An Empiricist Schema of the Psychophysical Problem Gustav Bergmann Philosophy of Science, Vol. 9, No. 1. (Jan., 1942), pp. 72-91. Stable URL: http://links.jstor.org/sici?sici=0031-8248%28194201%299%3A1%3C72%3AAESOTP%3E2.0.CO%3B2-U Philosophy of Science is currently published by The University of Chicago Press.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/ucpress.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Fri May 18 08:19:17 2007
An Empiricist Schema of t h e
Psychophysical Problem*
BY
GUSTAV BERGMANN H I S paper does not purport to present a fully developed argument. Rather, it intends t o delineate, in broad strokes and in a synoptic manner, the total frame of reference of Scientific Empiricism as far as it is necessary for an integration of the several contributory clarifications of the complex, usually referred to as the psychophysical problem. Several considerations seem to justify such an attempt. I t is desirable, from time to time, to free the results of logical and methodological analysis from their specialized contexts. Only thus can they be presented in a way which brings t o the fore their continuity with more traditionally worded epistemological thought. By doing this such a presentation should also help t o dissolve the current objections of formalism, scientism, and philosophical irrelevance so frequently based on misunderstanding. I t is only fair to admit, however, that these misunderstandings have been facilitated by the preoccupation with the technical, and the brevity with regard t o the nontechnical, which characterize the modern analysts. T h e perennial mind-body issue is a conspicuous point in case. *References in footnotes have been omitted; the argument itself is believed to show clearly to which writers i t owes most. I wish, however, to acknowledge m y debt to my friend, Professor W. S. Sellars from whom I have learned, by disagreement, in two years of assiduous discussion.
72
G. Bergmann
As a matter of fact, there exists a wide-spread belief that the related theses of physicalism and behaviorism, each and in isolation, are offered as a solution or rather, as it has become customary to say, a scientific dissolution of the problem. Similarly, the thesis that strictly linguistic analysis is the epistemological approach least subject to the psychological fallacy, is construed as a formalistic assertion that no problem is left a t all, i.e., that there is nothing worth while to discuss within the traditional complex. As far as I can see, all these interpretations are mistaken, while the criticism of what they assume to be asserted would not be without foundation. T h e refutation of all these misinterpretations, as well as some casual support of them, can be found in the writings of the empiricist group itself, though possibly not always in a sufficiently explicit fashion. All that will be said here is at least implicitly contained either in those writings or in the work of the pragmatists. T h e brevity of the argument, or even the complete lack of it, which will be noted at several places of the present schematic organization, should, therefore, not be ascribed to a dogmatic attitude; its only cause is reluctance toward repetitiousness. What follows is essentially a synoptic schema in the form of an annotated chain of theses. Familiarity with what these theses refer to in a merely allusive manner is assumed. Three principles, however, are so fundamental for the contemporary Neo-Humean approach that they should be explicitly mentioned a t the very beginning. Loosely worded for the purpose of an introductory stage setting, these principles are: first, the impossibility of synthetic-a-priori knowledge; second, an empiricist meaning criterion as the touchstone for all factual cognitive meaning; and, third, the exclusion of all ontological statements. One of the main causes of uneasiness is the alleged claim of the linguistic approach to supersede and entirely eliminate epistemology. Such a claim is not made, and if it were, would not be justified; this in spite of some ambiguous utterances to the contrary. Actually the formalistic turn towards language has no other purpose than to provide an adequate instrument for a theory of human understanding. I n a broader context, and apart
Scientific Empiricism
from its considerable intrinsic interest, linguistic analysis has as its main function not to supersede epistemology, but rather t o guard against the ontological and psychologistic pitfalls which so abundantly beset the epistemological problem. For the empiricist philosopher, this problem consists in distinguishing between, and tracing the strands of, the analytic and the synthetic, the necessary and the contingent, the formal and the factual elements in our knowledge or, for that matter, in our language. Whatever terms are used here to refer to this basic dichotomy is of little importance. I t s clarification is the task of epistemology itself, and cannot be achieved through a sweeping characterization. Two remarks, however, shall be added. First, it must be said that the complexity of the logical problems involved does not, t o my mind, detract from the importance of the episten~ological endeavor. According to recent results the concept of analyticality is, in a technical meaning of the term, indejnite and, moreover, rather precariously perched on an ever widening scale of languages. I n such a situation a logician might very well maintain that analyticality is merely truth by convention. T h e opposite view, however, is equally justified; logicians holding it fail to see, and probably rightly so, what else could possibly be asked for under the circumstances beyond such an indefinite and relativistic criterion. After everything will be said, it might turn out that both parties say exactly the same thing, a t least to everybody not too engrossed in the ingenious technicalities of the issue. If the clarification of this point, which is now in progress, should result in making us increasingly conscious of the pragmatic limitations which even the logician can not escape, such an outcome would be in line with a genuine empiricism, no matter how important a place logic has in its structure. Secondly, after the analytical character of logic and arithmetic has once been recognized, and after the completion of their formal structure has been handed over t o the specialist, it is no longer this type of tautology which commands the main interest of epistemology proper. At this stage interest turns to that class
G. Bergmann
of statements, of which Ayer's example is a convenient representative: 'A thing can not be a t two places at the same time'. These are the statements which have been made analytical by virtue of the relations between the introductory chains for the derived terms (in the example: thing, place, time) occurring in them, and where the specific form of this structure has been suggested by the prospective empirical application of the language. Therefore, a statement like: 'A thing-time-space language is applicable' contains an epistemological clarification, and renders a t least part of the possible cognitive content which might be read into the ontological thesis of realism. On the other hand, the peeling off of strata of analyticality by means of the thought experiment technique corresponds to the psychologistic attempts of isolating ('s trata of experience". With the thought experiment technique, as opposed to metaphysical speculation, we are at present fairly familiar because of the ample use which has been made of it on the methodological level. T h e distinction just made between 'methodological' and 'epistemological' is merely one of loose convenience. T h e former is preferably used for an analysis in terms of a thing language, i.e., what Carnap calls a physicalistic analysis. Undoubtedly this, too, is a genuinely epistemological enterprise, as shown by such well-known examples as the theory of measurement and the socalled operational analysis of physical time and space with their main concern for the distinction between the analytical and the empirical. Though it is the reconstruction of the thing language as a derived level within a 'sense data' analysis, which is traditionally thought of as the typical epistemological problem, there is no difference whatsoever between what is actually done in either case. I realize that part of this point is sometimes granted, though with the important qualification that from penetrating to the deepest layers the empiricist bars himself by the silent presuppositions of his method. This criticism is closely tied u p with certain strands in the psychophysical complex; an attempt to answer it is therefore implicit in this schema as a whole. What
76
Scientific Empiricism
these preparatory remarks insist on is merely that Der Logische Aufbau der Welt is an epistemological undertaking, no matter how formally it is deemed necessary t o proceed. Let us begin, then, by indicating some theses which concern linguistic analysis and are relevant for the development of the schema. I n the following, 'language' is always thought of as a completely analyzed and formalized structure. Of course, no such empirical language exists, but still there are several reasons why reference to it should not be considered as an artificial, pseudo-rationalistic device. First, it is not held that there exists one and only one correct way of ultimately and completely analyzing our empirical language or, for that matter, of reconstructing it. I n the present context, the two expressions, analyzing and reconstructing, have exactly the same meaning. Every analysis of the empirical language, no matter how incomplete and partial, aims at, and actually provides, the means for a t least a model reconstruction, and vice versa. As indicated by this remark, it is not maintained either that a perfect empirical language can ever be reached. Quite to the contrary, a principle of broadest tolerance as to possible model formalizations of the empirical language, which in turn is in a state of continuous remaking, is being advocated. Furthermore, none of these analyses including the 'sense data' analysis, confusingly so called, can lay claim to any epistemological excellence in itself. Decisive epistemological clarification results only from a study of the relationships between several possible analyses. Not the least benefit reaped from this study is that we learn to avoid certain ambiguities. I mean those fleeting transitions which are mediated by the use of the same symbol, where the user has once the one analysis in mind, and once the other. This is of particular importance with respect to one of the most basic distinctions zuithin any particular analysis, the distinction between primitives and derived terms. Not before the relativity of these purely grammatical categories was realized, could the rationalistic-ontological bias which vitiates Wittgenstein's contribution be eliminated. For neither atomic sentences
G. Bergmann
77
nor atomic facts in an absolute sense have any place within an empiricist epistemology. A certain term like 'table', for instance, might occur in a primitive sentence of a physicalistic language ('x is a table'), while it is at the same time a highly derived compound (empirical construct) in a sensualistic analysis. I t is, of course, never a primitive name in the strict sense in which only 'this' is. For the sake of continuity with the traditional terminology, it might also be mentioned that the term 'primitive name', if so restricted, is synonymous with one of the meanings of 'particular'. I n another usage, however, a derived term might also be called a 'particular', provided that its referent possesses enough of those features, frequently spatio-temporal individuality and continuity, which suggest a substantive use. Ayer has recently called our attention again t o the difficulties which arise from the failure to distinguish between the grammatical and the empirical (falsely, ontological) meaning of the word' particular'.' Further, it should be noted that the primitive names of any given analysis can be used as arguments only in accordance with the rule of types. And it goes almost without saying that the specific syntax of the primitive functions is not part of the logical, but rather a scientific calculus; e.g., the following fragment of a grammar of the color predicates: '(x is red) implies (x is not blue)'. T h a t this latter insight is almost forced upon us is one of the great advantages of logistics. I t enables one to put his finger exactly on the spot (intuitive induction) where the psycho~ogisticfallacy creeps into modern intuitionistic approaches. Moreover, this obvious restriction in the use of the primitive names and functions automatically forestalls a host of further errors. Consider, for instance, in a 'sense data' analysis, the possible cognitive meanings of the typically philosophical predicates: 'real', 'objective', 'subjective', 'being perceived'. It is obvious, and has been pointed out again and again, that in such an analysis all these A third possible precise meaning of 'particular' would refer to the descriptive symbols of a language in general without restriction to the proper names. 'Descriptive' is used in Carnap's sense. What is here and later on referred to as 'primitive functions' corresponds to Carnap's undefined descriptive predicates and functors.
Scientific Empiricism
meanings lie on a very high level of derivation. One of the possible meanings of 'being real', e.g., is to be a particular of the second type mentioned above (spatio-temporal individuality and continuity). Hence expressions like 'the data are real' are either not sentences a t all or, a t best, add a new, merely tautological meaning to the confusing variety of usages. I n a merely suggestive and literally false way this state of affairs can be epitomized by a formula: Nothing can be predicated about the data as such. But that is exactly what every kind of ontology tries to do. T h e fatal substitution of the expression 'data' for 'primitive names' and the sensualistic-psycl~ologistic deviation which it implies leads to the next remark. Secondly, and this is one of the most crucial points, no analysis of the empirical language rests upon any epistemological assumptions, nor has it anything to do with perception or a theory thereof, whatever the expression 'theory of perception' might mean. This assertion is, indeed, the most important aspect of the thesis that language analysis is strictly formal. With respect to its first half, i.e., those alleged assumptions, it is safe to say that they all resort ultimately to what one is tempted t o call, in the terminology of their advocates, the ontological priority of the epistemological subject. I t was just to discourage errors of this sort, and in order t o indicate more clearly one of the fundamental tenets of Scientific Empiricism, that Carnap dropped the deceptive label of methodological solipsism. T h e fundamental principle involved is the denial that there is anything like an epistemological subject, that there is any necessity for, or any sense in creating this fiction. But to push this point farther one has to enter into the psychophysical complex, and reference must therefore be made again to this schema as a whole. At its present stage the discussion is mainly concerned with the second half of the italicized sentence, viz., that there is no relation whatsoever between language analysis and theory of perception. All that can possibly be meant by the latter expression belongs to the science of psychology. T o keep this in mind is of particular importance in dealing with the so-called 'sense data' analysis. Terminological continuity is the only reason why the
G. Bergmann
79
dangerous expression 'sense data' is being used here a t all, subsequently even without quotation marks. T o clarify this point some preparation is necessary. A sense-data language is a reconstruction of the empirical language in which the terms of common sense are on a derived level. I t s primitive predicates and relations are words like 'hot', 'green', 'simultaneous', and so on, without the subjective ('I7,'You') and the objective ('thing', etc.,) implications which are already contained in the common sense statement: 'Somebody sees something green'. Thus it is naturally the best tool to use, if one wants to investigate those features (empirical laws) which determine the grammar of the common sense level in statements as: 'A thing can not be a t two places a t the same time'. Aphysicalistic language, on the other hand, is an analysis whose primitives refer to physical objects, properties of, and relations between them. I t s primitives therefore correspond, roughly, to a certain level of derivation within a sense-data analysis. I n a sort of crude illustration, one might visualize three horizontal planes. On the highest level of such a scheme one will find the most highly compounded empirical constructs of the sciences including the science of psychology, on the medium level one will find the thing language of common sense, and on the lowest the primitives of the sense-data analysis. Physicalistic analysis, then, moors its primitives on the medium of these three planes and solves its task by filling the interval between the upper two levels with its hierarchy of empirical constructs (derived terms). Sense-data analysis anchors its primitives on the lowest plane and extends fro; this basis to the top. Remembering the vague sense in which the two expressions 'methodological' and 'epistemological' have been previously opposed to each other, one can allocate the space between the upper two planes to methodology, and the space between the two lower ones to "what is traditionally thought of as the typical epistemological problem " (p. 75). This representation also indicates the only sense in which the one analysis may be said to be contained in the other, and hence the latter to be more comprehensive or deeper than the former. Let us finally mention that the primitives of an analysis are some-
80
Scientific Empiricism
times also spoken of as its meaning basis, particularly if attention focusses on application. We are now ready to turn to perception. The givelz, the data, what we immediately perceive, both common sense and psychology tell us, is what we immediately respond to, either by verbal symbols or by imagery, or by both, with or without concomitant overt response. If the symbols thus called forth were distinguished, let us say, by a red mark, one could plot a perceptual horizon against the background of our tripartite scheme. Using one version of a mongrelized logico-psychological terminology, one could also say that the referents of the terms on the horizon are perceived, while the terms above it are consciously inferred. Another version would insist that the referents of the lowest plane alone are perceived and that everything above it is consciously or unconsciously inferred. Subtleties as to a differential perceptual (ontological?) status of the primitive names and the primitive functions are here neglected (Universals ! Intrinsic relations !) . This version plays upon the fact that by a conscious effort we are able, within certain limits, to 'press down' the perceptual horizon or, a t least, to 'reflect' upon it in such a reductive intent. With the average adult the horizon lies somewhere near and roughly parallel to the intermediate plane. But on the other hand, it is a well-established scientific fact that all responses, perceptual or otherwise, which are made to the referents of derived terms, i.e., derived with respect to the sense-data analysis, are something learned. Likewise, it is common-sense knowledge that training pushes the horizon upwards, even with adults, while we have some reason to assume that in sub-humans and in infants the response does not even reach the thing level. What 'epistemological' structure can be built upon such shifting ground? At this point another remark is needed to prevent misunderstanding. The foregoing observations on perception have not been offered as a scientific proof of a philosophical thesis. Such proof is impossible, for neither can science in a direct manner support epistemology, nor epistemology science. All they can and, indeed, must be expected to do, is to dovetail. To illustrate this figure: The circumstance that the sense-data analysis is the
G. Bergmann
8I
"deepest" known t o us, dovetails with the scientific fact that organisms derive all their factual knowledge from sensory cues and from sensory cues alone. T h e scientific problems of perception in particular can not even be adequately formulated, before it has been realized that language analysis and data are two completely different things. Fortunately the distinction is a t least implicit in the behavioristic methodological frame of contemporary psychology. If this frame is abandoned, however, difficulties immediately arise. These difficulties Gestalt psychology has never been able to overcome, and it certainly is significant that all the writers of that group are more or less heavily engrossed in an intuitionistic philosophical background. This approximately is their predicament: If the phenomenological term 'data' is used synonymously with the linguistic term 'sensualistic meaning basis,' certain expressions concerning data as, e.g., that we are aware of them, become pseudo-statements. This has been mentioned before (p. 78). If however, under the suggestive impact of unanalyzed everyday usage some meaning is read into them, they frequently turn out to be false, because there is, in normal perception, no perception (awareness) of the 'elements'. Hence, so it is argued, the alleged data are not the real data, and so forth, and so forth. This should suffice to indicate the complete epistemological irrelevance of the Gestaltist argument. I t s irrelevance for theoretical psychology has become apparent in the course of the recent methodological discussion within that field. T h e actual contribution of the students of Gestalt thus seems to be confined to their experimental findings. N o other opinion could ever have arisen, were it not for the very specific course of the historic development of ideas inside Germany. Attention must now be called to a certain usage. Reference has consistently been made to the empirical language and, in contradistinction, to the several possible analyses of it. Three circumstances sometimes obliterate the thesis underlying . - this usage. First, the one empirical language is, under the impact of discovery and changing theoretical interpretation, in a state of continuous flux and remoulding. Only relatively small fragments of it have actually become systematized, and the cement between
Scientific Empiricism
these bricks is still provided by skill rather than by articulate cognitive knowledge. We know as long as we are not asked, and when we are asked we do not know. These interstices of ignorance are the shelter for the 'residual irrationality' of the emergentist doctrines. T h e danger of rationalism, on the other hand, which lies in the fiction of the empirical language has been pointed out before. Socio-psychological connotations of the words 'rational' and 'irrational' have been played upon in the last remark. For it is the rationalist who pins his hope upon the residual irrationality of vitalism, who clings to certain obvious misinterpretations of modern physics, to Gestalt doctrines, and so on. The less these views dovetail with science, the more outspoken becomes the trend towards the frankly hierarchical ontology of Aristotle with the ablest among the modern rationalists. This has at least the merit of making the issue clear-cut. Therefore it must be pointed out again and again that there is no extra scientific stake in the vitalism-mechanism argument. Formally speaking, this argument concerns itself with the relation between two scientific calculi within the empirical language, not, as epistemology does, with possible language analyses, their interrelations, and the ensuing illumination of the basic features of experience, upon which such a study throws the sharpest light. Second, it is customary to refer to any particular analysis of the empirical language, or rather to the elucidations and the model reconstruction based upon it, as a language; in this sense there is a multiplicity of languages. Numerous instances of this convenient use will be found in the present paper. T h e third circumstance requires more caution. For the purpose of organizing theoretically parts of the empirical language, scientific calculi (hypothetico-deductive systems, system languages, theories) are used, and these, too, are sometimes called languages. Looked a t from a formal angle, any scientific calculus is indeed a selfsufficient symbolic structure very much like the empirical language, only much more articulate and much less comprehensive. Like the empirical language and any of its model reconstructions it also contains a logical calculus. I t is therefore not inappropriate to call it a language. But still any such theory becomes
G. Bergmann
applicable only through co-ordination (identification) of its terms xvith certain families of empirical constructs, and is consequently exchangeable with relative ease. Here is another manifold of languages, and again the word has a slightly different connotation. Let it also be noted that it is sometimes very clarifying to treat certain masses within the empirical language as if they were theories, even if they are far from exhibiting that articulateness and consistency which an actual calculus must possess; simply because they have a somewhat similar position within the total structure. The similarity consists in their being relatively isolable and tied together by mutual dependency like the empirical constructs of those masses which correspond to the terms of a genuine scientific calculus. The 'mentalistic language' is the most important instance of such a linguistic mass2 The reference to the mentalistic mass as a potential calculus provides the cues for the next point: the integration of logical behaviorism into the schema. I n this context it is important to note that the preceding observation about the theory-like closure of the mentalistic mass has as such nothing to do with logical behaviorism. This latter term has been coined in order to distinguish an epistemological thesis from questions of method within psychology. Insofar as it serves this purpose there is no reason to abandon it. I-Iowever, it must be emphasized that the core of so-called logical behaviorism is factual rather than logical in the strict sense of the latter term, as the expression itself could suggest. The main concern of logical behaviorism is the meaning analysis of the mentalistic terms, i,e., the tracing, at least in rough contours, of the introductory chains of the empirical constructs which belong to the linguistic mass I have called mentalistic. As a matter of fact, but as a matter of fact only, all these terms are located above the thing plane, if I may avail myself again of the visualization previously used. This is, after all, the very obvious factual nucleus around which the metaphysics of solipsism (consistent idealism) has developed. But this is also the incisive fact which justifies the thesis of physicalism, that the physicalistic language provides a satisfactory meaning basis for 2G. Bergmann, The Subject Matter of Psychology. Phil. of Sci., 7, 1940,415-433.
84
Scientific Empiricism
the whole of our scientific knowledge. Thus, the two theses of physicalism and logical behaviorism are closely related, though, as one sees, not identical. I t also becomes apparent that physicalism is essentially a methodological thesis. The importance of its meaning basis as the only objective one, as scientists use the expression, has been exhaustively discussed. With these considerations the development of the schema has reached a point where the so-called double-language approach to the psychophysical complex can be discussed. The two languages in question are, in fact, merely two layers or sublanguages, i.e., actually or a t least potentially, interpreted scientific calculi of the one empirical language. What the approach accomplislles can be subsumed under two headings. First, it vigorously and ably points at that feature which has been mentioned as the factual core of logical behaviorism in the preceding paragraph. Second, it develops certain analytical consequences of this fact. If the two types of introductory chains, those for the mentalistic and those for the physical terms, both start from the same basis on the intermediate plane, then the following must be the case: With an increase in factual, scientiJc knowledge the mentalistic mass will progressively approach the status of an actual theory and, by the same research, a more and more comprehensive connection between mentalistic and physical theory will be established. The expectation of such a psychophysical parallelism or, as Tolman very adequately and with less anticipation calls it, behavior-physiological parallelism is, indeed, not only a consequence of the meaning analysis of the mentalistic terms, but it also dovetails perfectly with contemporary science. Still, this parallelism is essentially an affair of facts and science. The more clearly this will be stated by Scientific Empiricism, the less will its philosophical opponents be tempted into dilettante disparagements of contemporary science. James' criticism of some of their counterarguments as bashful spiritualisms is not less to the point today than it was at his time. On the other hand, the epistemologist rightly resents it if a mere selection of facts and frames of reference is presented to him as the whole of the analysis he can perform, no matter how incisive and obvious these facts and
G. Bergmann
frames appear to both common sense and science. The doublelanguage approach, it seems to me, has at least invited such a misrepresentation. As a matter of fact, it is not easy to see how it could fend off the criticism that it is merely a 'formalization' of the brain-mind and shares the realistic bias of this earlier conception. Again, linguistic analysis permits one to locate the danger spot. The double-language approach tends to raise to the dignity of two empirical languages two calculi which are fitted to different masses of the coherent hierarchy of the one empirical language. These two 'languages' are then assumed to speak about the same. Such an unknown common referent is clearly indicative of a realistic bias, or rather, it is a vestige of a realistic ontology. The last argument hinges on the following distinction: The fact that the terms of the two theories have a common basis on the intermediate plane must not be interpreted as an ontological parallelism or identity; this in spite of the circumstance that certain consequences of this state of affairs can be spoken of, vaguely, as an intertranslatability or correspondence or, still more vaguely and rather confusingly, as an isomorphism. The Gestaltist doctrine of isomorphism, finally, is in its cognitive content about equivalent with the thesis of physicalism. Unfortunately, however, this nucleus is embedded into a holistic ideology and a fragmentary intuitionistic epistemology which prevent the students of Gestalt from an adequate grasp of scientific methodology even where no philosophical problem is involved. Hence the rather paradoxical fact that Koehler, himself the protagonist of a rather naive physicalism, directs the brunt of his criticism against what he conceives of as the elementaristic physicalism of the positivists. Moreover, it is only fair to say that such a scientific parallelism or intertranslatability, if these ambiguous terms have to be used a t all, is the common basis of all contemporary psychology and not the exclusive tenet of a particular school. If, however, the word 'isomorphism' means to suggest more, e.g., a structural similarity between the phenomenological protocol on the one hand and the physiological process on the other, then it must be said that our present knowledge by no means favors or even suggests such a hypothesis. This has been amply clarified by Boring. These remarks complete the scaffolding which is necessary for an organization of the psychophysical complex. I t is realized,
86
Scientific Empiricism
however, that a t present no attempt a t such an organization can yield more than a very rough, tentative, and even hazy sketch. Twofold ignorance against which we still labor makes for this limitation; ignorance in the science of psychology, and ignorance with regard to the analysis of the empirical language. T h e coexistence of these two lacunae is not incidental (Wittgenstein, Tractatus, 4 . 121). ~ One way of approaching the problem is to determine the syntactical status of the personal pronouns 'I' and 'You'. Clearly, the grammar of these terms belongs to that part of the language which "has been made analytical", and the study of it involves an epistemological clarification (p. 75). T h e following six points will indicate the thesis concerning the position of the pronoun '1': ( I ) I n a sense-data analysis the pronoun 'I' and its derivatives, as 'my', are derived terms; (2) the height of these terms in the hierarchy of the empirical constructs is roughly t h e same as that of thing-names; (3) 'subjective' and 'objective' are relational terms of the same level, as closely interwoven with 'I' as 'thing' is with the spatio-temporal relations; (4) word combinations referring to 'my data' are excluded; (5) it is not asserted that analyses in which 'I' is a primitive are impossible; (6) nothing is asserted as to whether or not 'I' lies on the perceptual horizon on which we live. T o tie up with previous remarks: T h e exclusion of the word combination 'my data' is but an application of the rule epitomized by the formula that nothing can be predicated about the data as such (p. 78); the thesis as a whole exposes the fallacy which is inherent in the fiction of the epistemological subject; comparison of what is and what is not asserted shows, first, that epistemology must include a study of all possible language analyses, and second, confirms that the distinction between analysis, theory of perception, and phenomenological description is of the essence of the argument. Further elucidations can be organized around three questions which suggest themselves a t this point. First: How are the opposite theses, the various forms of intuitionism and, in particular, Kantianism, to be formulated within the schema? It is relatively easy to indicate the intuitionistic
G. Bergmann
view. As far as I can see, it amounts to the assertion that 'I' will be found among the primitives of any possible analysis. T h e transcendentalist position is, naturally, much harder to translate. One might vaguely depict the transcendental Self as a basic term below the lowest plane of the sense-data analysis. Second: What light is thrown upon the objection of implicit solipsism? From his standpoint, the empiricist can neither assert nor deny an ontological position. But he can and should locate as precisely as possible what one might call, in a figurative way of speaking, the projection of such positions into his universe of discourse. There are, indeed, several points within the empiricist schema which together constitute a projected denial of solipsism: the claim that there is an analysis in which 'I' and 'You' are both derived terms and thus on a par with respect to primitivity; the exclusion within such an analysis of expressions which purport to predicate that the 'data' are mine; the applicability of a timespace-thing language, i.e., that projected assertion of the realistic thesis which has been referred to in the introduction; finally, those factual features within the psychophysical complex which will be mentioned presently. But there is also the fact underlying the thesis of logical behaviorism. Obviously it is this fact which accounts for the difference in height between the constructs 'I' and 'YOU' within the hierarchy of empirical constructs. Such rewording should clarify the pictorial use of the expression 'height'. This difference in height is undoubtedly the projected common-sense core of solipsism. I t is rather ironical, therefore, if the charge of materialism is based upon the behavioristic thesis or its physicalistic corollary. I n spite of this antimaterialistic bias of many of its critics, Scientific Empiricism should be particularly careful and explicit in its denial, loosely speaking, of the idealistic position; for if one considers the intrinsic affiliation of philosophical thought rather than to study it in the larger context of a universal history of ideas, contemporary empiricism seems nearer to non-romantic idealism than to the realistic-materialistic elements with which it has been saturated during its positivist and pragmatist period. Let it be said again that an analytical interpretation of ontological positions, as here attempted, is not itself
Scientific Empiricism
an ontological position, and that the last, purely historical remark is not an idealistic thesis. Third: What is the 'dovetailing' scientific evidence for the thesis? Modern Psychology, in its heroic efforts to obtain scientific knowledge rather than prescientific insights, finds itself for the most part confined to the animal laboratory and to the study of very elementary human responses. Little support, therefore, can be derived from it. On the level of psychological insight, however, the dovetailing evidence is impressive. Here is the place, where the whole argument about the Self, as raised and digested by the empiricist succession, can be inserted into the schema, from Hume's chapter on personal identity, through Mill, to Mach's elements and James' Thought-the-Thinker. If the modern empiricist fails to avail himself of this material, because he is imbued with the ascetic rigors of scientific standards of which those ingenious elucidations naturally fall short, he would deprive himself of the most convincing support. But one must never forget that both argument and counter-argument about the Empirical Self are psychological. Who mistakes them for epistemology proper falls victim to some form of psychological or genetic fallacy. Turning to the pronoun 'You', the schema naturally accepts the result of the behavioristic meaning analysis, here also referred to as the difference in height between the constructs 'I' and 'You', and pungently epitomized in James' simile of the automatic sweetheart. No attempt will be made to recommend this thesis by repeating well-known arguments; this not only because such an attempt goes beyond the limits of a schema, but also because the behavioristic thesis has become sufficiently elastic since the rigoristic use of the epithet 'meaningless' has been supplemented by an appreciative, pragmatic analysis of the nonreferential meanings of language. The one point to be made here is of a different nature. It concerns another factual feature which is not explicitly covered by the behavioristic meaning analysis. Neither, and more important, is this feature analytically contained in the thesis of logical behaviorism, as great brevity just on this point and the use of the adjective 'logical' could suggest. T h e
G. Bergmann
89
credit for having seen first that this factuality requires a very explicit acknowledgment as such, goes among all the critics of Carnap, as far as I know, to Felix Kaufmann. Assume that an observer truly describes his experiential flux by asserting, e.g., the compresence of a bright, red, cubic object and of a human organism, a subject, in a position to perceive this object. In this context, 'position to perceive' is, of course, a physicalistically introduced expression below the mentalistic mass. Such situations including the protocols of the verbal responses of the subjects are the material which suggest the introductory chains (definitions) for the terms of the mentalistic mass, including such terms as the second-type particular YOU'.^ More clearly, these situations yield the empirical laws which determine the grammar of the introduced terms as well as those empirical laws which are recognized and formulated as such. Let it be mentioned here again that the distinction between these two types of empirical laws, very familiar from scientific methodology, corresponds to what in other terminologies is called strata of experiencea4 The oversimplification of the illustration is obvious, and attention has already been called twice to the present lack of a developed behavioristic psychology of the higher processes. Cautiously worded the thesis will therefore only assert, that the following will be the case: (a) If the mentalistic mass is introduced and integrated into the empirical language in such a way, a part of the language reproduces in a certain sense itself, or as one might also put it, the language bends back into itself. T o explain this figure: Whenever the statement asserting the compresence of the bright, red, cubic object and of the perceiving subject is true, then the statement that the 'You' which has the subject in question That 'I' and 'You' are particulars of this type is theprojected assertion of the existence of minds. In a suggestive way of speaking, minds are as real as forces, atoms, and electrons, which are indeed constructs of about the same height as 'You'. See also the paper on "The Subject Matter of Psychology". In the terminology of that article, the bending back of the language is, approximately, both a factual and a formal isomorphism. The point now stressed is its factuality. See also G. Bergmann and K. W. Spence: Operationism and Theory in Psychology. Psych. Rev., 48, 1941,1-14.
90
Scientific Empiricism
in its introductory chain 'perceives' a bright, red, cubic, object, where 'perceives' is this time a term of the mentalistic mass, is also true. This is the meaning of the simile that the perceptual level, as a matter of fact, and with obvious modifications, reproduces itself. (b) If the observer makes his own organism the subject of the experimental situations, then his protocols will, as a matter of fact, not deviate from the protocols he has obtained from other subjects. The introductory clause 'will be the case' and the repeated emphasis of the inserted clauses 'as a matter of fact' contain the gist of the argument. These are of course the same features which account for the appeal of the so-called argument from analogy. They have also been anticipatorily enumerated among the projected denials of solipsism (p. 87). Finally, they account for the circumstance that physicalistic analysis happens to be, in a certain, more than scientific sense, exhaustive. Thought experiments illustrating the factuality of these features offer no particular obstacles to our imagination. Carnap's blind physicist who discovers light or, similarly, the only human observer in a Martian world which gets its 'sensory cues' from magnetic phenomena only, are points in case. As a matter of fact, the imaginary man on Mars would be alone, but he would have no reason to become a solipsist, though solipsism might, as a metaphysics, appeal to him more than it does in a world of ours. Only a few remarks remain to be made. T h e syntactical expression of the last mentioned two basic facts is the machinery of conjugation which allows for the composition of any of the pronouns with the same verb. As far as the perceptual verbs like 'seeing', 'hearing', 'feeling' are concerned, a particular difficulty can be traced back to what has been called the bending back of the language into itself, that de-facto identification in common usage of two distinct layers of the empirical language. Because of this circumstance it seems natural to use those verbs in order to differentiate between 'things' and '1', and it is then not easy to see that this differentiation is based upon structures below the intermediate plane. These structures, together with the three facts support the introduction of the perception verbs and cannot,
G. Bergmann
91
therefore, be analyzed in terms of them. Finally, another relevant aspect of all three factual features, the two now under consideration and the third one, which supports the behavioristic thesis, shall a t least be mentioned. It seems to have played a certain r61e in recent discussions. I am referring to the factuality of the individualization which makes for the truth of the statement: 'My experiences are not yours, and yours are not mine'. This statement is merely an illustration, and 'experience' in it a derived psychological term not to be confused with 'data' which, as we have seen, is not a good expression to use a t all. What i t is supposed to illustrate is this: I n an a n t world in which 'I' could feel 'your' pain and vice versa, the words 'I' and 'You' would possibly not occur or, if they occurred, would not mean what they mean in our language; their syntax would be different.5 At this point thought becomes hazy. State University of 10wa 5This shift accounts also for the inevitable failure of all attempts to raise the traditional argument for the existence of other minds to the status of a hypothesis. Such an attempt has again been made in Ayer's last book, Foundations of Empirical Knowledge.