in Pacific Philosophical Quarterly 61 (1980): 3-30.


Wilfrid Sellars

I. Behaviorism as a Program

1. Many who went through the behaviorist revolution -- whether as more or less innocent bystanders, or torch bearers -- and watched it settle down into an established orthodoxy dominating the scene from the commanding heights of learning theory, have been puzzled by subsequent developments. Although nothing as systematic as a counterrevolution has occurred, free thinking is rampant. And if closely knit groups are not lacking who seek to restore law and order and put psychology on its proper path, they look to a variety of models.

2. Some of the ideas which have come to the surface in this period of anarchy would startle, indeed shock, the pioneers of behaviorism. They would find them of a piece with what they had regarded as modes of explanation which either don't explain or are so messy that it is difficult to determine exactly what did the explaining, how it explained and what was explained.

3. They had sought a methodology which would remedy this situation by generating hypotheses which were intersubjectively confirmable, and the content of which was formulated in such a manner that the nature and locus of its explanatory power would be manifest.

4. I emphasized above that what was sought was a methodology. This is not to say that the behavioristic movement was lacking in substantive commitment concerning the objects -- docile organisms -- which constituted its scientific domain. Clearly it was anti-dualistic -- even physicalistic, to use a contemporary term -- in its orientation, fully expecting that the application of its methods would yield results which would harmonize with a thoroughly naturalistic conception of the world. Yet as it became increasingly aware of its primary methodological character, behaviorism placed more and more emphasis on the autonomy of psychology as a science.

5. For, and this was an important insight, even when, as in the case of chemistry, one is prepared to envisage an identification of chemical objects, processes and properties, and properties with complex micro-physical objects, processes and properties, and, in this sense, unify the ontologies or domains of the sciences of physics and chemistry; to achieve this goal would not be in any useful sense to unify the sciences. For an essential part of a science, e.g. chemistry, is its methodology -- its mode of access to its subject matter, including its experimental apparatus and the structures of concepts in terms of which its questions are formulated.

6. Methodological autonomy of chemical inquiry -- of chemistry as a science -- serves as a powerful constraint on the form a unification of the theories of the two sciences might take. For it is chemical objects and processes as conceptualized in the methodological context of chemical theory which must find counterparts in the domain of micro-physics, conceptualized in the context of the methodology of micro-physics, if the envisaged unification is to take place.

7. Thus, while Behaviorists undoubtedly cherished the vision of a sometime identification of psychological objects, processes and properties with complex physiological and neuro-physiological objects, processes and properties, they fought a determined fight to preserve the methodological autonomy of psychology as a science. Any unification of the subject matters of these sciences would have to harmonize with the results of psychological investigation, achieved in accordance with the methodology of psychology as a science.

8. Thus, while there was, so to speak, an 'ideological' conviction that the results which would be achieved by a behavioristically oriented psychology would harmonize with a thoroughly naturalistic or physicalistic conception of the world, there was an equally strong conviction that psychology had its contribution to make to any adequate conception of the world as including sentient organisms.

9. But what, specifically, was the methodological program of Behaviorism? and how, specifically, was it motivated? I will take up the second question first. Here the central point is that although the Behaviorist did not deny that each human subject has an access to his or her psychological states which is not a matter of inference from publicly observable features of their bodily states, this 'privileged access' is neither clear, distinct, adequate nor infallible. Introspection is not a simple transposition of psychological reality into the cognitive order; a direct apprehension of the facts as they are. It is a conceptual response to psychological states and the concepts included in this response are common sense psychological concepts and, as such, no more adequate to an understanding of what is really going on than are common sense concepts pertaining to the middle sized physical objects of everyday experience. Common sense psychological concepts are an ill-disciplined manifold; vague, open-textured, controlled by obscure analogies and subject in their first person use to special sources of error.

10. As for explanations in common sense psychological terms, they are at their best in connection with human activity at a level on which people approximate their potentiality as rational animals. The more rational people are, the easier it is to explain what they do. The more difficult task is that of finding, classifying and explaining lapses from rationality. Explanations in terms of such categories as belief, desire, aversion, hope, fear, perception, memory, ends, means -- and such lapses as are implied by impulse, wishful thinking, suppressed emotions, carelessness -- are often satisfying. They certainly assimilate the problematic to the familiar.

11. There is nothing intrinsically wrong with the practice of classifying phenomena to be explained in terms of the categories of the theory which explains them -- in this case the theory of rational action and lapses therefrom. But where, as here, the theory has evolved in a poorly understood way by a kind of natural selection, the dangers of circularity and emptiness are ever present.

12. A darker side of the picture begins to emerge when we consider psychological facts which, though equally in need of explanation, do not fit -- in any neat way -- into the above framework. To take an example which is very much in the news -- How does an infant learn a language? Can an answer to this question be given in the categories of common sense psychology?

13. What might such an explanation be like? 1 shall not attempt to answer this question in any detail -- though I shall be concerned with the topic of language acquisition at a later stage of my argument. For the moment I shall limit myself to pointing out that there is available at the level of common sense the model of hypothesis formation and the search for confirming (or disconfirming) evidence.

14. According to this model the infant -- confronted by salient linguistic configurations of sounds, formulates increasingly subtle hypotheses about them and the contexts in which they occur; and accepts, rejects or modifies these hypotheses. Each new occasion provides more grist for the child's inductive mill.

15. Now it leaps to the eye that this kind of explanation assumes that the child is no merely potential rational animal. Its rationality is full fledged. It operates with concepts and logical forms which have a high degree of sophistication. Of course a common sense explanation of language acquisition along these lines would undoubtedly attempt to minimize the rationality of the child by emphasizing the sporadic, careless, undisciplined character -- in short the childishness of the process -- thought not to the extent of completely undercutting the applicability of the question, hypothesis, weighing evidence, acceptance, rejection, modification model.

16. If the question is now raised -- How does the infant acquire the conceptual abilities mobilized by his inductive reasonings? one answer is clearly precluded: "in the course of learning a language." Nor could these abilities be acquired by a process of inquiry or reasoning. The pressure towards an innatist account of the child's logical powers would be almost irresistible.

17. Notice that the Behaviorist need not claim that common sense does not contain the resources for a non-rationalistic account of the process of language acquisition. He is simply pointing out that whether or not it contains such resources they are not mobilized and deployed by common sense, and that the construction of such an alternative non-rationalistic account would require the deliberate and systematic development of a radically different framework of psychological explanation.

18. What might such an alternative-explanatory framework be? Before answering this question the Behaviorist turns to a second example of the unsatisfactory character of common sense modes of explanation. He does so because although the idea that the child learns to understand and use a language by a process of inductive reasoning within a framework of innate logical abilities strikes him as flawed by circularity -- in that the acquiring of a language is explained by postulating an unacquired (innate) languages-like structure -- he must admit that the account is not absurd. There is nothing self-contradictory about the idea of an innate language-like structure (Mentalese), and there might just be no other way of explaining language acquisition.

19. A second example is more immediately promising. It suggests both a methodology and a pattern of explanation. It put behaviorism on the track which, largely, it still follows. For it concerns the phenomenon of animal learning.

20. A common sense explanation of the acquisition by a child of the abilities which animals acquire in experimental situations (e.g. learning a maze) can be expected to be of a piece with its explanation of language acquisition. But what of animals? Does it even make sense to attribute to animals procedure of hypothesis formation and confirmation which employs an innate conceptual structure? Perhaps it makes sense -- but barely. The pressure to find another mode of explanation is overwhelming. As for methodology: in the nature of the case it could not be introspection.

21. As for the pattern of explanation, the Behaviorist came up with stimulus-response-reinforcement theory, which works reasonably well in simple, tightly controlled cases. The essential features of this explanatory pattern appeared at an early date in Thorndike's explanation of how cats learn to escape from specially contrived boxes.

22. The hope, of course, was that if this mode of explanation works in simple cases, in connection with animals, it would also work in complex cases, and then again in simple cases in connection with humans and finally in connection with complex human cases. Language learning might be explained in terms of complex structures of S-R connections, each of which was stamped in by a reinforcing reward.

23. As for methodology -- the study of animal behavior was particularly well suited to an experimental technique in which questions and hypotheses are formulated in terms of publicly observable variables.

24. Today it is obvious at a glance that in a sense the Behaviorist methodology was unnecessarily restrictive. Once it became clear that the concepts of physical theory could not be explicitly defined in terms of observables, the Behaviorist was open to the challenge -- why is it not legitimate to introduce concepts in Behavioristics which are not explicitly definable in terms of observable behavior described in carefully aseptic (non-anthropomorphic) terms? As it was sometimes put: Why not introduce hypothetical constructs into Behavioristics (compare the postulational concepts of physical theory) instead of restricting yourself to intervening variables (the mark of which -- though the terminology was misleading -- is definability in terms of observables)? To be sure, in the resulting controversy the phrase 'intervening variables' was often used with sufficient looseness by its proponents that it could cover 'variables' which were not, strictly speaking, definable in observation terms. The latter, however, were treated in an instrumentalist spirit, as merely 'mathematical,' and denied psychological reality. Thus, in a sense, the issue was not clearly joined. Nevertheless the methodological point of the stress on the concept of an 'intervening variable' was to block a return to Mentalistic concepts.

25. For this was indeed the real issue. The Behaviorist might grant that he needed concepts which could not be defined in observable terms. But (a) he would go as far as he could without them; (b) he insisted that the above concession not be construed as opening the way to a free use of Mentalistic concepts. A priori restrictions on the kinds of concepts to be introduced into psychological theory were out of place. But from the standpoint of methodology the binding principle was to be: Don't simply borrow concepts and principles from the framework of introspective knowledge. Use all the analogical and suggestive power of Mentalistic concepts and principles, but be sure that the concepts and principles you introduce have no more Mentalistic structure than can be justified in terms of their ability to explain observable behavior phenomena.

26. As I see it, this was -- and remains -- the methodological stance of a sophisticated behaviorism. And while it is not as restrictive as the outlook which is typical of the more orthodox forms of the movement, and can be subscribed to by many who are uncomfortable about the historical connotations of the term 'Behaviorism,' it by no means trivializes the latter. It formulates methodological policy which is so widely shared as to justify the statement that Behaviorism is indeed alive and well in psychology.

II. Language and Thought: A Behaviorist Strategy

27. Thinking as an activity is often contrasted in a global manner with perceiving and doing. The paradigm for this sense of the verb 'to think' is problem solving.

28. But in another sense thinking is itself a constituent of perception and action, and is, indeed, but a proper part of problem solving.

29. The term 'perception' ranges over a spectrum of psychological states, but in its most developed form, by analogy with which we attempt to understand the others, it clearly includes a conceptual element, an awareness of something as being of a certain kind, as having qualities and standing in relations. This conceptual element was traditionally regimented by the term 'perceptual judgment,' e.g. the judgment: "This is a red apple."

30. Perception also involves a sensory element, to which the perceptual judgment is a response. In what, exactly, this sensory element consists and how, exactly, these two elements are related are central topics in perception theory on which much remains to be said.

31. Thinking is also essentially involved in action. One can, of course seek to solve problems about what to do. But, even in impulsive action, there is a conceptual element. One conceptualizes an action as something to be done here and now. This conceptual element was traditionally regimented by the term 'volition,' e.g. the volition: "Now I will signal a left turn."

32. Actions are responses to volitions, as perceptual judgments are responses to sensory stimulation.

33. I have explored on other occasions the various dimensions of perception theory and action theory from the standpoint of a relaxed behaviorism.1 On the present occasion I shall concentrate on conceptual activity as such, abstracting as far as possible -- which, of course, is not very far -- from its involvement in perception and action.

III. Thinking and Speaking: The Classical View

34. The classical theory of mental activity recognizes analogies between certain properties of conceptual states and certain properties of the linguistic utterances which express them, judgments as well as sentences have subjects and predicates. Both have logical form and quantificational structure. Both involve distinctions of tense and of mood.

35. It is characteristic of the classical theory of mental activity to regard the syntax and semantics of conceptual episodes as primary, and those of linguistic episodes as derivative. The latter have the grammar they do because this makes it possible for them to express thoughts having what might be called an intrinsic grammar, i.e. a grammar, the possession of which is not to be explained in terms of the grammar of anything else.2

36. The classical theory takes there to be an intersubjective domain (the exact ontological status of which it usually does not attempt to specify) of what might be called thinkables. There is, for example, the thinkable

that 2 + 2 = 4.
When a person thinks that .2 + 2 = 4., he or she is in some way directly related to the thinkable. If, as seems legitimate, we speak of the mental term of this relation as an awareness, and continue to think of conceptual episodes as having structure, we might try to capture the grammatical character of the awareness as a matter of its being of the 2 + 2 = 4 kind. We would be baptizing the awareness by putting to special use the form of words which, in English, express an awareness of this kind.3

37. The classical theory construes the relation of utterances to thinkables as the logical product of a relation between utterances and thoughts on the one hand, and the above characterized relation between thoughts and thinkables on the other.

38. Emphasis is placed on the role of language in communication. Language is viewed as essentially an instrument. Furthermore, linguistic activity consists essentially of actions. Thus, although communication (and deceit) may be impulsive, we understand it best when we understand it at its best, i.e. in cases where one deliberates about necessary and sufficient conditions for realizing one's intention that a hearer becomes aware of and, perhaps, believes certain thinkables.

39. Classical theory at its best is sensitive to the distinction between act as action and act as actuality. Actions are actualities, but actualities are not, in general, actions. Actions, to follow Ryle, are those actualities which appropriately can be spoken of as deliberate or impulsive, careful or careless, etc.

40. A feeling of pain is an act by virtue of being the actualization of a potentiality, but it is obviously not an action. It makes sense to speak of a passive act, but not (save in a highly derived sense) of a passive action.

41. Mental acts, in the classical sense, are not, as such, mental actions. There are, of course, such things as mental actions, e.g. deliberating about what to do or trying to solve a mathematical problem.

42. And, of course, there are mental non-actions. Thus one can not deliberately notice something -- though one can deliberately pay attention. Again, it is notoriously a source of paradox to treat volitions as actions. Nor, when relevant distinctions are drawn, is inference properly subsumed under action.

43. Obviously, therefore, if all linguistic activity fell in the category of action, there would necessarily be some thoughts which could not be identified with linguistic episodes. This consideration has served to reinforce the classical view that thought is essentially non-linguistic, and that the syntax and semantics of thought (its intentionality) is primary.

IV. A Contrasting Strategy: Verbal Behaviorism

44. 1 have already called attention to the fact that classical theories recognize "analogies between certain properties of conceptual states and certain properties of the linguistic utterances which express them."

45. The limiting case of a theory which recognizes such analogies is one which construes them in terms of identity. According to it, grammar of conceptual episodes is analogous to the grammar of linguistic episodes because conceptual episodes simply are linguistic episodes.

46. According to this theory, which I have called Verbal Behaviorism,4 thought episodes are, in the first instance, candid linguistic utterances by one who knows the language. Let me call them 'thinkings-out-loud' -- though they could, of course, be in the vocabulary of the deaf and dumb.

47. Again, in the first instance, they are linguistic non-actions -- as contrasted with the illocutionary and perlocutionary acts (actions) stressed by those who look at language through the spectacles of communication theory.5 There would be noticings-out-loud, inferrings-out-loud and willings-out-loud, all of them actualizations of linguistic abilities, but none of them actions.

48. There would also, of course, be linguistic actions -- not only such other-regarding actions as Austin so lovingly botanizes, but also such actions as deliberating-out-loud and attacking-mathematical-problems-out-loud.

49. The concept of thinking-out-loud is, according to the verbal behaviorist, the primary concept pertaining to conceptual activity. All other concepts of thinking are to be understood in terms of their connection with it.

50. Primacy in this sense is not to be confused with primacy in other senses. Compare the primacy 'in the order of knowing' of the concept of a middle sized perceptible physical object with the primacy 'in the order of being' of concepts of micro-physical entities.

51. The verbal behaviorist insists that he doesn't invent this concept out of whole cloth. We all have this concept, but, as is our philosophical wont, confuse it with other concepts when we attempt to theorize.

52. But if thinking-out-loud is the primary concept pertaining to conceptual episodes, not every concept of a conceptual episode is the concept of a thinking-out-loud. There is, in the second place, the concept of a proximate -- a 'tip of the tongue' -- propensity to think-something-out-loud. Such propensities amount to subjunctive conditionals as

If Jones were in a thinking-out-loud frame of mind, he would think-out-loud: "the bus didn't stop."

53. Such propensities are episodic in the sense that one comes to have them and ceases to have them.

54. Needless to say propensities can be remote as well as proximate, and remote to various degrees. Objects can acquire propensities pertaining to the acquisition of other propensities. One thinks of a nesting of conditionals. Magnets have the propensity to attract iron filings. Iron has the propensity to become magnetized in certain specifiable conditions.

55. Thus there is space for complicated structures of conceptual episodes, construed as shifting propensities of various degrees of remoteness pertaining to thinking-out-loud. To grasp these structures one need only reflect on their counterparts in the classical theory of mental activity -- for, if the verbal behaviorist is right, the shifting sands of mental activity, with which we are so commonsensically familiar, simply are the shiftings of proximate and remote propensities pertaining to thinking-out-loud.6

56. Verbal Behaviorists are inevitably confronted by the following off-the-cuff objection: "Surely thinking-that-p can't be just a matter of uttering 'p.' One isn't thinking out loud that-p unless one knows the meaning of the sounds one utters, and this knowing the meaning isn't a matter of utterances and propensities to utter. It is a genuine conceptual act which alone makes the utterance a thinking out loud."

57. The answer, of course, is that there is all the difference in the world between parroting words and thinkings-out-loud. In the latter the utterances cohere with one another and with the extra-linguistic context in accordance with the syntactical and semantical rules of the language. 'Knowing the meaning' of linguistic expressions is, in Ryle's terminology a case of knowing how, and is to be carefully distinguished from 'knowing the meaning' in the sense of being able to make correct lexicographical statements about them.

58. A more insightful criticism is that whereas it is a familiar fact that our thoughts succeed one another with lightning-like speed, overt speech moves as slowly as the tongue. Furthermore, it is argued, we can think two thoughts at once, but speak only one. How can this be reconciled with verbal behaviorism?

59. The answer is that propensities can come and go with the necessary speed, The propensity to think-out-loud 'It is raining' can come into being at the time it begins to be actualized. It continues to exist as it continues to be realized. Here it is important to note that the continuing propensity should not be identified with a sequence of separate propensities to utter constituent syllables, though it includes them. Again, propensities for different behaviors can coexist even though they cannot be simultaneously realized.

60. It might be objected that much of our thinking does not seem to take place in words. We often know that we have been thinking and, indeed, what we have been thinking, although no words have 'gone through our mind.'

61. Answer. Just as to be flammable is not to contain a hidden flame, nor an electron's propensity to jump to another orbit, a concealed jumping, so a propensity to verbalize is not an 'inner' verbalizing. It may be accompanied by verbal imagery, but does not require it, and is certainly not identical with it.

62. The verbal behaviorist may elaborate this point into a general counter attack by arguing that something like the above confusion would account for the attractiveness of the classical theory. The latter, after all, construes thoughts as 'analogous' to, but not identical with, linguistic episodes. Can this not be understood as an attempt to put into one picture of thoughts as episodes, features which they have as episodes of thinking-out-loud and features which they have as episodic propensities to think-out-loud, i.e. features belonging to radically different categories?

63. Perhaps the strongest argument for verbal behaviorism is that it makes clear how we know about thoughts. For thoughts as thus construed are, in their primary mode of being, publicly observable episodes of thinking-out-loud, people non-parrotingly saying something.

64. When we overhear Jones muttering 'I'll get him if it's the last thing I do,' we are literally hearing him plot revenge.

65. We can also hear ourselves think-out-loud. Yes, it will be said, but how does the verbal behaviorist account for the fact that we can know what we are thinking, when we are not hearing ourselves think-out-loud? How is the awareness of a propensity to think-out-loud to be understood?

66. The first part of the story, of course, is that we can have publicly available evidence of the existence of propensities. For example, one can infer from such behavior as x's gritting his teeth, turning pale and clenching his fists, that he has propensities to lash out in the manners characteristic of rage.

67. Assume that anger simply is a togetherness of such propensities. There would seem to be no problem about how one could come to be able to infer that a person (including oneself) is angry.

68. But what about the non-inferential knowledge we can have that we are angry? The answer calls for a quick glance at language acquisition. Consider the case of the child who already has the ability to use sentences of the form "x is angry" in a way which mobilizes publicly observable criteria. The child has "I am angry" in its repertoire, but so far only for inferential use.

69. Now the child's parents can know inferentially when the child is angry. And they can, by behavior toward the child of a kind which neither they (nor the child psychologist) need clearly understand, but which involves something like the methods of positive and negative reinforcement, bring it about that the child responds directly to his anger propensities by uttering "I am angry."7

70. The child acquires the ability to think-out-loud 'I am angry' in the non-inferential mode.

71. Notice that the child isn't acquiring the propensity to say "I am angry," when he notices his anger. This would put the cart before the horse. The noticing simply is the actualization of the acquired propensity to say 'I am angry' as a direct response to the anger itself.

72. This strategy applies, mutatis mutandis, to the acquisition of the ability to notice one's propensities to think-out-loud.

V. Behaviorism and Language Acquisition

73. I have already emphasized, in the opening section of this essay, that the child does not acquire a language by forming hypotheses about the correlation of the verbal episodes to which it is exposed, with each other and with non-verbal features of its environment -- any more than does a rat when learning a maze.

74. There are, of course, analogies between

(a) the ability to respond to objects in ways which discriminate between different degrees and kinds of similarity and difference
(b) the ability to formulate hypotheses which involve generic concepts.
(a) is more primitive than (b), and the latter would be impossible without it.

75. The rat does not form the general concept of a triangle when it acquires the propensity to jump at a trap door on which is inscribed a triangle of some kind or other.

76. Yet it jumps "as though" it were a rat-shaped homunculus who has said to himself "I will jump at the one with a triangle on it."

77. One can call such an ability to discriminate degrees of similarity and difference, "having the concept of a triangle." But this is a denomination by analogy which papers over essential differences.

78. Again, there are analogies between basic inductive reasoning and certain forms of operant conditioning. Thus inductive reasoning can contribute to the adequacy of the manner in which we represent our environment. It enables us, so to speak, to construct more complete and more accurate maps.

79. In this respect it plays a role which is analogous to that played by the maze-learning behavior which leads a rat to form a "cognitive map" of its environment.

80. It is important to think of language acquisition as involving the reinforcement (and the extinction) of whole strings or patterns of behavior rather than as a matter of setting up isolated stimulus-response connections and then stringing them together.

81. Too much emphasis on responses to environmental stimuli can have as its consequence an over-emphasis on connections between words and things -- as contrasted with the acquisition of togetherness patterns (syntactical patterns, including con-sequential patterns) pertaining to strings of words in the acquisition of which the character of the non-linguistic environment has been made irrelevant.

82. I obviously have in mind the acquisition of sentential and inferential patterns.

83. Acquiring negative uniformities, propensities to avoid certain modes of togetherness, is obviously just as important as acquiring tendencies to favor positively specified modes of togetherness.

84. By giving proper stress in one's conception of language acquisition to the acquisition of positive and negative linguistic propensities which have no direct connection with specific environmental variables, one will be in a position to recognize from the start both the possibility and the importance of environment-free (as contrasted with 'stimulus bound' or 'tied') linguistic behavior.8

85. 1 have long stressed that inference patterns -- including the patterns characteristic of causal inference and explanation -- are at least as essential to verbal behavior and, as we shall see, the meaning of linguistic expressions, as are word-object connections.

86. The emphasis on word-object connections and, in general, on those features of verbal behavior which involve responses to environmental stimuli, can lead to formulations which are particularly annoying to the humanistic ear.

87. Thus one may come to think of verbal behavior as 'caused' or 'controlled' by external events.

88. The word 'cause' is a notorious source of philosophical confusion; particularly when construed as on an interventionist model -- transeunt as contrasted with immanent causation.

89. The trainer does, indeed, intervene and is, in a legitimate sense, one of the causes -- one of the contributing factors -- in language acquisition. But the child is another. It is the source of the behavior which gives trainers clues as to what to do next.

90. The upshot of the training is that the child becomes a thinker, exhibiting rational discourse in sequences which are not fixed responses to intra-organic episodes. It is important to note that while, if scientific determinism is correct every item of verbal behavior has a causal explanation, this must not be confused with the idea that every such item is a conditioned response to a stimulus.

91. Thus the inferential sequences involved in rational discourse would have causes. But the correct description of these cause-effect connections would, to the extent that they are available, belong in a conceptual framework,'below' that of behavioristics.9

92. The child, so to speak, is trained to be free.

93. Note also that to speak of the fruits of language learning as a system of 'habits' is to use a misleading metaphor. The child does not acquire a system of dispositions and propensities, and it is tempting to use the term 'habit' in this connection.10

94. But not everything the child acquires is a disposition. It also acquires abilities -- thus the ability to think. And only the above noted confusion between the metaphysical thesis that every linguistic episode has (in principle) its causal explanation, with a substantive (and mistaken) claim in behavioristics to the effect that every linguistic episode is a conditioned response to a stimulus, would inspire the idea that all verbal behavior is the manifestation of habit.

95. An aphorism to the effect that man as thinker is entirely a creature of spontaneity would, of course, be equally false.

96. According to verbal behaviorism, properly understood, man as thinker is both a creature of habit and a creature of spontaneity.11

97. Finally, it is of equal importance to conceive of the acquisition of the language of action as developing pari-passu with the acquisition of the language in which the environment is described.

98. The child must learn to represent itself and its actions as well as its environment.

99. It would be a serious mistake to suppose that a language is learned as a mosaic is constructed: first language about things, then language about self, then . . . ; or first the object language, then a meta-language . . . ; or first descriptive expressions, then logical words, ....

100. The language learner gropes in all these dimensions at once. And each level of achievement is better pictured as a falling of items belonging to different dimensions into place, than as an adding of new rooms and stories to a building.

VI. Behaviorism and Awareness as

101. An awareness of an item as green is a response to the item as green. But it isn't an awareness of it as green simply by virtue of being a response to it as green.

102. An iron filing can be said to respond to a green magnet as a magnet. It doesn't respond to the magnet as green, and, indeed, it would respond in the same way if the magnet were of any other color.

103. But though the filing responds to the magnet as a magnet, we don't say, except in a metaphorical way, that the filing is aware of the magnet as a magnet. I shall refer to the sense of 'responds to as' in which the filing responds to the magnet as a magnet, as the causal sense of this phrase.

104. By virtue of what is a response an awareness? The problem is clearly a special case of that of distinguishing, within a naturalistic framework, between natural signs and semantical signs.

105. As a minimum we can say that to be an awareness, a response must be a manifestation of a system of dispositions and propensities by virtue of which the subject constructs maps of itself in its environment and locates itself and its behavior on the map.

106. Such representational systems or 'cognitive cartography' can be brought about by natural selection and transmitted genetically, as in the case of the 'language' of bees.12 Undoubtedly a primitive form of representational system is also an innate endowment of human beings. The concept of innate abilities to be aware of something as something, and, hence, of pre-linguistic awarenesses, is perfectly intelligible.

107. People obviously have innate psychological abilities. Not even the most radical empiricist has ever denied this. The proper empiricist stance has been to keep innatenesses and their degree of sophistication to a minimum.

108. Primitive representational abilities show analogies to the more sophisticated forms which became available with language acquisition (otherwise they wouldn't be representational). But while they may, to use Leibnitz' phrase, ape reason, by achieving results which can be achieved by reason, they should not be construed as more akin to reason, than it is necessary to suppose them to be for the purpose of the explanation of behavior. Thus it is only in extremis that one would ascribe to animals and infants the assignment of degrees of confirmation to general hypotheses.

109. An interesting case of responding to something as something is that of a rat which has acquired the propensity to leap at panels with varieties of triangles painted on them. As in the case of the filings and the magnet we can describe the rat correctly as responding to triangles as triangles. But we should be cautious about concluding that simply by virtue of this training, the rat has acquired an ur-concept of triangle. A much greater degree of integration of this pattern into the rat's mapping behavior, and its satisfaction of its needs by goal directed activity, would be necessary to justify this move.

110. I have long argued that our concept of inner representational episodes ('thoughts') is an analogical extension of our notion of meaningful linguistic episodes. I have distinguished between the problem of how we come to be able to think of 'inner episodes,' and that of how we come to be able to have 'inner episodes.'

HI. Notice that at this point we have gone beyond verbal behaviorism proper as we have hitherto characterized it. The inner conceptual episodes I now have in mind are no longer proximate and remote propensities to think-out-loud. They belong to a theoretical framework which purports to explain the comings and goings of verbal propensities in terms of finer grained structures, as micro-physical theory explains the powers and propensities of middle-sized objects. Nevertheless, the episodes postulated by the theory are, at least initially, taken be analogous to verbal episodes.

112. The substantive framework of verbal behaviorism is now viewed as methodological prelude to a full-blooded theory of representational systems, much as macro-physics is a methodological prelude to a unified theory of the behavior of physical things.

113. It is useful to view this extended theory of conceptual episodes as a placeholder for the future achievements of a neurophysiological theory of representation.

114. As it currently exists, this theory of 'inner representational episodes' has little more explanatory power with respect to conceptual thinking at the verbal level than does substantive verbal behaviorism. Its primary structure is essentially that of the classical theory, though it is self-consciously formulated as a theory, not a report of Cartesian givens.

115. This freedom from Cartesian methodology, however, enables it to make a unique contribution. It attempts, with increasing success, to make sense of the concept of more and less adequate representational systems, and to view the sophisticated structures of the classical theory as belonging to a hierarchy of representational systems, the lower forms of which are only remotely analogous to linguistic structures.

116. Thus viewed from the standpoint of methodology, verbal behaviorism is perfectly compatible with the idea that there are pre-linguistic representational activities.

117. Indeed, the ability to have primitive representational episodes might be not only pre-linguistic, but innate. On the other hand, ability to represent these representational episodes with any degree of adequacy might presuppose a degree of sophistication which comes only with the mastery of language.

118. What about the ur-awareness of representational states? Here the important thing is not to suppose that an ur-awareness of a representational state which is, in the causal sense, a response to it as an inner representational state, must also be an awareness of it as an inner representational state.

119. A child can have an ur-awareness of an ur-awareness which is adequate to play the role of self-awareness in primitive mapping activity, without being aware of it as a 'mental' activity; indeed without being aware of it as an 'inner' as contrasted with 'overt' representational activity. This contrast is a highly sophisticated one.

120. Thus an ur-representation of ur-representations as representations should not be supposed to represent them as judgements, or inferences, as involving singular terms, predicates, or modalities. Just as animal representational systems are, so to speak, minimal, so the animal's representation of its mapping activities, its meta-representations are equally primitive or minimal. Analogous though it is, in certain respects, to meta-linguistic activity, the analogies should not be pushed beyond necessity.

121. The myth of the given is hovering around because it has been tempting to suppose

(a) that the characteristic features of developed representational activity -- formulated in terms of the categories of intentionality -- are present in any activity which is worthy of being called representational.

The truth of the matter is that the generic concept of a representational system admits of many gradations between primitive systems and the sophisticated systems on which philosophers tend to concentrate.

(b) that a response to representational activity as representational activity must involve an adequate awareness of it as the kind of representational activity it is.

Thus the Cartesian is tempted to think that an awareness of a mental state is a direct grasp of its character as the specific kind of mental slate it is.

122. A basic form of the Mvth of the Given is the idea that

if we are directly aware of an item which in point of fact (i.e. from the standpoint of the 'best explanation') has categorial status C, then one will be aware of it as having categorial status C.

123. Closely related to this is the idea that

the 'directness' of an awareness, its confrontation with that of which it is the awareness, guarantees that if the confronted item has the property of being φ, and if the awareness is a response to the item as, in the causal sense, φ, then the awareness is an awareness of the item as φ.

124. This idea easily develops into the givenness account of concept formation. According to the latter one acquires concepts of perceptible qualities and relations from experiences of being directly aware of instances of these qualities and relations

125. This account has had disastrous consequences for the theory of meaning. According to it, the concept of φ is, so to speak, a conceptual atom. The one-on-one picture of apprehension and item apprehended leads to the idea that inferential relationships, and, in particular, non-logical or 'material' inferential relationships are not essential features of basic concepts.

126. Thus, although it is obvious that being (an expanse of) red implies not being (an expanse of) green, the above picture suggests that the concept of red has no intrinsic connection with this fact.

127. No one would say that the fact that a certain conceptual item stands for negation is independent of inferential relations to items standing for other logical connectives. It is obvious that the meaning of logical words, the aboutness of logical concepts, essentially involves these inferential relationships.

128. And exactly the same is true of basic concepts generally, a fact which is obscured by the above picture of concept formation.

129. I have been highlighting the idea of prelinguistic direct awarenesses. Let me emphasize that although they are pre-linguistic, they are not pre-symbolic. They involve the occurrence and functioning of signs in that broad sense of 'sign' to which Peirce devoted so much loving (if often obscure) commentary.

130. Thus a prelinguistic awareness of something as red is a primitive member of the "this is red" family of representations. If we signal that an item has a function in a representational system akin to that performed in our language by a certain sentence, by enclosing the latter in dot-quotes to form a predicate of functional classification which applies to items which perform a function which is relevantly similar, we can represent the occurrence of a direct awareness of something red as red as follows:

Although t is a response to α as red in the causal sense, this is a small part of what is true of t, for the ability to have a representation of the ·this is red· kind essentially involves propensities which relate it to other elements of the representational system to which it belongs.

VII. Behaviorism and Meaning

131. Verbal behaviorists are sooner or later confronted by the problem of relating the concepts which they find it natural to employ in describing the involvement of linguistic expressions in behavior to the concepts and categories used by grammarians and logicians.

132. Concepts pertaining to syntax cause them little immediate trouble, although in the long run they hope to arrive at an understanding of why natural languages exhibit the syntactical structures and transformations they do.

133. Semantical concepts and categories, on the other hand, are notorious sources of discomfort. Attempts to provide behavioristic definitions of such terms as 'meaning,' 'sense,' 'reference,' and 'denotation' -- not to mention 'truth' -- have been startlingly inept.

134. Yet the behaviorist feels that there must be some close connection between these two families of concepts, each of which concerns essential features of language.

135. The mentalist, on the other hand, feels no qualm, about treating semantical terms as standing for ultimate properties of mental events, properties, indeed, the having of which makes them mental.

136. But this line of thought is not open to verbal behaviorists, for whom concepts pertaining to thoughts simply are, at bottom, concepts pertaining to linguistic behavior.

137. Their lack of success in defining semantical concepts in behavioristic terms appears as a threat to their very enterprise, for it would seem to mean that linguistic episodes have non-behavioristic attributes which are nevertheless essential to them.13

138. Fortunately a prima facie14 case can be made for the thesis that the verbal behaviorist has been confronted by a false dilemma. I shall sketch this case with respect to the semantical concept of meaning, as represented by the context

E (in L) means . . .
and standing for, as represented by the context
E (in L) stands for . . .

139. Ostensibly these contexts are relational, i.e. contexts of the form

According to the thesis I wish to advance nothing could be further from the truth.

140. If so, then the question 'Is the relation in question definable in behavioristic terms or is it a unique non-behavioral relation?' does not arise.

141. But before embarking on the task of defending the above thesis, it will be useful to set the stage by calling attention to a grammatical concept which is essential to a naturalistic theory of reference.

142. Consider the statement

Man is an animal
The subject term 'man' obviously falls in the general category of singular terms, as is shown by the fact that it is followed by a verb in the singular.

143. But if the statement has the surface grammar of a singular statement, it -- shall I say "obviously" -- has the depth grammar of a general statement, thus

Men are animals

144. The topic of what grammatical connections between the predicate 'man' and the predicates 'man' and 'animal' underwrite the transformation of "men are animals" into "man is an animal" belongs to the grammatical theory of classificatory systems.

145. 1 have called such terms as 'man' in the above statement15 distributive singular terms (DSTs), for the predications made of them distribute over the many objects which satisfy the predicates from which they are formed.

146. Parallel considerations apply to such statements as

'And' (in E) is a conjunction
just as 'man' in "man is an animal" is not a name, but a DST formed from the predicate 'man,' so " 'and' " in the above statement is not a name, but a DST formed from the predicate " 'and.' " Thus the above statement is equivalent to
'And's (in E) are conjunctions.

147. Other relevant examples of statements with DSTs as subjects are

The lion is tawny
The elephant has a long memory
These are equivalent respectively, to
Lions (typically) are tawny
Elephants (as a rule) have long memories

148. The parenthetical comments remind us that it would be a mistake to regard the original statements as paraphrases of the unqualified

All lions are tawny
All elephants have long memories.

149. With these elementary -- but invaluable -- points under our belts we can return to our central topic.

150. Consider

'Und' (in G) means and
It leaps to the eye that " 'und' " is functioning here not as the name of an abstract-linguistic object -- the class of which individual tokens of the German word are members -- but rather a DST formed from the sortal predicate " 'and ' "

151. But the most striking fact about the original statement is the unusual role which is being played by the English word 'and.' Thus, it is not in any straightforward way playing the role of a sentential connective.

152. A natural move is to construe the context in which 'and' occurs as a quoting one. One is tempted to rewrite the original statement as

'Und' (in G) means 'and' (in E)
and to paraphrase this, in turn, as
'Und' (in G) means the same as 'and' (in E)

153. But something has clearly been lost. For the original statement doesn't merely tell us that 'und' and 'and' have the same meaning, it is, in some way, designed to give this very meaning -- which

'nein' (in G) means the same as 'nyet' (in R)
clearly does not.

154. I pointed out in paragraph 139 above that statements of the kind we are considering are ostensibly relational statements. We are tempted to speak of the meaning relation and to search for its terms.

155. I shall present, however, without further ado, an alternative account which undercuts the centuries-old puzzles which this search has generated.

156. According to this alternative, the English word 'and' is, indeed, occuring in a quoting context. But the context is not

. . . means the same as . . .
but, simply
. . . . means . . .

157. Quoting contexts, however, are often such that to leave them unchanged while adding quotes to the item which is, in effect, being quoted, changes the sense. Thus it is incorrect to keep the term 'means' while adding quotes to the 'and.'

158. To pick the situation up by the right handle: If we add quotes to 'and' must we drop the word 'means' -- in favor of what?

159. The answer is both simple and powerful: In favor of the, copula. Mobilizing the already detected DST structure of the subject of the original statement, we get

'Unds' (in G) are . . .
What are they? Obviously they are not 'and's. They do not have the sign design of 'and's.

160. But to think that quoting has as its primary function to form expressions for sign designs is to take far too myopic a view of the power of quoting devices. Quoting does form sortal predicates from linguistic exemplars. But the criteria for satisfying the sortal need not be the design of the exemplar. It obviously can be the function of the exemplar. Indeed the criteria can abstract from the design.16

161. To make a long story short, our original statement has the deep structure

'Und's (in O) are ·and·s
where the criterion for being an ·and· is to be an item which functions in some language or other in a way which is relevantly similar to the way in which 'and's function in our language.

162. Our meaning statement gives the meaning of 'und' (in German) by presenting us with an exemplar and telling us that if we want to understand how 'und's function (in German) we should rehearse in imagination the cluster of functions characteristic of 'and.' See, for example, the transformation authorized by De Morgan's formulae.

163. To draw a distinction between propensities to exhibit patterned behavior which arc governed by rules, and those which are not, requires an account of the causal involvement of rule sentences in the generation of behavior which conforms to them. This topic, which is central to the theory of normative discourse, is a difficult one in its own right, and I can do no more on the present occasion than suggest that here, as elsewhere, the distinctions and principles of classical mentalistic psychology will be found to contain a large measure of truth which needs only demystification to be incorporated into behavioristics.

164. Just as to be a pawn is not to have a certain shape, but to be directly or indirectly involved with other pieces in a system of behavioral propensities specified by the rules of chess, so to be an ·and· is to be involved in a system of behavioral propensities conforming to the logical rules of the language.

165. And just as we can draw a conceptual distinction between propensities which conform to the constitutive rules of chess, propensities which conform to the tactical and strategic rules for winning at chess, and the idiosyncratic propensities of individual chess players, so we can distinguish in principle between propensities which conform to the constitutive rules of a 'language game,' propensities which conform to tactical and strategic rules for effectively playing the game, and propensities which reflect the happenstances of individual players.

166. Needless to say, the tidiness of our distinctions is a constructed tidiness, and not a datum plucked from the flux of experience as one might pluck a plum from a pie.

167. Nor does the ability to draw these distinctions in the abstract, carry with it a tidy decision procedure for determining, from sample behaviors, the rules of the game -- if any -- which certain people play.

168. The task of the anthropological grammarian is far more difficult than that of the anthropological student of games which bear a family resemblance to chess.

169. In just what ways it is more difficult, and in just what ways there are constraints on what the anthropological grammarian can -- even in principle -- hope to achieve, are topics of current interest on which one can say nothing without saying a great deal indeed.

170. But these issues do not touch directly on the viability of verbal behaviorism as a philosophy of mind.

171. There is, however, an indirect connection to which I shall devote my concluding remarks.

172. It will be remembered that 1 undertook to explore not only the context

E (in L) means . . .
but also
E (in L) stands for . . .

173. Consider, for example,

'dreieckig' (in G) stands for triangularity
Ostensibly this, too, is a relational context.

174. Notice that not every open sentence which has two 'openings' stands for a relation. Thus

... or ---
if ... , then ---
do not stand for relations -- unless one so stipulates, and by doing so, rides roughshod over important distinctions.

175. The essential feature of relations proper is that they relate individuals. Otherwise put, relation words are predicates and are properly flanked by individual constants.

176. If we look, with this in mind, at the statement we are considering, the idea that it is a relational statement seems to pass this crucial test.

177. Of course, if we take into account our previous discussion of DSTs, we may be prepared to regard our statement as a transform of

'drcicckig's (in G) stand for triangularity
But this, by itself, is quite compatible with the idea that standing for is a relation.

178. Now, when the question "Is stands for a relation?" is raised, just what is at stake? Let me remind you of the problem posed to the verbal behaviorist by semantical concepts. I wrote (in paragraph 137 above): "[The verbal behaviorists'] lack of success in defining semantical concepts in behavioristic terms appears as a threat to their very enterprise, for it would seem to mean that linguistic episodes have non-behavioristic attributes which are nevertheless essential to them."

179. This same problem arises in heightened form in connection with the ostensible relation of standing for. It can scarcely be denied that it is an essential feature of the semantic functioning of the German word 'dreieckig' that it stands for triangularity. Its involvement in German verbal behavior would necessarily be different if it stood, instead, for, say, temperance.

180. Yet no behavioristic definition of 'stands for' has gotten off the ground.

181. Once again, the behaviorist is ostensibly confronted by a dilemma: Either define in behavioristic terms a relation which can obtain between utterances involving 'dreiecking' and the property of being triangular, a relation which can assume the place of stands for in the family of semantic concepts; or grant that a relation essential to the functioning of 'dreieckig' in German verbal behavior can not be expressed in a behavioristic vocabulary.

182. The way out of this apparent dilemma is, interestingly enough, a more elaborate form of the previous strategy. The clue is provided by the fact that there is obviously a strong equivalence between the two statements

'dreieckig' (in G) means triangular
'dreieckig' (in G) stands for triangularity

183. One looks for a way of tracing them to the same deep structure.

184. Now if the previous argument is correct, the deep structure of the meaning statement is

'dreieckig's (in G) are ·triangular·s
How might we get from this to our stands for statement? Consider the following
Pubs arc poor man's clubs
If this sentence is viewed against the background of institutional sociology, do we not find a transformation into
The pub is the poor man's club
appropriate? Here we have a statement of which both the subject and the predicate involve distributive singular terms.

185. If we apply this type of transformation to

'dreieckig's (in G) are ·triangular·s
we would get
The 'dreieckig' is the German ·triangular·.

186. We would be home free if we could construe triangularity as having the depth grammar of a DST. Roughly,

Triangularity = the ·triangular·.

187. Now 1 have argued on a number of occasions17 that this is the truth of the matter, and that it makes possible a fundamental breakthrough in the ontology of attributes and, par passu, in the philosophy of mind.

188. The gist of the matter is that "... triangularity merely looks (to the eye bewitched by a certain picture) to be a name. It merely looks as though it referred to something non-linguistic. Applying to expressions in any language which do a certain job, its inter-linguistic reference is confused with a non-linguistic reference. Again 'stands for' merely seems to stand for a relation. It is, as 'means' proved to be, a specialized form of the copula."18

189. What is the point of there being two surface transformations of the deep structure

'dreieckig's (in G) are ·triangulars·s?
The answer is that whereas the surface grammar of 'means' statements is suited to the task of giving meanings, the surface grammar of 'stands for' statements is suited to connecting functional classifications of linguistic expressions with predications of truth. Thus, the singular term 'triangularity' fits both into the context
'dreieckig' (in G) stands for triangularity
triangularity is exemplified by (i.e. is true of) a19

190. Now the relevance of all this to behaviorism can be made vivid by considering a question which I posed to Professor Quine in the letter to which he refers in his remarks ("Sellars on Behaviorsim, Language and Meaning") on my contribution to the symposium. The relevant passage of my letter (of March 28, 1978) reads as follows:

I suspect that our major disagreement continues to lie in the area of ontology. I simply do not see how to fit platonic objects (classes and sets) into a naturalistic framework. Bluntly put: If sets are basic objects, how does the mind get in touch with them? (See my "Empiricism and Abstract Entities" in the Carnap volume.) I would use the same strategy with sets as I do with attributes and propositions. Statements about triangularity are to be parsed out as statements about any ·triangular·. We "get in touch with triangularity" by acquiring the ability to use ·triangular· tokens.

191. Quine counters (in "Sellars on Behaviorism, Language and Meaning") with

My reply is another question: "How does the mind get in touch with neutrinos?" What we have is a ponderous theory consisting of interlocking laws and concepts. It is in contact with sensory evidence only at its edge. The doctrines of elementary particles and spin in cosmic radiation occupy a remote part of this theoretical structure, and so do the doctrines of real numbers and sets. Epistemologically, sets differ from neutrinos only in being somewhat less analogous to observable bodies.

192. To this Duhemian ploy I have replied as follows:20

Thus, suppose a platonist with respect to attributes and/or classes to be asked: "Must there not be matter-of-factual relations between abstract entities and human minds by virtue of which abstract singular terms acquire a hook-up with the world?" Might not our platonist reply: "It is our theory as a whole, including its logical apparatus and such sortal predicates as 'molecule,' 'positron,' 'attribute,' 'proposition,' 'class,' 'class of classes,' etc., which confronts the tribunal of experience. Our language hooks up with positrons and classes alike by virtue of the application of the theory to experience. It is only the will-o'-the-wisp of the analytic-synthetic distinction which keep one from recognizing that 'class' and 'proposition' are in a continuum with 'current' and 'positron.' It is simply a matter of degrees of theoreticity, of remoteness from the occasion-sentences elicited by sensory stimulation."

The following consideration, however, should generate a measure of skepticism with respect to this facile gambit. The theory-whole has specific things to say about the causal relations which connect micro-physical objects with the sensory stimulations which bombard the sensory surfaces of experimenters looking at bubble chambers and photographic plates. The theory explains how we are in touch with micro-physical objects.

Thus, in addition to the Duhemian point that expressions for micro-physical particles acquire a hook-up with micro-physical particles by virtue of belonging to a theory which is applied as a whole, the theory offers a causal account of the specifics of the hook-up. This is not the case with such terms as 'number,' 'class,' 'attribute,' and 'proposition.' This fact introduces a radical discontinuity into Quine's Continuum, one which has important consequences for the problem of abstract entities, for ontology and, above all, for the philosophy of mind.

University of Pittsburgh
Pittsburgh, Pennsylvania


* A revised version of a paper presented in a Symposium on Behaviorism and Philosophy at the University of Missouri (Columbia). April 1978.

1 On perception, see, for example, "Some Reflections on Perceptual Consciousness." in Selected Studies in Phenomenology and Existential Philosophy, edited by R. Bruzina and B. Wilshire, The Hague, 1977. On action, see "Actions and Events," in Nous, 7, 1973 [reprinted in Essays in Philosophy and its History, Dordrecht, 1976].

2 That is, of any other episodes. For in its platonistic -- as contrasted with conceptualistic -- form, the classical theory posits a domain of independent entities (attributes, relations, propositions, possible worlds, etc.) which have syntactical and semantical properties.

3 Some philosophers (e.g. Gustav Bergmann) postulate unstructured awarenesses. According to them, all structure belongs to the other term of the relation. But this is not the orthodox form of the classical theory.

4 For example, see "The Structure of Knowledge," in the Machette Foundation Lectures for 1971 at the University of Texas. These lectures were printed in Action, Knowledge and Reality, edited by H. N. Castaneda, Bobbs-Merrill, 1975. The reference is to lecture II, "Minds."

5 Notice that as the phrase 'locutionary act' is used it tends to straddle the distinction between a linguistic actuality and a linguistic action.

6 I have added these two paragraphs in reply to Quine's comment on the vagueness of the concept of verbal propensities. (See the fifth paragraph of his remarks "Sellars on Behaviorism, Language and Meaning")

7 That the acquisition of this linkage of propensities is grounded in neurophysiological episodes is a part of the story which goes beyond my immediate purposes. For an exposition which takes it into account see my essay "The Structure of Knowledge." in Action, Knowledge and Reality, H. N. Castaneda, ed., Bobbs-Merrill, 1975. The reference is to pp. 316-31.

8 For an early account which stresses the difference between 'free' and 'tied' verbal behavior, see my "Language, Rules and Behavior," in John Dewey: Philosopher of Science and Freedom, edited by Sidney Hook, New York. 1949.

9 For an elaboration of this point, see my "Philosophy and the Scientific Image of Man," in Frontiers of Science and Philosophy, edited by Robert Colodny, University of Pittsburgh Press, 1962 [reprinted as Chapter 1 of Science, Perception and Reality, London, 1963]. See especially part four, "The Scientific Image."

10 In my early writings I succumbed far too often to this temptation.

11 I am clearly touching on themes central to perennial controversies over freedom and determinism. For a discussion of these issues in the spirit of the preceding paragraphs see my "Fatalism and Determinism," in Keith Lehrer, ed., Freedom and Determinism (New York, Random House. 1966): also Alan Donagan, "Determinism and Freedom: Sellars and the Reconciliationist Thesis," in Action, Knowledge and Reality (cited above, footnote 4), and my "Reply to Donagan," Philosophical Studies, Vol. 27. 1975.

12 For a development of this point, see "Some Reflections on Language Games," in Science, Perception and Reality, pp. 324 ff.

13 Not even the suggestion that semantical properties are 'epiphenomenal' or, perhaps, 'emergent' would give them aid and comfort, for the task of finding correlations between these equally essential semantical and behavioral properties would remain to haunt them.

14 I say 'a prima facie case' because the issues involved are so central to metaphysics that a conciliance of many lines of thought, many of which can not even be hinted at on the present occasion, would be necessary to bring conviction. For systematic attempts to develop these lines of thought, see Science and Metaphysics, London and New York, 1968, and Naturalism and Ontology, Reseda, Calif. 1979.

15 By contrast, in the statement 'man is a species,' the term 'man' is not a singular transform of the predicate 'man' but rather, as will become clear, of a predicate which refers to this predicate.

16 Though, of course, the criteria will require that the items which satisfy the sortal have a design which is capable of embodying the function.

17 Originally, in "Abstract Entities," Review of Metaphysics, 16, 1963 [reprinted in both Philosophical Perspectives, Reseda, Calif., 1979, and Essays in Philosophy and its History, Dordrecht, 1974].

18 "Meaning as Functional Classification," Synthese, 27, 1974, 417-437.

19 A systematic working out of this strategy will be found in Naturalism and Ontology. See particularly Chapter IV, paragraphs 77-84.

20 Naturalism and Ontology, Chapter I, paragraphs 32-34.