothetical* (i.e.,
involve the supposition of entities or processes not among the observed). As they put it in another way:
...we propose that the term 'hypothetical construct* be used to designate theoretical concepts which do not meet the requirements for intervening variables in the strict sense. That is to say, these constructs involve terms which are not wholly reducible to empirical terms; they refer to processes or entities that are not directly observed (although they need not be in principle
For the former views, see P. E. Meehl and K. MacCorquodale, "On a Distinction Between Hypothetical Constructs and Intervening Variables," Psychological Rev., Vol. 17, (1950), pp. 1-40. For the latter view, see G. Bergmann, "Theoretical Psychology," Annual Review of Psy.. Vol. 4, (1953), pp. 435-458.
■^Meehl and MacCorquodale, "On a Distinction Between Hypothetical Constructs and Intervening Variables," p. 610. 173
unobservable); the mathematical expression of them cannot be formed simply by a suit able grouping of terms in a direct empirical equation.^
This last quotation makes it clear what Meehl and
MacCorquodale intended. While the referents of certain theoretical terms possess the characteristics of inter vening variables in the manner Tolman suggested, there is still a class of theoretical terms whose meaning goes beyond the functional relationship established between the antecedent (independent) variables, and consequent
(dependent) variables. This latter cluster of theoretical terms was what Meehl and MacCorquodale called "hypotheti cal constructs." Thus, in the case of hypothetical con structs as distinguished from intervening variables, one is told that the latter are mere abbreviations; they are in principle eliminable and thus are only convenient
"summaries of thought" and have no "excess meaning." On the other hand, hypothetical constructs are not mere abbreviations but imply some physiological mechanisms; their referents exist in the organisms in a way in which those of intervening variables do not. Put differently, the proponents of the distinction between hypothetical
14Ibid., p. 606. 174 constructs and intervening variables contend that hypo thetical constructs carry surplus meaning. They involve, as Feigl would contend, an "existential hypothesis"^ about a molecular state which may be either directly or indirectly confirmed. It is in this sense that concepts in psychology, as used by the proponents of hypothetical constructs, are considered to have "excess" or "surplus meaning."
The whole point behind the distinction between hypo thetical constructs and intervening variables is now obvious: it behooves psychologists to concentrate on physiological psychology rather than on pure behavior.
Recall that the level at which Tolman constructs his system is the molar level. Yet Tolman was willing to admit that it is possible to construct a behavioristic theory couched in terms of physiology. This level was called "molecular" or "microscopic;" according to Tolman, it "attempts a complete neural, glandular, and visceral picture."^ This kind of behaviorism is rejected by
^Feigl, "Existential Hypothesis," pp. 35-60. '
- ^ T o l m a n , "Psychology versus Immediate Experience," p . 1 0 1 . 175
Tolman. He prefers to leave it to the physiologists.
It seems that the prime factor behind Tolman*s rejection of physiological psychology is the relative inacces
sibility of physiological states and processes.
Let us now sketch the alleged differences between
intervening variables and hypothetical constructs and evaluate the claims made by MacCorquodale and Meehl.
(a) Excess meaning
The claim that only hypothetical constructs have excess meaning is countered by simply pointing out that
intervening variables likewise have it. Hull’s mathe- matico-deductive theory of learning, for example, invokes
intervening variables (like habit strength, inhibitory potential, excitatory potential) in the formulation of its fundamental laws of simple classical and instrumental conditioning. In addition, Hull has sketched and employed composition laws or rules which, like the vector addition of forces in Newtonian mechanics, enables one to derive a large number of laws for many kinds of complex situations.
(Two of these are Hull's rule governing the "summation" of habits and the law stating what happens if incompatible response tendencies occur.) Any direct or phenomenological 1 7 6 theory (a theory not requiring a micro-model) thus always carries excess meaning in the sense that it predicts laws unlike those upon which it is based. In Hull's theory, predictions have been attempted for the complex situations of discrimination and maze learning. And hence it is obvious that as soon as his intervening variables are put to use they automatically acquire excess meaning.^7
So much, then, for the claim that one can dis tinguish hypothetical constructs from intervening vari ables by appeal to excess meaning. Both kinds of vari ables possess it.
(b) Referential status of abstract terms
Advocates of hypothetical constructs argue that intervening variables, unless physiologically "located" and hence "physically existing," lie somewhere in a con ceptual No Man's Land bn whose shifting shores we should not embark. This claim can be met head-on by clarifying the status of the referents of abstract, or defined theoretical, terms in science.
» l-7Bergmann, "Theoretical Psychology," pp. 445-446, from which I have borrowed heavily. See also, Spence, "Types of Constructs in Psychology," III, 4. 177
This penchant for the literal "localization" of properties named by defined predicates is clearly mis guided. As Bergmann puts it:
Only individual physical objects, stones, trees, animals, nerve fibers, occupy literally a place in space. Their properties and the rela tions that obtain among them do not. This is as true of characters which we immediately ob serve as of the more "abstract" ones. Logically, there is no difference in this respect between colors and, say, boiling points or friction coef ficients. A colored object is in space; looking at it, one "sees" its color directly and thus comes to think and speak, loosely, of colors being in space. The mistaken belief that only what is thus "placed" or "concrete" is "objec tive" or "real" has been the source of much bad philosophy.... Most abstract or theoretical terms...are either explicitly relational, like 'friction coefficient,' or they name, like 'boil ing point,' relational properties, that is, characters which, though they are attributed to a single object, involve, as may be seen from their definitions, relations to others. I can see no difference in all these respects between the boiling point of a certain piece of wax and the strength of a certain habit£one of Hull's* intervening variables^, as the behavior theor ists define it, in a certain rat.^®
Thus it is methodologically unsound and misleading to speak of either simple or relational properties of things or states as either literally localizable or physically existing. The neutral term "exemplification" at least
Bergmann, "Theoretical Psychology," p. 443. verbally avoids this pitfall, A defined predicate that contains nonvacuously at least one that is unde fined names a character, an instance of which is exempli fied, (Contrast this formulation with the confused statement: The referents of defined predicates do or do not exist, or are or are not in space.) We should like wise urge the methodologist to speak of relational proper ties or characters as being exemplified (or not), but if so, by more than one physical object. That there instances are not located in space does not mean that the relational properties are any less "objective" or (physically) "real,"
It is, therefore, merely verbal and conceptual confusion to criticize proponents of intervening variables simply because the intended referents of (say) "discriminanda" or "habit strength" are mental or molar, exemplifying either instances of mental properties (states or traits) or instances of relational properties of animal bodies
(acts or behavior patterns).
This love for localization and penchant for exis tential hypotheses has thus misled some methodologists.
Proponents of intervening variables need not fear their conceptual No Man*s Land where the referents of abstract terms are not "concrete" or "localizable." 179
(C) Abstract terms in physics and psychology
Advocates of hypothetical constructs have argued that
these constructs can be introduced and partially inter preted, like some of the basic terms of an axiomatic cal
culus of the kinetic theory of matter, by correspondence postulates. As Meehl and MacCorquodale explain it:
Once having set up sentences (postulates) containing these hypothetical words, we can arrive by deduction at empirical sentences which can themselves be tested. But the words themselves are not defined directly by or reducible to those empirical facts.^
One can attack Meehl and MacCorquodale's claim that hypothetical constructs are postulated entities by pointing out that there is a considerable difference between the microterms used by physicists and the terms used by physiologists. That is to say, a protagonist of intervening variable theory might agree that postulates are needed in theoretical micro-physics but not in theo retical physiological psychology. According to this cri ticism, hypothetical constructs such as "anxiety" or
"expectancy" in psychology are not postulated theoretical entities. They function quite differently from
^Meehl and MacCorquodale, "On a Distinction Between Hypothetical Constructs and Intervening Variables," p. 598. 180
theoretical terms such as "molecule," "neutron." An answer
to this attack might simply consist in pointing out that
terms like "anxiety," "expectancy" are surely terms that are frequently used by psychology in explaining human behavior. They are obviously not observation terms like
"red," "green." Hence they are theoretical terms and could be introduced by postulates.
Thus I think that the claim that physics differs radically from psychology— postulates are needed in theo retical physics but not in theoretical psychology--is not at the heart of the issue. Rather the point I would like to bring out is that intervening variables can also receive their meaning by means of postulates.20
After all this, perhaps, one should take Bergmann's position that "the whole controversy of intervening variables versus hypothetical constructs is a pseudo- 21 issue, the distinction itself is a pseudo-distinction."
While one appreciates Bergmann*s point, X think one can still maintain the distinction. When, for examples
vn See C. L. Hull's book, Essentials of Behavior. (New Haven: Yale Univ. Press, 1951), passim.
^Bergmann, "Theoretical Psychology," p. 44. 181
Tolman first introduced intervening variables, he
stressed that they provide natural definientia for both mental states (sensations, images, conscious demands or
drives) and mental traits or capacities (demand capacities,
discrimination capacities, personality traits). The
traits and capacities are, he argued,^the relatively
direct product of heredity, training, and maturity, while
the mental events are products of environmental stimuli
and physiological drive combined with the traits and capacities. I suggest that we thus distinguish between mental variables and physiological variables rather than between intervening variables and hypothetical constructs.
Those, like Tolman, who stressed the need in psychology for "mental variables," were simply pointing out that, in attempting to specify the mathematical form of the empirical laws relating the independent environ mental and individual-difference variables to the depen dent response variables, one had to break down the initial functional relations into successive sets of component functions. The constructed "mental variables" thus mediated between the several independent variables and the be havioral response that the laws were designed to predict.
In this context of discovery, their oblique referents were 182
mental events and capacities but, because of the under
lying psycho-physiological parallelism which behaviorism
espouses as a factual thesis, "mental variables" were
assumed to be eventually eliminable. The early Tolman,
for example, can be interpreted as programmatically hoping
that, however necessary "mental variables" are in the
actual discovery of the form of laws of behavior, they would eventually be replaced by molar behavioral variables
once the laws were found. Proponents of "physiological
variables," on the other hand, can be interpreted as hoping
that all such mediating variables would eventually find
their locus in physiological states and processes, and as believing--in keeping with their physiological reduction-
ism— that all talk about mental states and capacities would ultimately be replaced by talk about their physiological 99 counterparts. ^
However, unimportant my distinction between "mental
variables" and "physiological variables" might appear in
the context of the justification or reconstruction of be haviorism, I think it might be helpful in characterizing
^For extended treatment of this conception of re placement, see Hinshaw, "Science versus Ontology," Philosophy of Science (forthcoming). 183 different approaches in the actual development of concept and theory formation in the context of discovery for psychology. Hence while all behaviorists abstractly affirm that future behavior is in principle a function of present and past behavior, physiology, and environment, some (like Tolman) have stressed a molar behavioral approach which had to invoke "mental variables" both to detect the form of behavioral laws and to avoid unwitting appeal to "action over temporal distance," while others have stressed a molecular approach by invoking "physio logical variables"— thus hinting at ultimate physiological reduction. My distinction, X contend, makes it far easier for the student or historian of psychological science to point out these different approaches with their respective emphases•
With all this discussion of intervening variables and hypothetical constructs behind us, I would now like to turn to some of the influences of the logical positivist- operationalist methodological views on Tolman.
3. Tolman, operationalism and logical positivism
It is my contention that neo-behaviorists were influenced by certain methodological views put forth by logical positivism and Bridgman. Tolman is no exception, 184 for he definitely attempted to stabilize his writings within the ambiance of the logical positivist-operationist model of science. I shall now consider a representative sample of this dual influence on Tolman.
Tolman, at one time, called his psychology "opera tional behaviorism" in order to give credit to the logical positivists and to Bridgman for the term "operational."
Tolman's admission of this dual influence is put in the following manner:
. . . the term "operational" has been chosen with two different meanings in mind. In the first place, I have chosen it to indi cate a certain positivistic attitude now being taken by many modern physicists and philosophers for which Professor Bridgman has selected this word "operational" . . . But in the second place, because of what seems.to me a second connotation which in connection with the word "behavior" it tends to have. For behavior is the thing observed also turns out to be essentially an activity whereby the organism in question "operates."^3
At the end of the article "Operational Behaviorism and
Trends in Psychology," Tolman recapitulates one of his central points by saying:
23T0 iman, "Operational Behaviorism and Current Trends in Psychology," p. 115. 185
...these intervening variables are to be defined wholly operationally--that is, in terms of the actual experimental operations whereby their presences or absences and their relations to the controlling independent variables and to the final dependent behavior are determined.24
The above two quotations serve to indicate that two
of the central goals which Tolman attempted to attain
were (a) finding an adequate observation in order to
securely anchor psychological statements in his theory and
(b) finding ways of introducing intervening variables in
theoretical psychology. It is my contention that with
respect to both goals, Tolman patterned his views after
those of the logical positivists and Bridgman.
A. Observation base.
Concerning the observation base for psychological
statements, Tolman emphasized that the independent vari
ables as well as dependent variables be explicitly linked
to observables. That is to say, psychological statements
containing dependent variables or independent variables must be formulated such that their referents are publicly
observable data; otherwise they are unfit for behavioristic
psychology. An example of an independent variable is,
24Ibid., p. 129. 186 of course, the notion of stimulus. In this connection, it is demanded that stimulus variables be explicitly linked to the observation language of physical science.
Thus independent variables in Tolman1s scheme are factors amenable to measurement or empirical events which serve as the operationally identifiable antecedent condition of the behavioral events which the theory is supposed to predict. In his presidential address to the American
Psychological Association, Tolman listed the following independent variables.
I. Environmental Variables
M: Maintenance schedule
G: Appropriateness of the goal object
S: Types and modes of stimuli provided
R: Types of motor response required
^(0B0): Cumulative nature and number of trials
P: Pattern of preceding and succeeding maze units.
II. Individual Difference Variables
H: Heredity
A: Age
T: Previous training 187
E: Special endocrine, drug or vitamin conditions.25
Dependent variables are terms which refer to classes
of events that the theory is designed to predict. In
Tolmanrs scheme, the consequent condition which is measured in an experiment is some kind of behavior.
Examples of dependent variables are tendencies, traits,
and skills. According to Tolman, "the final dependent
behavior has three component aspects... a) direction
b) quantity or persistence and c) efficiency or skill.
Thus Tolman*s attempt to insure that his independent
and dependent variables are "directly observable events"
appears to be motivated by the logical positivists* meaning criterion and Bridgman*s operationalism. We noticed
Tolman suggested that his independent and dependent
variables be linked to "observables" in order to satisfy
the requirement of an adequate observation base. Implicit
in this discussion is Tolman*s commitment to the empiricist
criterion of meaningfulness. Although Tolman has not
25 Tolman, "The Determiners of Behavior at a Choice Point.** Psychological Reviewr Vol. 45, (January, 1938), Reprinted in Behavior and Psychological Man, p. 149.
^Tolman, ’’Operational Behaviorism and Current Trends in Psychology," p. 119. 188 specifically asserted which of the various meaning criteria he was appealing to, one can still detect the general tenets of logical positivism in his writings, especially the commitment to the view that cognitive meaningfulness is predicated only of sentences that are at least in principle verifiable. Hence all significant statements of psychology containing dependent as well as independent variables must at least meet this minimal requirement of the meaning criterion. If they do not, they are then to be excluded from behavioristic psychology.
■ I think, then, that Tolman in his search for an adequate legitimate observation base for psychology did adopt a verifiability criterion of meaning. Which one?
I am not sure. There is no specific evidence in his writings regarding his commitment to any version of the meaning criteria put forth by the logical positivists.
Hence one is not in a position to marshal against Tolman the various criticisms associated with the empiricist criterion of meaningfulness. However, should one attri bute a verifiability meaning criterion to Tolman in the manner suggested by logical positivists, one should then, as indicated in Chapter II, answer adequately the various criticisms leveled at the verifiability criterion. But 189
be that as it may, 1 believe that Tolman was not inter
ested in propounding a theory of meaning. In this
respect, he reminds us of Bridgman who (as indicated in
Chapter XV) did not intend his brand of operationalism to
be construed as a theory of meaning. What Tolman is
committed to is the minimal requirement demanded of any
scientific theory, namely, that for something to be worthy
of scientific consideration, it should at least be in
principle verifiable or confirmable. And if psychology
is to be granted the status of science, as Tolman wants
it to be, then it must satisfy the verifiability require ment.
In addition to the verifiability requirement, we must also consider the "operational requirement" as pro
pounded by Bridgman. Recall that Tolman, at one time,
described his brand of psychology as "operational be
haviorism," By that he meant "one which seeks to define
its concepts in such a manner that it can be stated aid
tested in terms of concrete repeatable operations by
independent observers.1,27 He goes on to add that "the
behaviorism which I am going to present seeks, then, to
27Ibid. use only concepts which are capable of such concrete
operational verification."28 So far this accords well with Bridgman's famous principle: "the concept is
synonymous with the corresponding set of operations."
This principle of Bridgman is, as we indicated in Chapter
IV, restrictive. Too much is demanded by the operational principle when it requires us to recognize that a concept has meaning only if it stands for definite, concrete operations performed by human beings.^ What this opera
tional principle should require is not that all concepts,
say, in psychology should stand for precise physical operations. Rather, the operational principle should make methodologically oriented psychologists aware of the need for asking that the observation base of any concept be made as explicit as possible. This much should at least be emphasized. Otherwise, if this is not done, then serious doubts would be cast on the enterprise carred out by psychologists, like Tolman, who insist that the method adopted in behaviorism is parallel to the one in physical science. I take it this to be what Tolman should have
28Ibid.
2^See Chapter IV, Section 1. 191
stressed, i.e., that the observational bases of any concepts be made as explicit as possible. Whether this
is what he intended to do, I am not sure. At any rate we can at least detect the dual influences of Bridgman and logical positivism on Tolman in his search for an observation base for psychological statements containing dependent and independent variables. With respect to the dual influence, Tolman considered both the verifi ability criterion and the operational principle indispen
sable tools for rendering psychology scientific. And it certainly provided Tolman with a frame of reference which enabled him to gain penetrating insights into many theoretical problems in psychology.
B. Ways of introducing intervening variables.
With respect to the other issue, namely, Tolman*s deep concern with the various ways of introducing intervening variables, logical positivism and operation al ism again provided Tolman with a frame of reference.
Of special significance is Tolman*s use of "operational definition." This comes to fore when Tolman discusses the way (s) of introducing intervening variables into a
theory. To such matters we now turn our attention.
In his earlier writings, Tolman used "defining 192 experiment" instead of "operational definition." However, in his later writings, Tolman explicitly uses the tern
"operational definition." Hence Tolman, like Bridgman, appears to be using the term "operational definition" in at least two ways. It was indicated In Chapter IV that Bridgman has used "operational definition" in two * OA ways. w First, it was used to refer to a sentence con necting a term to be defined with other terms describing laboratory procedures and manipulations. This use is comparable to what Tolman referred to as "defining experiment." As he puts it
...each "intervening variable" is defined by a standard experiment In which its correlative independent environmental variable is systematically varied.31
The intervening variable "demand," for example is explained by providing a "defining experiment" in terms of a graph where Tolman kept all independent variables relevant to the food demand constant except M I.e.,
(maintenance schedule). If we repeat this experiment several times, we obtain a curve (function) between
"demand" and "M".
^See Chapter IV, Section IV.
^Tolman, "The Determiners of Behavior at a Choice Point," p. 157. 193
According to Tolman, the curve (function) serves as
the definition for the intervening variable "demand."
Tolman also thought that this procedure for defining
"demand" could be used to define other intervening vari
ables. As he put it: "But this procedure, which I have
thus outlined in some detail for demand, could also be * used in analogous fashion for defining each of the
other intervening variables.Thus the definitions in
terms of the experimental operations needed to indicate
the functions (curves) of these intervening variables would
constitute their operational definitions.
Second, the term "operational definition" was used
by Bridgman to refer to mental operations such as
pencil-and-paper operations, verbal operations. Hempel
referred to such operations as "symbolic operations."
Concerning this second use of "operational definition,"
one would have to depend on the logical positivists and
not on Bridgman— as I indicated on Chapter IV; in this
connection, Tolman*s second use of "operational definition"
is explicitly exemplified in the article "Studies in
Spatial Learning I. Orientation and the Short Cut."
Tolman*s definition of "expectancy of food at
32ibid.. p. 159. 194 location L" provides us with an illustration. The first definition provided by Tolman was:
When we assert that a rat expects food at L, what we assert is that if (1) he is deprived of food, (2) he has been trained on path P, (3) he is now put on path P, (4) path P is now blocked, and (5) there are other paths which lead , away from path P, one of which points directly to location L, theh he will run down the path which points directly to location L. When we assert that he does not expect food at location L, what we assert is that, under the same conditions, he will not run down the path which points directly to location L.33
This definition of "expectancy of food at location
L" is an instance of an explicit definition of a theo retical term. One can symbolize this definition of
"expectancy of food at L" by letting "Q3 (x)" stands for
"rat x expects food at location L;" MQ^(x,t)" stands for
"the conjunction of the conditions (1), (2), (3), (4),
( 5 ) i.e., as holding for rat x at the time t;
"Q2 (x,t) stands for "x runs down the path which points directly to location L." In symbolic notation, we have the following Carnapian* rendition:
^Tolman et al., "Studies in Spatial Learning I. Orientation and the Short-Cut," Journal of Experimental Psychology, Vol. 3, (1946), p. 14. 195
(1) Q3 (x) S (t) (Q1(x,t)=>Q2(x,t) )
However, as Carnap has shown in "Testability and Meaning,"
the explicit definition view of theoretical (dispositions)
terras is too restrictive. Tolman, aware of the diffi culties connected with the explicit definition of inter
vening variables, offered the following definition as an * improvement over (1)
If x is deprived of food and x has been trained on path P and x is now put on path P and path P is now blocked and there are other paths which lead away from path P, one of which points directly to location L, then x runs down the path which points dir ectly to location L x expects food at location L.34-
In symbolic notation, it is rendered as
(2) (x) (t) (Q-j/x.t) » (Q3 (x ) = Q 2(x,t) ) )
(A direct transcription of the right hand parenthesis yields "(Q2 (x,t)==Q3(x) ).” But since this is analyti cally equivalent to "(Q3(x) = Q2(x,t) ) ," I have trans cribed it in this manner to parallel a bilateral reduc
tion sentence.)
Thus (2) provides us a meaning of only if the antecedent conditions "Q^" is fulfilled. In terms of the
illustration provided by Tolman, (2) would specify the
34Ibid., p. 15. 196 the meaning of "x expects food" only for rats that have met the antecedent conditions (1) - (5). If the ante cedent conditions are not fulfilled, the meaning of the
"x expects food at location L" is left undetermined. The advantage of (2) over (1) is that it does not completely limit the meaning of the terms "Qo" once and for all, but J * permits room for expanding the meaning on the basis of accumulated scientific knowledge. It also avoids the difficulty of the vacuous truth of in explicit definition. Hence reduction sentences guarantee that certain scientific concepts (psychological terms included) are only partially determined; their meaning Is left
"open."
The foregoing discussion clearly indicates Carnap's influence on Tolman. The second sense of the operational definition that Tolman made use of in "Studies in Spatial
Learning" is a bilateral reduction sentence discussed in
Carnap's "Testability and Meaning." At this state, it would be worthwhile to pause and consider some criticisms levelled at Tolman*s use of the term "operational definition." 197
C. Criticisms of Tolman1s use of operational definitions
One critic^-* of Tolman1 s use of "defining experiment" offered the following objection. If one were to ask
Tolman why an experiment is needed in order to define intervening variables, he would most likely tell us that his use of "definition" parallels the use of a physicist * in, for example, attempting to define temperature by the expansion of the column of mercury, where the appropriate use of a mercury thermometer is sufficient to determine the temperature of a body. The critic continues by claim ing that if this "defining experiment" is to be an experi mental question, there must be then some independent ways of verifying statements about defined variables. With respect to temperature, one can provide a rough way of verifying by "sensation" temperature differences or euqalities. But with respect to such independent tests in the case of "demand," one is faced with a difficulty.
We cannot use independent tests as (in the) temperature case for, as the critic puts it, "there is no such thing as a sense Impression of a rat's demand.
35 A. Pap, An Introduction to the Philosophy of Science. (New York: The Free Press, 1962), p. 391
36Ibid., p. 392. 198
Tolxnan was aware of the above difficulty. He attempted to answer it by admitting that the functional relations between the intervening variables and the other variables is arrived at by sheer guess. As he puts it:
...X in my future work intend to go ahead imagining how, if I were a rat, X would behave as a result of such ...on the basis of such imaginings, I shall try to figure out some sort of f^ rules or equations.^7
Another criticism relating to Tolman*s definition of "expectancy of food at location L" came from Meehl and MacCorquodale,who thought that Tolman*s definition failed to specify "what else may have been in the rat*s history and what restrictions are placed on path P*s
stimulus properties in contrast to other paths." Con
sider an example which they cited. Xf path P is white and blocked, and there are other non-white paths which
lead to location L, then, if the rat runs "the path which leads to location L" is this evidence that the
S^Tolraan, "The Determiners of Behavior at a Choice Point," p. 164..
38w. k . Estes, et al., Modern Learning Theory, (New York: Appleton-Century-Crofts, Inc., 1954), p. 180. 199
"rat expects food at location L?" Again, what if there were a cat along path P? Are we going to say that the
retreat of the rat was a sign that he did not expect
food at location L?
I take it that the point that Meehl and MacCorquodale were stressing was that these other conditions should be
stated before one can concede the truth of the sentence
"x expects food at location L." And since'the conditions
they mentioned will break down the test provided by
Tolman, one must then not consider Tolman*s definition as providing us with an adequate meaning of "the rat expects
food at location L." As they explain it:
It seems, for example, unlikely that any reduction-sentences for such constructs as "expectancy," "habit," or "hunger" will in the long run lead to a predictively powerful theoretical structure if they are framed so as to ignore such nuisance vari ables as "fear," "fatigue," or "strength of competing exploratory response R^."^
Be all that as it may, I think Meehl and MacCorquodale missed the point about the intent of reduction sentences,
Carnap has pointed out that many of the definitions that
scientists use to connect theoretical and observable terms
39ibid., p. 233 200
are incomplete: they may, if you like, be called "partial
definitions." Hence operational definitions which can be
viewed as a kind of reduction sentences are in this sense
incomplete: whenever they introduce any theoretical term,
they do not completely specify or limit the meaning of the
term defined, once and for all. They allow room for %* expanding the meaning of the term pending further data.
Hence Tolman*s definition of "expectancy of food at loca
tion L" is to be viewed only as a "partial definition."
It holds conditionally, that is to say, provided the test
conditions stipulated are fulfilled; otherwise, the meaning is left undetermined. I agree that Tolman*s
example presupposes other "test conditions.'* But these other test conditions could be given by including more
reduction sentences. And, thus, I think Tolman's defi- nition avoids the criticisms put forth by Meehl and
MacCorquodale.
D. Review of the ways Tolman introduced intervening
variables.
At this stage, it is perhaps worthwhile to pause and consider the different ways Tolman attempted to introduce intervening variables. First, one of the earliest ways of introducing intervening variables was 201 to say that they are functionally related to the inde pendent variables and to the dependent variable. As
Tolman puts it in speaking of intervening variables:
...the sole cash-value of mental processes lies, I shall assert, in this their character as a set of intermediating functional pro cesses which interconnect between the initiating causes of behavior, on the one hand, and the final resulting behavior itself, on the other. ^
Remember that for Tolman the initiating causes of behavior are what he called on other occasion "independent variables" and the final resulting behavior is what he called "the dependent variable." Also keep in mind that these two variables are "operationally defined." Here we have in Tolman1s scheme the view that intervening variables are to be anchored to observables on both sides when they are employed in any scientific theory. This view amounts to claiming that intervening concepts should be closely tied to observables if one is to insure a maxi mum of empirical significance. This view, incidentally, led Tolman to say that significant concepts in psychology must be operationally defined. Though Tolman did not bother in his earlier writings to articulate the logical
40Tolman, "Operational Behaviorism and Current Trends in Psychology," p. 117. 202 link between theoretical terms and observational terms, we noticed, however, that in his article "Studies in
Spatial Learning" how logically oriented his discussion appeared in introducing the intervening variable "expec tancy of food at location L." His first attempt in explaining "expectancy of food at location L" reminded us to of the earliest attempts provided by logical positivists, where it was thought that theoretical terms can be explicitly defined by means of observation terms. By the same token, Tolman1s intervening variables are viewed as explicitly defined by means of observation terms. Put in this manner, intervening variables become meaningful only as a symbolic construct or as a functional shorthand expression of observable terms.
Second: The other method by which Tolman attempted
to introduce intervening variables was in effect by means of Carnap’s bilateral reduction sentences. It seems clear
that though intervening variables are still tied to the plane of observable, their meanings are only partially specified; their meanings are left "open."
Third: Tolman throughout his career did not intro duce intervening variables by means of postulates. His failure to develop anything like a logically integrated 203 theory became one of the criticisms of Tolman. Paren thetically, many psychologists expected Tolman to formal ize his theory like Hull, another leading behaviorist.
But Tolman realized this defect when he later wrote: "I myself can neither get very interested in nor completely understand such more refined logical distinctions."^*
It is important to note that although his intervening variables were securely anchored to the observable vari ables, i.e., dependent and independent variables, still
Tolman did not think it possible to replace his inter vening variables by observational terms. In "Operational
Behaviorism," he gives one reason why Intervening vari ables are not replaceable by observational terms. As mentioned in Section I, above, the original mathematical function required to connect the independent and dependent variables in Tolman*s scheme was broken down into succes sive sets of simple component functions. The reason
Tolman broke the original function into a simple component functions was due to the complexity of the mathematical function. As he puts it:
^E. C. Tolman, "Principles of Purposive Behavior." In Psychology: A Study of a Science, ed. S. Koch, II, (New York: McGraw-Hill, 1959), p. 149. 204
It Is in fact so complicated that we at present seem unable to state it in any single simple statement. We find, rather, that we have to handle it by conceiving it as broken into successive sets of component functions.^
Hence Tolman, like the logical positivists, consid ered it unfeasible to dispense with theoretical terms.
They both thought that theoretical terms are indispensable for the fabric of science. Tolman*s underlying reason for the indispensability of intervening variables was based on practical factors. Hempel, in his discussion of Tolman, tells us
This argument, then, attributes to the introduction of unobservable theoretical entities an important practical role in the context of discovering inter dependencies among observables... ^
However, Hempel*s reason for the indispensability of theoretical terms in science rested primarily, as we indicated in Chapter III, Section 1, on the assumption that a scientific theory, among other things, has the task of establishing not just deductive relations but also inductive relations in the explanation and prediction of empirical findings.
^Tolman, "Operational Behaviorism," p. 117.
^Hempel, "The Theoretician's Dilemma," p. 203. 205
E. Concluding note and summary.
The preceding discussion clearly indicates that
Tolman, a leading exponent of neo-behaviorism, has been
influenced directly or indirectly by the scientific move ment of operationalism and the philosophic movement of
logical positivism. Undoubtedly this dual influence has * spread to other schools of thought.^
Before concluding this chapter, I would like to
^Recently, for example, there has been an interest to reformulate the main tenets of psychoanalysis in opera tional language. Mr. A. Ellis in his article "An Opera tional Reformulation of Some of the Basis of Psychoanaly sis" in Minnesota Studies. Vol. I, pp. 131-134., attempted to show how one can translate all statements containing psychoanalytic terms into statements which make use of ob servational terms. I shall not attempt to show this, but simply state the principle underlying this kind of analy sis. Ellis* operational criterion is aptly summarized as "This then is modern operationalism or empiricism. As applied to psychoanalysis, it means that psychoanalytic principles should be stated in terms so that they are in some final analysis in principle confirmable in terms of some ultimate observables." (see p. 133). Hence one can obviously detect the spread of the logical positivist- operationist methodological views to psychoanalysis. As Bergmann has noted: "To-day every scientific psychology is practically behavioristic and in its methodological dis cussion psychology seems to have reached a complete under standing with logical positivism." See Bergmann "On Some Methodological Problems of Psychology," Phil, of Science. Vol. 7, (1940). Reprinted in Readings in the Phil, of Science, Feigl and Brodbeck (eds.), p. 628. On another occasion, Bergmann emphasized that "No body of empirical knowledge can be built up without operational definitions of the terms in use." See Bergmann and Spence, "Opera tionalism and Theory in Psychology," p. 2* 206 mention two points in connection with Tolman.
First, as we indicated earlier in our discussion,
throughout his career Tolman has insisted that the level
at which his behaviorism is constructed is the molar
level. Tolman went on to introduce his intervening vari
ables in terms of macroconcepts. Another form of behav- * iorism admitted by Tolman, in addition to the molar level, was what he called "molecular" or "microscopic" psychology which is described as "essentially physiological."
Tolman, however, preferred to leave this latter level to
physiologists. After Meehl and MacCorquodale*s paper,
Tolman changed his mind. In 1949, Tolman agrees with
Meehl and MacCorquodale, Hence he is willing to accept
their claim that intervening variables can be couched in
terms of physiological terms. In this sense we are then
dealing with a hypothetical construct of some kind. That
is to say, intervening variables are now "loosely defined
hypothetical constructs."^
Second, Tolman in 1959 re-evaluated his original
thoughts about intervening variables. He tells us that
the intervening variables are derived "from intuition,
^Tolman, "Discussion," Journal of Personality, Vol. 18, (1949), pp. 48-50. 207 common sense, and phenomenology."4^ It seems in this connection that Tolman now wants to deny his earlier claims that intervening variables are linked to the em pirically dependent and independent variables. That is to say, he doubts whether these intervening variables could be solely defined in terms of manifest behavior.
Rather, they are as he puts it:
...merely an aid to thinking ("my thinking" if you will). All anyone really sees are the empirically stipulated independent and dependent variables. In developing notions of what happens in between...all I really am doing is setting up a tentative logic (or psychologic of my own) for predicting what the dependent behavior should be and how it should be affected by variations in such and such sets of independent variables.47
This quotation clearly indicates Tolman's rejection of his earlier views that "defining experiments" could adequately introduce intervening variables. Put in this manner, it seems that operationally defined intervening variables are not adequate.
It Is interesting to note that even among logical positivists there is also a new appreciation of
^Tolman, "Principles of Purposive Behavior," p. 98.
47Ibid., p. 148. 208
introspection. Carnap in "The Methodological Character of
Concepts" tells us
Although many of the alleged results of introspection were indeed questionable, a person*s awareness of his own state of imagining, feeling, etc., must be recognized as a kind of observation, in principle not different from external observation, and therefore as a legitimate source of knowledge, though limited by its subjective character.^8
Even Bridgman in The Wav Things Are (1959) has assigned an importance to introspection and value to the introspec- tional report. The following quotations reveal Bridgman*s position.
Insistence on the use of the first person either explicitly or implicitly, will inevitably focus attention on the indi vidual. . .Neglect of the role of the individual, with resulting overemphasis on the social, may well be one of the fundamental difficulties in the way the human race handles it mind. ^
i Another word for which I believe the private aspect is much more important than ordinarily realized is "proof** ...Here I shall only reiterate ray opin ion that a proper appreciation of this will alter the common picture of science
^Carnap, "The Methodological Character of Concepts," pp. 70-71.
^P. W. Bridgman, The way Things Are. (Cambridge: Harvard University Press, 1959), p. 5. 2 0 9
as something essentially public into something essentially private.
Hence we notice certain apparent anti-behaviorist
tendencies on the part of Tolman, Bridgman and Carnap.
Although Tolman has again and again formulated his con
ception of the proper methods and goals for scientific
psychology as behavioristic, toward the end of his career
he seemed to attach some importance to the "private data of experience." This was significantly noticeable in his
admission that intervening variables are derived "from
intuition, common sense, and phenomenology." Yet, there was no indication on the part of Tolman how such a deriva
tion could be implemented. Tolman seems to be committed
to saying, at least throughout most of his work, that if
psychology is to attain scientific status, it must proceed
along behavioristic lines; if psychology includes private
experience, it must then forfeit its scientific status.
Should this describe Tolman*s general position, then in
conceding that intervening variables are derived from
"intuition, common sense, phenomenology11 as he is willing
to admit, it would appear that the behavioristic psychol
ogy Tolman espoused should forfeit its scientific status.
50Ibid., p. 237. 210
Yet I do not think that one should be led to such a
conclusion--even in light, I should add, of the recent 51 emphasis by some psychologists like S. Koch. One can
see why Tolman sidestepped any talk about "immediate
experience," "introspection." Yet, it is a mistake to
think that he was denying the importance of immediate
experience. The following quotation supports my con
tention.
Immediate experience, as initially given, is not my private world. It is not something to be studied primarily by psychology. It is, rather, an initial, common matrix out of which both physics and psychology evolved.
As psychologists we do not seek to relive and describe the other man's immediate experiences. Such a reliving must be left,
-^Recently S. Koch, in "Psychology and Emerging Conception of Knowledge," in Behaviorism and Phenomen ology , ed. T. W. Wann, (Chicago: The Univ. of Chicago Press, 1964), has pointed out the need to come to grips with what he calls "experientialism" or "phenomenology." As he puts it: "I want to make it plain that I think there to be an intense current need for particular analyzes of many issues connected with the use of experiential variables in psychology...the fate of psychology must be very much bound up with progress toward their resolution." (pp. 34-35). Koch's recom mendation is indeed worthwhile; however, as far as behaviorism is concerned, Koch's suggestion is of minor significance.
^Tolman, "Psychology versus Immediate Experience," p. 96. 211
as we have said, to metaphysics, or to poetry, or to common sense...that is, to whatever disciplines as may concern them- 1 selves with immediate experience, per se.
Hence it is quite obvious from these quotations that
Tolman was not denying the importance of "immediate ex perience." He was simply taken it for granted. In this
sense, Tolman, like Watson, recommended that if psychology
is to achieve the status of a science, like chemistry or physics, it should not consider private data of experience as its subject matter. Rather the primary subject matter of psychology is exactly the same in kind as that of all other sciences. Even the logical positivists as well as
Bridgman would concede to Tolman the importance of
immediate experience. But if one is to draw a conclusion
from the foregoing, one can perhaps do no better than what
Spence has already pointed out:
The Philosophers of Science, particularly the logical positivists, philosophically- minded scientists such as Bridgman...have succeeded in making the point that the data of all sciences have the same origin, namely, the immediate experience of an observing person, the scientist himself. That is to say, immediate experience, the initial matrix out of which all sciences develop, is no longer considered a matter of concern for the scientist qua scientist. He simply
53Ibid., p. 101 2 1 2 Cakes it for granted and then proceeds to his task of describing the events occurring in it and discovering and formulating the nature,of the relationships holding among them. ^
Thus I deem it quite preposterous on the part of
anyone to deny immediate experience. Whether psycholo
gists should study and delve deeper into the analysis of * immediate experience (as Koch is apparently inclined to)
depends upon the standpoint which one takes to be the
proper concern of psychologists. As far as the neo-
behaviorists are concerned, they deemed it unscientific
to delve into an analysis of immediate experience. It is
simply out of touch with what they consider to be the
proper methods and goals for psychology as a science. It
is assumed by neo-behaviorists that their brand of psych
ology must be couched in concepts which are a part of what
the logical positivists called "the physical thing
language." Undoubtedly it is held that a psychological
language consisting of such undefined terms as "conscious
ness," "introspection," "id," etc., are such that senten
ces containing them are untestable. Hence it seems that
54k . Spence, "The Postulates and Methods of •Behav iorism,*" Psychological Review. Vol. 55, (1948). Reprint ed in Reading in the Philosophy of Science. Feigl and Brodbeck (eds.), pp. 572-573. 213
since the nature of scientific inquiry demands some kind of verification, a science of behavior is by assumption unfit for scientific considerations should it encompass psychological concepts which are devoid of any public empirical meaning. In this manner, it is easy to see why operationalism in psychology belongs to the logical posi-
tivist tradition and see how operationally defined con
cepts should figure in the statements which are verifiable.
That there are problems connected with the behavioristic program, especially in connection with dispensing with psychological concepts in favor of operationally defined 55 ones, has been already pointed out in Chapter V.
But now the question to be answered is why Tolman in
1959 tells us that intervening variables "are not pri marily neurophysio logical as it is suggested by Meejil and
MacCorquodale..., but are derived from intuition, common
sense, a little sophomoric neurology,‘and my own phenom enology."-^ One consequence of this admission is that intervening variables are no longer operationally defined, i.e., tied to empirically stipulated variables. Is Tolman,
S^See Chapter V, Section III.
56Ibid., p. 98. 214 then, forfeiting the scientific status that he so much fought for in his behavioristic psychology? Recall that, at the beginning of this section, I contended that one need not arrive at such a conclusion. 1 still think so despite Tolman's admission that intervening variables are dervied from "intuition, common sense, and phenomenology." * What Tolman is doing in this connection is to indicate to us that perhaps one should in his scientific theorizing pay attention to "the context of discovery" and not just to "the context of justification." That is to say, Tolman appears to be telling us that the process by which scien tific hypotheses embodying intervening variables is, after all, important; for the discovery of hypotheses requires insight, intuition and ingenuity. My sole reason for thinking that Tolman is emphasizing the context of dis covery is borne out by the following quotation
They (intervening variables) do not lead to any then-and-there precise further quantitative predictions. They do, how ever (when mixed with a healthy brew of intuition, common sense, and phenomenology), lead to a suggestion for further types of empirical relations to be tested.57
57Ibid. 215
Whether Tolman would agree to what I am claiming, I am not sure. Yet I feel somehow confident that the importance of the. context of discovery led Tolman in 1959 to his final position on intervening variables.
This should not, however, lead us to minimize
Tollman* s scientific stand in psychology. Hempel, for one, in his book Philosophy of Natural Science emphasizes the context of discovery in addition, of course, to the con text of justification. As Hempel puts it:
Scientific hypothesis and theories are not derived from observed facts, but invented in order to account for them. They constitute guesses at the connections that might obtain between the phenomena under study, at uniformities and patterns that might underlie their occurrence. "Happy guesses" of this kind require ingenuity...58
Thus I think it is fair that the program for psycho logy set down by Tolman constitutes a method for the development of "a science of behavior." Which route one should take in order to develop such a science is a moot question. Regarding behavior it is often contended that the ideal form of explanation consists in the "reduction"
58 C. Hempel, Philosophy of Natural Science. (N. J.: Prentice-Hall, Inc., 1966), p. 15. of behavior to the physiology of organisms. The predominant appeal of the physiological reduction model
is its.accord with the common conviction that in physiology
(as in other physical sciences) we are dealing directly with the substantial constituents of physical reality.
The physiological psychologist may feel justified in * regarding his explanations as telling us "how things
really are.11 Nevertheless, his explanations are justified only if they work, that is, only if they also enable us to make reliable predications relating physiological processes and behavior. As opposed to the physiological reductionist, the pure behaviorial theorist like Tolman
thinks that he can do the job of predicting behavior just as well by using his molar level, regardless of how remote it might be from the current language of
"real entities," so long as it is developed according to rigid criteria such as those Tolman set down for us.
Indeed, the behavior theorist who has achieved predictive efficacy by breaking away from the unreliable intro spective language of 19th century psychology (which pur posed to report real though subjective conditions) will be little inclined to abidcate this advance. But beyond inclination, he can appeal to the criteria for a 217 scientific vocabulary which does not demand that •* physiology necessarily underlie all psychological terms.
Further, as the respective disciplines now stand, there is no evidence to suggest that either the pure behaviorist or the physiological reductionist is more capable than the other to make accurate predictions. While it is true that the physiologist may claim that he has "more basic" explanations than the pure behaviorist, there is no scientific basis for this claim.
By way of summary, X have discussed in this chapter the influence of logical positivism and operationalism on neo-behaviorism. As basis for my analysis, I concen trated on Tolman. In connection with Tolman, I discussed the following:
First, Tolman*s position on intervening variables where it was pointed out that his early view was to consider them as functionally related to the independent and dependent variables.
Second, I discussed the controversy of intervening variables versus hypothetical constructs. In this con nection we partially agreed with Bergmann*s contention that the "distinction is a pseudo-distinction.** However, I suggested to distinguish between mental variables and 218 physiological variables rather than between intervening variables and hypothetical constructs. Our distinction might be helpful in characterizing different approaches
in the actual development of concept and theory formation
in the context of discovery for psychology.
Third, I discussed some of the influences of logical positivism and operationalism on Tolman. It was pointed out that Tolman*s concern' in (a) finding an adequate base
in order to anchor securely his psychological statements
and (b) his deep concern with finding ways of introducing
intervening variables in theoretical psychology were deeply influenced by the methodological views put forth by logical positivism and Bridgman.
Fourth, in the concluding note, I attempted to
defend Tolman*s claim that behavioristic psychology need not forfeit its scientific status despite Tolman*s
admission in 1959 that intervening variables are derived
from **intuition, common sense and phenomenology." I
contended that the prime factor behind this was Tolman*s
concern to emphasize the "context of discovery" in the ' * formulation of scientific hypothesis. CHAPTER VII
CONCLUSION
The preceding study aimed at establishing the dual
influence of logical positivism and Bridgman's operation-
ism upon behaviorism. Specifically, the object of the
study was to point out the dual influence upon Tolman. In
the course of our study, we encountered three revolutions
in three major fields, namely, physics, psychology and philosophy. Although these three movements occurred almost independently of each other, yet to use S. S.
Stevens' words, "a general community of spirit among them
led directly to extensive cross-fertilization."^ Opera-
tionism, as propounded by Bridgman in 1927 in his Logic of
Modem Physics, was a movement which focused its attention upon getting rid of the last vestiges of metaphysical obscurity still remaining in some parts of science, namely, physics. Logical positivism was a movement
3-S. S. Stevens, "Psychology and the Science of Science," Psychological Bulletin. Vol. 36, (1936). Reprinted in Theories in Contemporary Psychology, p. 74.
219 220 directed against speculative philosophy. The avowed intention of the logical positivists was to replace phil osophy. by the logic of science and, for Carnap, the "logic of science is the syntax of the language of science."^
Behaviorism as initiated by Watson early in this century was a movement directed against the old mentalistic psych- * ology, i.e., introspectionist schools of thought.
In the remainder of this chapter, I shall, once again, attempt to (1) clarify the historical influences and (2) ask some philosophical questions about the result of the influences mentioned in (1).
1. Clarification of historical influences.
A. Watson. It is clear that neither logical positivism nor Bridgman*s pperationalism had any direct influence on
Watson inasmuch as he published his first behavioristic views In 1913, in an article entitled "Psychology as the
Behaviorist views it." This is almost two decades before any official publication by the logical positivists and
Bridgman appeared. This is not to say, however, that
Watson could not have been influenced by the 19th century positivistic movement spearheaded by Mach, Avenarius,
o Carnap, Logical Syntax of Language. Sections 72, 73. 221
Pearson, Duhem. In addition it is not at all unlikely
that Watson was influenced by James' pragmatism.
B, Bridgman. Bridgman was directly influenced by
Einstein's views on relativity which brought him to con
sider operational analysis in physics. The Logic of
Modern Physics, however, was not influenced by either * logical positivism or Watson's behaviorism. Neither has
Bridgman influenced Watson. Yet one can say that Bridgman
has influenced logical positivism by his use of the term
"operational definition." As Frank puts it:
Reichenbach had explicitly pointed out that what is needed is a bridge between the sym bolic system of axioms and the protocols of the laboratory. But the nature of this bridge had been only vaguely described. Bridgman was the first who said precisely that these "rela tions of coordination" consist in the descrip tion of physical operations. He called them, therefore, "operational definitions." This name has been generally accepted.3
On the other hand, Schlick in "Meaning and Verifica
tion" gives us a different outlook about Bridgman. As he
puts it: "Professor Bridgman's book on The Logic of
Modern Physics is an admirable attempt to carry out this
program Einstein's initial adoption of a meaning criterion for "non-local simultaneity'-!/^ for all concepts
^Frank, Modem Science and its Philosophy, p. 44. 222 of physics."** Thus it seems that Schlick does not think
Bridgman deserves credit for influencing logical positiv ism. However, although Bridgman might have influenced the logical positivists through his use of the term "opera tional definition," still, as indicated in Chapter IV,
Bridgman has ignored striking differences in the manner by which various theoretical terms are introduced or anchored. As a technique for concept formation, opera- tionalism stressed one of the relationships between theo retical terms and observation terms— the operational definition for concepts appearing in the empirical hier archy of terms,
C. Logical Positivism. Although logical positivism influenced neo-behaviorism, it is clear, as indicated earlier, that they had no direct influence on Watson,
Logical positivism, for the most part, occurred "indepen dently" of Watson*s behaviorism and Bridgman’s operation- ism. Yet it should be kept in mind, as indicated in
Chapter V, that some logical positivists such as Carnap and Hempel were attracted to Watson's behaviorism. Unlike
Watson, however, they were not at all concerned with
^Schlick, "Meaning and Verification." p. 149. developing a psychological theory. Rather they were
primarily interested in the logical analysis of psych ological propositions in order to discover whether they
display any relationship to the ones in the physical
sciences.
D. Neo -behaviorism (Tolman): With Watson one comes across one of the earliest attempts to rid psychology of
its speculative vestiges. Neo-behaviorism could also be
viewed as a second attempt to cleanse psychology from its past and provide us with a more effective scientific methodology. Tolman, no doubt, was influenced by Watson.
For like Watson, he, too, emphasized that psychology
should be treated as one among other scientific inquiries
But unlike Watspn, who constructed his theory upon the molecular level, Tolman built his theory upon the molar level. Aside from Watson, logical positivism and
Bridgman*s operationism directly influenced Tolman. It was pointed out in Chapter VI that Tolman's deep concern in (a) finding an adequate base to anchor securely his psychological statements and (b) finding ways of intro ducing intervening variables in theoretical psychology were deeply influenced by the methodological views put forth by logical positivism and Bridgman’s operationalism 224
With all this historical background behind us, I would like to address myself to some philosophical questions.
2. Some philosophical questions about the result of the influences mentioned in (1.)
A. Is logical positivism bad philosophy?
Before attempting to answer this question, we should inquire what we mean by "bad?" By the term "bad" we mean
"restricted," "limited." Hence bad philosophy is restric ted. limited, if (1) it does not attempt to resolve onto logical and epistemological questions and (2) it is not an ontology which, in its genuine attempt to treat questions of first philosophy, cannot dialectically answer all the philosophic questions about the assay of what there is and of knowing. Now to the question whether logical positi vism is bad philosophy. One quick reaction to the ques tion is, Yes. For it neglects inquiry into metaphysics, epistemology and ethics. Recall that the early positiv ists declared that these traditional inquiries are mean ingless. Later, as we already know, the logical positiv ists found it necessary to distinguish various components of meaning, whereby they claimed that these traditional inquiries were devoid of any cognitive content, although 225 the concepts involved might still have strong psycho logical overtones. Thus it is my contention that logical positivism is bad philosophy since it rejected study of traditional inquiries. On the other hand, logical posi tivism, as a form of scientism, is not bad philosophy, since it equates philosophy with logic of science. Phil- * osophy was replaced by the logic of science, and the
i "logic of science is the syntax of the language of science."^
B. Can good psychology stem from bad philosophy?
The answer would simply be, No. Good psychology (neo behaviorism) simply did not. It stemmed rather from rigorous examination of methodological formulations of science. Good psychology stemmed from philosophy con strued as the philosophy of science. And philosophy of science (the logic of science) is, according to positiv ism, neither an ontological nor an epistemological analy sis of what exists or of the structure of knowledge- situation; it is an analytical description, within the context of justification, of the status and the structure of scientific concepts, laws, theories and of broad
^Carnap, Logical Syntax of Language, sections 72, 73. 226 factual theses which scientists take for granted.
C. Can bad psychology stem from bad philosophy?
In order to answer this question, let us consider meta physical materialism. Advocates of metaphysical material ism deny that there are any nonmaterial entities, events or processes. Put differently, metaphysical materialism * simply denies the existence of minds (conscious states)•
Such a position is obviously absurd and would, if adopted by behaviorists, lead to disastrous consequences. For instance, if metaphysical materialism were adopted by behaviorists, it would certainly hamstring the whole behavioristic program in completely neglecting mind or conscious states in any form. Also adoption of meta physical materialism by behaviorists would throw out the psycho-physiological parallelism which behaviorism espouses as a factual thesis. Thus I think, bad psychology £ would stem from bad philosophy.
D. Why is science (stemming from bad philosophy) bad?
This question is similar to B. or C. and already answered.
®By happenstance, perhaps Hobbes* metaphysical materialism would lead to better psychology or science of man than Hegel's objective idealism. 211'
E. What import or bearing does scientific method
(or do the results of theoretical physics) have upon philosophy proper— upon certain large questions of know ledge and reality?
Sir James Jeans,? for example, concludes that we can no longer pretend to understand the "real nature of things" ♦ since theoretical physics only describes "our observations on nature." These processes in the "inaccesible sub stratum of nature" are thus best construed, according to
Jeans, as "mental" rather than "material." But his conclusion hinges on faulty common-sense analysis of, for example, "perceptions," "measures of things," and "micro scopic precision." Only by blurring the distinction between perceptual objects and perceptual apprehension, the distinction between measure as a relational property of a thing and the process or record of estimating that property does Jeans find support for his "mentalistic" or idealistic interpretation of nature. Likewise Jeans' claim that we must abandon both causal determination in our everyday world and any hope of precise knowledge of
?Physics and Philosophy. N. Y., 1943, passim. Also see Nagel's review in The Nation 156: (1943), reprinted in Logic Without Metaphysics. Glencoe, III., 1956, pp. 317-319. 228
it rests only upon inaccurate characterization of the relations between quantum mechanics and our macro-world.
What Jeans has done may best be described as a haphazard attempt as layman at rationalizing some popular version of "idealism" that he has acquired from his local heritage. He has not arrived at "idealism" by the methods V and theories of physics nor by those disciplined habits of mind characteristic of philosophers of science. From what he has argued, we can conclude nothing as to the bearing of physics on philosophy proper. And this fact supports my general claim that, strictly speaking, when science and philosophy are said to interact or to in fluence each other, only three sorts of things occur:
(1) science itself asks questions which require philosoph ical analysis for their solutions and these resolutions are the circumscribed task of the methodology of science.
(2) Philosophers may simply discard some of their ana lyses, because science has shown that the world is in some respect different from what philosophers had taken it to be--but not because science has proved philosophers' analyses wrong. (Consider, for example, the philosophical analyses of absolute and relational space, and the recent physical hypothesis of relational space-time.) 229
(3) Ontologists can be of distinct service to the philosophy of physics by clarifying, for example, the
status of theoretic entities like particle. By showing
that science unproblematically replaces the ordinary or perceptual object by a physical object (a set of micro particles) , ontologists take the wrong kind of dialec-
tical pressure off both theoretical physicists and phil osophers of science. Thus the theoretician need not
(indeed, he cannot) argue that science problematically constructs the physical particles or object out of per ceptual objects which, since they are themselves con
structions out of sense data, are not real— or, like
Mach, argue that therefore the particles are not real.
Relieved of such pressure, our theoretician can deal properly with those features of the particle which are peculiar to it and are in need of dialectical clarifi cation.®
®See Bergmann's "Physics and Ontology," reprinted in Logic and Reality, pp. 109-123. Also, see Hinshaw's "Science versus Ontology," Philosophy of Science (forth coming) . BIBLIOGRAPHY
A. BOOKS
Ayer, A. J. Language, Truth and Logic, New York: Dover Publications, Inc., 1946.
______. (ed.) Logical Positivism.~ Glencoe, Illinois: The Free Press, 1960.
Bergmann, G. The Philosophy of Science. Madison: University of Wisconsin Press, 1957.
Bridgman, P. W. The Logic of Modern Physics. New York: Macmillan Co., 1927.
______. The Way Things Are. Cambridge, Mass.: Harvard University Press, 1959.
Carnap, R. Philosophy and Logical Syntax. London: Kegan, Paul, Trench, Trubner and Co., 1937.
______. Logical Syntax of Language. London: Kegan; New York: Harcourt Brace, 1937.
Estes, W. K. et al. Modem Learning Theory. New York: Appleton-Century-Crofts, Inc., 1954.
Feigl, H . , and Brodbeck, M. (eds.). Readings in the Philosophy of Science. New York: Appleton-Century- Crofts, Inc., 1953.
Feigl, H. and Scriven, M. (eds.). Minnesota Studies in the Philosophy of Science. Volume I. Minneapolis: University of Minnesota Press, 1956.
Feigl, H. and Sellars, W. (eds.). Readings in Philosoph ical Analysis. New York: Appleton-Century-Crofts., Inc., 1949.
230 231
Frank, P. G. (ed.). The Validation of Scientific Theories. New York: Collier Books, 1956.
Hempel, C. G. Fundamentals of Concept Formation In Empirical Science. Chicago: University of Chicago Press (Int. Encycl. of Unified Science), 1952.
______. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. New York: The Free Press, 1964.
______. Philosophy of Natural Science. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1966.
Linsky, L. (eds.). Semantics and the Philosophy of Language. Urbana: University of Illinois Press, 1952.
Marx, M. H. (ed.). Psychological Theory. New York: The Macmillan Co., 1961.
Pap, A. An Introduction to the Philosophy of Science. New York: The Free Press, 1962.
Popper, K. R. Logik der Forschung. Vienna: Springer, 1935. Eng. transl.. The Logic of Scientific Discovery. London: Hutchinson, 1958.
______. Conjectures and Refutations: The Growth of Scientific Knowledge. New York: Harper & Row, Publishers, 1968.
Scheffler, I. The Anatomy of Inquiry. New York: Alfred A. Knopf, 1967.
Schlipp, P. (ed.). The Philosophy of Carnap. Illinois: Open Court, 1963.
Tolman, E. C. Purposive Behavior in Animals and Men. New York: Appleton-Century-Crofts, 1932.
______. Behavior and Psychological Man: Essays in Motivation and Learning. Berkeley: University of California Press, 1958. 232
Warm, T. W. (ed.). Behaviorism and Phenomenology. Chicago: The University of Chicago Press, 1964
Watson, J* B. The Wavs of Behaviorism. New York: Harper & Brothers Publishers, 1928.
______. Psychology From the Standpoint of a Behaviorist. (3rd. ed.), Philadelphia: Lippincott, 1929.
______. Behaviorism. Chicago: University of Chicago Press, 1930. *
B. ARTICLES
Bergmann, G. "On Some Methodological Problems of Psychology," Philosophy of Science. 7, (1940). Reprinted in Feigl, H. and Brodbeck, M. Readings in the Philosophy of Science. 627-636.
______. and Spence, K. "Operationalism and Theory in Psychology," Psychological Review. 48: 1-14 (1941).
•_____ . "Frequencies, Probabilities, and Positivism," Philosophy and Phenomenological Research. 6: 26-41 (1945).
______. "Comments on Professor Hempel's 'The Concept of Cognitive Significance,'" Proceedings of the American Academy of Arts and Sciences. 80: 78-86 (1951).
______. "The Logic of Psychological Concepts," Philosophy of Science. 18: 93-110 (1951).
______. "Theoretical Psychology," Annual Review of Psychology. 4: 435-458 (1953).
______• "Sense and Nonsense in Operationism," Scientific Monthly. 79, (1954). Reprinted in Frank, P. The Validation of Scientific Theories. 45-56. 233
______• "The Contribution of John B. Watson," Psychological Review, 63: 265-276 (1956).
______. "Ontology and Physics," Philosophy of Science, 26, (1961). Reprinted in Bergmann, G, Logic and Reality, (Wisconsin Press, 1964), 109-123.
Bridgman, P. W. "Operational Analysis," Philosophy of Science, 5: 114-131 (1938).
______. "Some General Principles of Operational Analysis," Psychological Review,052: 246-249 (1945).
______. "Some Implications of Recent Points of View in Physics," Revue Internationale de Philosophie. 3: 479-501 (1949).
______. "The Nature of Some of Our Physical Concepts," British Journal of the Philosophy of Science, 1: 257-272 (1951).
______. "The Present State of Operationalism," Scientific Monthly. 79, (1954). Reprinted in Frank, P. The Validation of Scientific Theories. 75-80.
Bridgman, P. W. "The Logic of Modern Physics after 30 Years," Daedalus. 3: 518-526 (1959).
Buress, C. E. "Operations, Constructions and Inference," Journal of Philosophy, 37: 393-401 (1940).
C a m p , R. "The Elimination of Metaphysics Through Logical Analysis of Language," Erkenntnis, 2, (1932). Reprinted in Ayer, A. J. Logical Positivism, 60-81.
_ . "Psychology in Physical Language," Erkenntnis, 3, (1932-33). Reprinted in Ayer, A. J., Logical Positivism. 165-198.
_ . "Testability and Meaning," Philosophy of Science, 3, (1936) and 4, (1937). Reprinted in part in Feigl, H. and Brodbeck, M. Readings in the Philosophy of Science. 47-92. 234
______. "Meaning Postulates," Philosophical Studies. 3: 65-73 (1952).
______. "On Some Concepts of Pragmatics," Philosophical Studies. 6: 89-91 (1955).
______. "The Methodological Character of Theoretical Concepts," in Feigl, H. and Scriven, M. Minnesota Studies in the Philosophy of Science. I: 38-76 (1956).
Chisholm, R. "Sentences about Believing," In Feigl, H., Scriven, M., and Maxwell, (eds.). Minnesota Studies in the Philosophy of Science. Vol. II, (1958), 510-520.
______. "The Problem of Counterfactual Conditions," Journal of Philosophy. 44: 113-128 (1947)..
Church, A. "Review of A. J. Ayer, Language. Truth and Logic, Second Edition," Journal of Symbolic Logic. 14: 52-53 (1949).
Cominan, J. W. "Mental Terms, Theoretical Terms and Materialism," Philosophy of Science. 35: 45-63 (1968).
Craig, W. "On Axiomatizability within a System," Journal of Symbolic Logic. 18: 30-32 (1953).
______. "Replacement of Auxiliary Expressions," Philosophical Review. 65: 38-55 (1956).
Ewing, A. C. "Meaninglessness," Mind. 46: 347-364 (1937).
Feigl, H. "Existential Hypotheses," Philosophy of Science. 17: 35-62 (1950).
______. "The Mind-Body Problem in the Development of Logical Empiricism," Revue Internationale de Philos- ophie. 4, (1950). Reprinted in Feigl, H. and Brodbeck, M. Readings in the Philosophy of Science. 612-626. 235
Feyerabend, P. K. "Explanations, Predictions, Theories," in Baumrin, B. (ed.). Philosophy of Science. The Delaware Seminar. Volume X, 1961-62. New York: John Wiley & Sons, 1963, 1-39.
Goodman, N. "The Problem of Counterfactual Conditionals," The Journal of Philosophy. 44: 113-128 (1947).
Hempel, C. G. "The Logical Analysis of Psychology," Revue de Synthese. (1935). Reprinted in Feigl, H. and Sellars, W. Readings in Philosophical Analysis. 373-384.
"Problems and Changes in the Empiricist Criterion of Meaning," Revue Internationale de Philosophie. 11. (1950). Reprinted in Linsky, L, Semantics and the Philosophy of Language, 163-185.
"The Concept of Cognitive Significance: A Reconsideration," Proceedings of the American Academy of Arts and Sciences. 80: 61-77 (1951).
"A Logical Appraisal of Operationism," Scientific Monthly. 70, (1954). Reprinted in Frank, P. The Validation of Scientific Theories. 56-69.
"Implications of Carnap’s Work for the Philosophy of Science," in Schilpp, P. A. The Philosophy of Carnap, 685-709.
______. "Postscript (1964) on Cognitive Significance," in Hempel, C. Aspects of Scientific Explanations. 120-122.
"The Theoretician’s Dilemma," in Feigl, H. Scriven, M. and Maxwell, G, (eds.). Minnesota Studies in the Philosophy of Science. Volume 11. Minneapolis: University of Minnesota Press, 1958. Reprinted in Hempel, C. Aspects of Scientific Explanation. 173-226,
Ilinshaw, V. Jr. "Levels of Analysis.11 Philosophy and Phenomenological Research. 11: 213-220 (1950). 236
Hinshaw, V. Jr. "Science versus Ontology," Philosophy Of Science, (forthcoming).
Koch, S. "Psychology and Emerging Conceptions of Knowledge as Unitary," in Wann, T. W. Behaviorism and Phenomenology. 1-41.
Lewis, C. I. "Experience and Meaning," Philosophical Review. 43, (1934). Reprinted in Feigl, H. and Sellars, W. Readings in Philosophical Analysis.
Madden, E. H. and Kiteley, M. J. "Postulates and Meaning," Philosophy of Science. 29: 66-78 (1962).
Marhenke, P. "The Criterion of Significance," Proceedings and Addresses of the American Philosophical Association. 23, (1950). Reprinted in Linsky, L. Semantics and the Philosophy of Language. 139-159.
Markinson, D. "Nidditch Defense of Verifiability," Mind. 74: 242-250 (1965).
MacCorquodale, K, and Meehl, P. "On a Distinction between Hypothetical Constructs and Intervening Variables," Psychological Review. 55, (1948). Reprinted in Feigl, H. and Brodbeck, M. Readings in the Philosophy of Science. 596-611.
______. "Edward C. Tolman," in Estes et. al. Modem Learning Theory. 177-266.
Morris, W. "Pragmatism and Logical Empiricism," in Schilpp, P. The Philosophy of Carnap. 87-98.
Neurath, 0. "Protocol Sentences," 3, Erkenntnis (1932-33). Reprinted in Ayer, A. J. Logical Positivism. 199-208.
Nidditch, P. "A Defense of Ayer's Verifiability Principle Against Church's Criticism," Mind. 70: 88-89 (1961). 237
0*Connor, D. J. "Some Consequences of Professor A. J. Ayer's Verification Principle," Analysis. 10: 67-72 (1950).
Russell, L. J. "Review of Bridgman's The Logic of Modem Physics." Mind. 47: 355-361 (1928).
Rynin, D. "Vindication of L*G*C*L P*S*T*V*SM," Proceedings and Addresses of the American Philoso phical Association. 30: 45-67 (1957).
Scheffler, I. "Prospects of a ModestvEmpiricism," The Review of Metaphysics, 80: 383-400; 602-625 (1957).
Schlick, M. "Positivismus und Realismus," 3, Erkenntnis (1932-33). English Translation in Ayer, A. J. Logical Positivism, 82-107.
______. "Meaning and Verification," Philosophical Review, 45, (1935). Reprinted in Feigl, H. and Sellars, W. Readings in Philosophical Analysis, 146-170.
Spence, K. and Bergmann, G. "Operationalism and Theory in Psychology," Psychological Review, 48: 1-14 (1941).
______. "The Nature of Theory Construction in Contemporary Psychology," Psychological Review. 51: 47-68 (1944).
______. "The Postulates and Methods of Behavioism," Psychological Review, 55, (1948). Reprinted in Feigl, H. and Brodbeck, M. Readings in the Philosophy of Science, 571-584.
Stevens, S. S. "Psychology and the Science of Science," Psychological Bulletin, 36, (1936). Reprinted in Marx, M. Theories in Contemporary Psychology, 47-76.
Tolman, E. C. "A New Formula for Behaviorism," Psycholo gical Review, 29, (1922). Reprinted in Tolman, E. C. Behavior and Psychological Man. 115-129. ______. "Psychology versus Immediate Experience," Philosophy of Science, 2, (1935). Reprinted in Tolman, E. C. Behavior and Psychological Man, 94-114.
______. "Operational Behaviorism and Current Trends in Psychology," Proceedings of the 25th Anniversary Celebration Inauguration of Graduate Study. Los Angeles: University of Southern California Press, 1936. Reprinted in Tolman, E. C. Behavior and Psychological Nan. 115-129.
______. "The Determiners of Behavior of a Choice Point," Psychological Review. 45, (1938). Reprinted in Tolman, E. C. Behavior and Psychological Man. 144-178.
______. Ritchie, B. F. and Kalish, D. "Studies in Spatial Learning I. Orientation and the Short Cut Journal of Experimental Psychology. 36: 13-24 (1946)
______. "Discussion," Journal of Personality. 18: 48-50. (1949).
______. "Principles of Purposive Behavior," In Koch, S. (ed.). Psychology: A Study of a Science. XX, New York: Me Graw-Hill, 1959, 92-157.
Ullian, J. "A Note on Scheffler on Nidditch," Journal of Philosophy, 62: 274-275 (1965).
Waiting, J. and Brown, R. "Amending the Verification Principle," Analysis, 11: 87-89 (1951).
Watson, J. B. "Psychology as the Behaviorist Views It," Psychological Review. 20: 158-177 (1913).
Wilson, F. "Definition and Discovery," British Journal for the Philosophy of Science, 18: 287-303 (1967).