It would indeed be nice if we could draw direct inferences about the nature of cognitive operations by analyzing the syntactic structure of utterances. But at the moment this remains a hope since no reliable methods of investigation have as yet been developed to establish the nature of this relationship. There are some troublesome facts which would seem to argue against the likelihood of a direct relationship between cognitive operation and syntactic operation. In the O'Donnell et.al. study 39 specific structures and functions were identified for attention and 36 of these were observed in every age sample; the three completely missing in kindergarten speech were not such used by older children either. Furthermore, the observed changes in frequency of occurrence of some specific patterns were not monotonic: some decreased in frequency with age to return at a later age with greater frequency. While these facts do not eliminate the possibility of a cognitive-syntactic relation, they do have two important implications. One is that we should be studying individual children longitudinally rather than crosssectional samples. The other is that the type of structure per se is not likely to be the crucial variable. What this should be is not easy to specify at the moment but I would guess that we need to pay attention to the more abstract operations involved in effecting syntactic transformations: flexibility (e.g. number of alternative transformations that are paraphrases of each other), nature of the generative rule under control (e.g. deletion vs. substitution vs. expansion), complexity of the syntactic operation (e.g. nature of selection rules and information content of the syntactic unit), and so on.

The second source of troublesome facts for the direct cognitive-syntactic relation hypothesis lies in the findings on the early syntactic development of young children. Long before a child enters kindergarten he has already mastered the major syntactic devices of his native language. He has aquired control over extremely complex rule governed operations in the speech area while remaining extremely immature - by adult standards - in other modes of cognitive operations. In the words of Simon (1962; quoted in Bruner, 1967), the child has not mastered the 'architecture of complexity for things', but only for words. The paradox between these two orders of complexity has led some writers (e.g. McNeill, 1966) to speculate on the innate component of a hypothetical "language acquisition device" which is said to equip the child with innate capacities for syntactic operations. The utterances of a four year old are sometimes so striking because they say such nonsensical things: their syntactic maturity puts us off guard and we forget for a moment how conceptually immature they actually are. Sapir (1921; quoted in Bruner, 1967) uses a nice analogy to comment on this decalage: "It is somewhat as though a dynamo capable of generating enough power to run an elevator were operated almost exclusively to feed an electric doorbell." The syntactic and conceptual levels of operation seem to run an independent course of development, until the experlences of the child in manipulating the environment moves on from the "electric doorbell" stage to something more like an electric kitty cart stage. At that point, as Bruner suggests, the independence of the two systems recedes and there obtains a state of "mismatch": "A child can say of two quantities that one is greater than another, a moment later that it is less than the other, and then that they are the same - using his words as labels for segments of experience. It is not until he inspects his 1anguage that he goes back to his experience to check on a mismatch between what he sees with his eyes and what he has just said. He must, in short, treat the utterance as a sentence and recognize contradiction at that level. He can then go back and reorder experience, literally see the world differently by virtue of symbolic processes reordering the nature of experience" (Bruner, 1967, p. 442).

What kind of a conceptual schema is implied by the syntactic achievements of the very young child? It has been established (see the review by McNeill, 1966) that even the earliest speech of children exhibit two of the fundamental properties of grammar: categoriality and hierarchical organization. Thus the distinction between "pivot" sad "open" class in the two-word constructions of children beginning to talk is based on a distinction of categories of syntactic function (e.g. the child's grammar allows for utterances such as allgone sticky, more hot, and pretty boot but not their reversal or "inappropriate" combinations such as more pretty). Neated conatructions whereby several words replace a single word while retaining the same property as a single constituent provides further examples of categoriality (e.g. as in the example given by Weir, 1962, and cited by McNeill, 1966: go throw go for them, go to the top etc.). Progressive differentiation of the "pivot" class exhibits the principle of categoriality within a hierarchically organized structure (e.g. P1, the original pivot class breaks up into articles, demonstratives and everything else or P2; P2 in turn breaks up into adjectives, possessives, and everything else or P3, and so on until the adult syntactic classes are fully approximated).

Grammatical categories have the property of discreteness. Thus, at the phonological level, phonemes (but not necessarily phones) are discrete conceptual entities; there are no intervening steps between one phoneme and another based on some degree of phonetic or perceptual similarity. At the syntactic level, words do not change from say noun to verb on some gradual basis: syntactic units are discreet. This property of discreteness is quite-different from the characteristics of perception in the sensory field: perceptual similarity is a gradual thing. At the semantic level too, discreteness seems to be the principle of operation: semantic markers are not gradual; a word is marked either for animateness or inanimateness, for a male or female feature, etc., and there are no words that blend gradually into one or the other.16

The conceptual schemata which make syntactic operations possible consist therefore of hierarchical organization, categoriality, differentiation, substitution, and discreteness. But these by themselves are insufficient to explicate the basic conceptual function of a sentence. One might describe this function as a process of predication. Linguists have argued (e.g. Greenberg, 1963) that there are at least three universal properties of a base-structure grammar: subject-verb relations, verb-object relations, and modification. Bruner (1967) argues that these three kinds of syntactic relations are describable in terms of the three logical operations of function, causation, and intersection, respectively. These are powerful intellectual tools with which the child is born and which equip him to master language and develop conceptual representations by which he can organize his experiential world.

It is safe to say, I think, that the utilization of these intellectual operations presuppose some kind of sensory input, although we know next to nothing about specific details of the role of sensory input. Children with drastically impaired sensory receptors and motor effectors are yet able to develop fully operational intellectual capacities, often without even a slow down in rate (see Lenneberg, 1967). I think it is necessary to assume that even the simplest utterances have an underlying and temporally antecedent conceptual event associated with them. I would deny however that conceptual events necessarily have an underlying and historically antecedent sensory or motor event associated with them. This is not to deny the possibility that some utterances have such prior external relation. And it is even possible that the early speech of children is typically of this sort. Children's early speech is characterized by a simple syntactic structure of the universal kind (subject-predicate, subject-verb-object) and by an absence of transformations (McNeill, 1966). The relative simplicity of this structure fits the conception we have of the immaturity of their intellectual operations which are characterized by the enactive and iconic levels of representation (Bruner, 1967), these being closely related to motoric and sensory events. Much of the speech function of such young children is thus ostensive, that is to say, externally related. There is a need here for extensive investigations on the very young child's perceptual organization and its relation to his ostensive speech (viz, function-predication, causation-verb-object relation, intersection-modification). This area of research is critical since, once transformations set in, the close relation between conceptual event and speech vanishes. As Lashley (1951, quoted in Bruner, 1967) put it: "The facility with which different word orders may be utilized to express the same thought [in mature speech] thus in further evidence that the temporal integration is not inherent in the preliminary, organization of the idea." In other words, the left-to-right temporal string in surface structure is not a basis for inferring the psychological organization of the conceptual event to which the utterance is related.

These considerations lead me to formulate a two-factor theory of the syntactic-conceptual interaction process. One factor in this interaction relates to the influence of syntactic operations upon conceptual organization. This results from the problem solving activity that is initiated by "mismatch troubles" between the syntactic and experiential schemas (as in the example given by Bruner in the above quotation). The other factor relates to the influence of conceptual events upon the syntactic structure of the utterance as it occurs during encoding. Here, it is the dynamo itself that is being fed. These two interaction processes appear to me to be fundamentally different. In the first, cognitive structure, driven by the complexities of syntactic organization, changes in fundamental ways in its capacity to handle symbolic operations. As the process unfolds, the child becomes capable of handling symbolic operations of increasing orders of complexity. Here (in the very young child) is where I would expect that syntactic operations offer a clue to cognitive operations, and I think it is significant that longitudinal studies at this stage offer a picture of acquential enlargement of syntactic repertoires (O'Donnell, et.al., 1967) unlike subsequent stages where no new syntactic patterns emerge (as pointed out in the discussion above). Once the mismatch stage is corrected, and the speaker's conceptual and syntactic schemata are brought into line, syntactic pattern per se offers little clues to the conceptual operation that underlines it, and we must look for other sources of information (perhaps the implicit and implicative inferences we are able to make on the basis of utterances in context).

Although I used the term "stage" for referring to loci within this two-factor account I do not think that these processes follow necessarily a developmental pattern. Adults are continually confronted with mismatch troubles. Habitual modes of conceptual organization may be based directly on semantic implications that are no longer applicable to changing aspects of the world, When these semantic implications are of a syntactic origin (as is the case for example with the tense system in Navaho, see Whorf, l956) the mismatch problem is usually discussed under the rubric of the Sapir-Whorf Hypothesis. When the semantic implications have a culturally determined affective origin, the mismatch problem is often discuased under the rubric of "general semantics" (see Korzybski, 1958 or Weinberg, 1959) or, in the more traditional psychological literature, under rubrics such as "set" or "labelling" (see for example the classic experiments by Carmichael, Hogan, and Walter, 1932 and others reviewed in Osgood, 1953), "cognitive dissonance" and related investigations (see Brown, 1965), and in an important sense, the whole field of psychotherapy.

I have used the term "encoding" throughout this paper to refer to the relation between a conceptual event and the utterance which represents it. I should perhaps clarify my conception of the nature of this representation process. It is not an encoding process in the strict sense of information theory whereby the translation process is fully determined by a set of rules previously stated and agreed to. There is no one-to-one relation between elements of the conceptual process and elements of the code (in the sense of translating English into Morse). This is why I described the meaning of an utterance as a derivation process based on inferential reasoning rather than a decoding process based on translation rules. The nature of the representational relation between an utterance and a conceptual event must be something very different from the representational process discussed by cognitive psychologists in the Behaviorist school (e.g. Osgood, 1953; Berlyne, 1965). These writers view meaning in general as a representational mediation process whereby some reduced and internal version of the total behavior of an individual in the situation comes to "represent" that situation by virtue of the aimilarity relation between thia laternal proceca and the original total behavior. To account for the meaning of events with no prior history of direct involvement of the organism (e.g. unicorns, far away places and times one reads about) these authors appeal to secondary conditioning device they refer to as "assign learning." But their central representational thesis remains land locked within the behavioral aspects of the organism to which "assigns" are tied. Meaning can never be anything but behavior and representation retains its "passive sense" 'something stands in the place of something else' (Furth, 1968) and this 'something' must be defined as a piece of behavior.

Such a conception of representation limits the meaning of an utterance to a signification in terms of pieces of behavior and behavioral dispositions. This is precisely what Skinner attempted to do in his functional analysis - without any success. It is a position with a double handicap: a limitation in practice due to the impossibility of specifying the reduction process from the total behavior to internal event and the impossibility of verifying the claim that these internal events follow the laws of stimulus and response that hold at the overt level - a limitation acknowledged by the proponents of the position, and a limitation in principle due to the weakness of behavioral concepts in accounting for mental events - a limitation disputed by its proponents.

I would not want to deny the possibility that this type of representation and internalization dose in fact occur. What Bruner has called the "enactlve" and "iconic" levels of representation may indeed be of this form. Knowledge at that level may be interiorized representations of motoric and sensory dispositions. This is the sense in which animals and computers have knowledge. But when we move up toward symbolic representation we switch to a qualitatively different order of complexity, to borrow Chomsky's phrase again, a complexity of the mental world which is not matched by motor and sensory functions. There is no need in this argument to raise the black flag of metaphysics which empiricists would take as their exit cue. This is simply an issue of the needed recognition of the complexity of mental operations, of the "ideational world," that the mind is more complex than the sum of its sensory and motor cortex.

At the end of Verbal Behavior Skinner presents "two personal epilogues" upon which I would like to comment. The first deals with Skinner 's evaluation of his attempt at a "scientific" description of verbal behavior. He confesses that what he had hoped to achieve by this book was "to get the reader to behave verbally as I behave" (1957, p. 455). What could he have meant by this statement? Since he considers "thinking" as a form of verbal behavior, I take it that he would maintain his statement even in the hypothetical situation where the reader would never have an opportunity to talk to others about the subject of verbal behavior (say because he is on an island all by himself) or where the reader is mute and knows no sign language. His statement could then be rephrased to say that his purpose was "to get the reader to think [about verbal behavior] as I do". But in what sense did he mean "as I do"? He did not merely mean (as he indeed states it explicitly) to teach the reader the use of new and specialized terms such as "mands", "autoclitics" and "intraverbals" for then he should be happy with this paper - which I am aure is not the case. He meant that the reader should use these terms in the manner he uses them, in fact, that the reader come to utter sentences that resemble his sentences. The ideal case would be for the reader to learn the book by heart and then to recite portions of it on proper occasions, somewhat like a priest quoting from the Blble. But even short of this ideal case, he might, I am sure, accept sentences that resemble those in the book. But how are we to specify this resemblance, and how would we know whether the statements are equivalent or paraphrases of each other and not contradictions (say by effecting negative transformations)? I submit that whatever answer there is to this question, or to the generaI problem of meaning, this case being a simple instance of it, must fall back on the intuitive assertions that speakers of the language may make - but this is precisely what Skinner left out of the account.

The second personal question he takes up in the Epilogue goes back to what motivated him to undertake this work. In his own account: "In 1934, while dining at the Harvard Society of Fellows, I found myself seated next to Professor Alfred North Whitehead. We dropped into a discussion of behaviorism, which was then still very much an 'ism,' and of which I was a zealous devotee. Here was an opportunity which I could not overlook to strike a blow for the cause, and I began to set forth the principal arguments of behaviorism with enthusiasm. Professor Whitehead was equally in earnest - not in defending his own position, but in trying to understand what I was saying and (I suppose) to discover how I could possibly bring myself to say it. Eventually we took the following stand. He agreed that science might be successful in accounting for human behavior provided one mate an exception of verbal behavior. Here, he insisted, something else must be at work. He brought the discussion to a close with a friendly challenge: 'Let me see you,' he said, 'account for my behavior as I sit here saying, No black scorpion is falling upon this table.' The next morning I drew up the outline of the present study" (1957, pp. 456457).

What degree of success has Skinner achieved? To evaluate this question I would like to make a distinction (as I have done elsewhere, see Jakobovits, 1966) between what Skinner does in verbal Behavior and what he says he is doing. By the criteria that he sets for himself and by his claim of what he has done, Verbal Behavior is a monumental failure. Take for instance his answer to Whitehead's challenge. Although he admits that he cannot account in retrospect for Whitehead's words since they were "emitted under a set of circumstances now long since largely forgotten", he argues hypothetically along the following lines: "A few relevant facts about the conditions under which Profesaor Whitehead made his remark are available. So far as I know there was no black scorpion falling on the table. The response was emitted to make a point - taken, as it were, out of the blue. This was, in fact, the point of the example: why did Professor Whitehead not say 'autumn leaf' or 'snowflake' rather than 'black scorpion'? The response was meant to be a poser just because it was not obviously controlled by a present stimulus. But this is, of course, the kind of material the Freudians relish, for it is under just such circumstances that other variables get their chance. The form of the response may have been weakly determined, but it was not necessarily free. Perhaps there was a stimulus which evoked the response black scorpion falling upon this table which in turn led to the autoclitic No. The stimulus may not have been much, but in a determined system it must have been something. Just as the physicist may suggest various explanations of the drop in temperature [in a problem cited earlier] in order to show that it could be explained in lawful terms, so it is not entirely beside the point to make a guess here. I suggest, then, that black scorpion was a metaphorical response to the topic under discussion. The black scorpion was behaviorism" (1957, p. 458).

Skinner's analysis of the problem seems reasonable enough to me but is entirely in contradiction with his stated purpose throughout the book, namely to eliminate from an account of verbal behavior any inferences about mentalistic concepts. Note the thesis of his argument above: "The response was emitted to make a point . . . The response was meant to be a poser . . ." (italics mine). It is clear that Skinner has most definitely not gotten rid of the notion of intention in his account of verbal behavior.

If we then disregard Skinner's statements about what he says he is doing and examine only what he has in fact done, then I submit that Verbal Behavior is to be considered a measured success. It provides the beginnings of a promising performance model of language in terms of a functional analysis of speech utterances, namely the three-way relation between the intention of the speaker, the utterance within the communicative context, and the meaning which the listener can derive on the basis of an inferential process. It is this kind of a model that I have tried to elaborate here.

(continue here)