On Chomsky’s Review of Skinner’s
VERBAL BEHAVIOR

Kenneth MacCorquodale
University of Minnesota

Criticism 1: Verbal Behavior Is an Untested Hypothesis 1, Which Has, Therefore, No Claim upon Our Credibility

Skinner’s reasons for avoiding the word “hypothesis” in this connection can only be guessed. Psychologists readily confuse “hypothesis” with “hypothetical” in the sense of “fictional”, and it is a strong point in Skinner’s hypothesis that it contains no reference to fictional causal entities. All of the events, processes, and mechanisms invoked are themselves empirical, and therefore the hypothesis containing them is in principle fully testable and possibly disconfirmable. A more potent reason for his avoiding the word, however, is probably that “hypothesis” has somewhat curiously come to imply the possibility of experimental test, which Skinner has not performed and which he does not seem to consider feasible, although Verbal Behavior is rich in observational evidence. According to his hypothesis speech is the product of the convergence of many concurrent and interacting variables in the natural environment, which does not sustain the experimental separation and detection of the relevant component variables. Yet anything less than concurrence and interaction of many variables would not, according to the hypothesis, generate speech. Skinner’s situation resembles that of the astronomer “explaining” tides as the resultants of many interacting attractions. No one has ever experimentally tested that hypothesis directly either, yet it is highly plausible and Supported by much observational evidence which is probably the strongest conclusion we shall ever be able to make for it.

Chomsky avoids the word “hypothesis” in favor of more picturesque terms: “[Skinner] utilizes the experimental results [of laboratory studies of infra-human, non-verbal behavior] as evidence for the scientific character of his system of behavior, and analogic guesses (formulated in terms of a metaphoric extension of the technical vocabulary of the laboratory) as evidence for its scope. This creates the illusion of a rigorous scientific theory with a very broad scope, although in fact the terms used in the description of real-life and of laboratory behavior may be mere homonyms, with at most a vague similarity of meaning ... with a literal reading (where the terms of the descriptive system have something like the technical meanings given in Skinner’s definitions) the book covers almost no aspect of linguistic behavior, and ... with a metaphoric reading, it is no more scientific than the traditional approaches to this subject matter ... (Chomsky, 1959, pp. 30-31. Italics added).” Which is really only to say that the technical language of Skinner’s system is used in a hypothesis about verbal behavior; all scientific terms in untested hypotheses are necessarily “metaphoric extensions” and “analogic guesses”. What is puzzling, therefore, is the pejorative aspect which “metaphor” and “analogic” assume in the passage quoted.

Even more puzzling is the giddy speed with which the argument moves from its insight that the terms in the hypothesis are for now metaphoric and analogic, proceeds to the possibility that this may prove to be all they are, and concludes flatly with the verdict that the technical terms used do not describe verbal behavior. This goes too fast! That remains to be seen. Until the hypothesis is tested the literal (non-metaphoric, non-analogic) applicability of its explanatory terms remains in doubt, at worst. Chomsky’s only real argument for his conclusion that the terms of the theory do not in fact apply to verbal behavior is given in the quotation above. It depends upon the arnazing possibility that “real-life” and laboratory behavior may be different, as if somehow nature maintains two sets of natural laws, one for laboratories and the other for the rest of the world so that any law observed in the laboratory is prima facie suspect when applied to events outside. Entrancing though this idea is, it seems unparsimonious to suppose it. That really does not sound like nature.

The fact is simply that we do not yet know if verbal behavior is within the domain of Skinner’s system and whether the technical terms stimulus, response, reinforcement are literally applicable to verbal behavior and correctly parse it into its functional parts of speech.

Chomsky raises special considerations for doubting that each particular term of the basic theory applies to the verbal case. These will be briefly noted.

The stimulus:

Chomsky holds Skinner severely accountable for hypothesizing certain stirrmlus-response relations in Verbal Behavior, such as “a piece of music” as a stimulus for the response “Mozart”, or a certain painting for “Dutch”, and a red chair for “red” or “chair”. “Since properties are free for the asking, we can account for a wide class of responses in terms of Skinnerian functional analysis by identifying the 'controlling stimuli.' But the word 'stimulus' has lost all objectivity in this usage.” He then goes on to say: “Stimuli are no longer part of the outside physical world; they are driven back into the organism (Chomsky, 1959, p. 32).” This is a non sequitur. Stimuli are “free for the asking” only in hypotheses. Their quid pro quo is payable in empirical demonstrations of the evoking power of the putative stimuli. None of the purported stimuli listed above seems outrageously improbable for those responses, and not until such an empirical test of their evocative control has failed is anyone entitled to conclude that these are not stimuli for those responses. Chomsky’s conclusion that a putative stimulus has lost its objectivity because it occurs in a hypothesis is merely muddled. Skinner did not hypothesize a (hypothetical) stimulus. The stimulus is as real as ever. He hypothesized that there is a controlling relation between the real stimulus and the real response. As for his conclusion that the stimulus in a hypothesized stimulus-response relation has somehow been “driven back into the organism”, the rationale is harder to reconstruct. Reading Chomsky on the subject of the stimulus here and elsewhere in his review arouses a growing suspicion that he imagines that by naming one stimulus for a verbal response we name its only stimulus, and that one stimulus somehow preempts a response. He criticizes Skinner’s characterization of the responses “Eisenhower” and “Moscow” as proper names, controlled by the man or the city, because one frequently says “Eisenhower” and “Moscow” when the man and the city are not present (Chomsky, 1959, p. 32). Indeed one does, but this only shows, as Verbal Behavior repeatedly and clearly insists, that a verbal response may be controlled by different stimuli on different occasions. Verbal behavior does not obey any “one response-one stimulus” rule and it makes no sense to speak of the stimulus for anything. “Eisenhower” and “Moscow” are said for many reasons, among which are the presence of the man and the city. Perhaps Chomsky’s conclusion that Skinner’s stimuli for verbal responses have receded into the mind of the speaker is based upon this point: if I say "Eisenhower" when there is no Eisenhower then he must be in my mind. Is that the difficulty? Only if one is misguidedly determined to preserve Eisenhower as the only stimulus for Eisenhower. It is really impossible to be sure. However clear it is in its conclusions, the review is not much help on matters of rationale.

Reinforcement.

Inevitably Chomsky finds Skinner’s functional definition of a reinforcer unsatisfactory (that it increases the strength of any operant which precedes it), saying that it is “perfectly useless ... in the discussion of real-life [sic] behavior, unless we can somehow characterize the stimuli which are reinforcing ... (Chomsky, 1959, p. 36).” He is complaining because reinforcers can only be postdicted from the lfct of reinforcement, since they cannot be “characterized” in terms of any universal, independently knowable correlated property, such as drive-reducing power. Many psychologists share this dissatisfaction. But the fault, if any, is in nature, not in our theories. Reinforcers seem in fact to have only one universal property: they reinforce, and no amount of dissatisfaction will either add a correlated property nor disprove the fact that they do reinforce.

To be quite correct, whether a specific stimulus will be reinforcing for the behavior of any specific organism can be predicted without actually trying it. That is, reinforcers can be predicted, since all reinforcers are either species-characteristic (the unconditioned reinforcers) or they, have, in the history of the behaver, been paired with an unconditioned reinforcer (the conditioned reinforcers). Both of these classes are knowable before any behavioral test of their effect upon behavior is made (although it is technically infeasible to enumerate the members of the second class in the human case.) Furthermore, as Premack’s data have shown, all reinforcing stimuli are at least partially transituational; they will reinforce any operant whose initial probability is less than the consummatory or preconsummatory behavior which the reinforcing stimulus itself occasions. Therefore, a prediction of future reinforcing effect must be made given a fact of past reinforcing effect for any stimulus as well as information concerning the momentary probabilities of the operant to be reinforced and the behavior occasioned by the reinforcer. These considerations, in addition to providing bases for prediction as to which stimuli will reinforce which responses, also act as constraints upon the illicit invocation of ad hoc reinforcers. Together they remove the concept of reinforcement from “perfect uselessness”. Reinforcement is a real and powerful behavioral influence. Its inclusion in a theory of verbal behavior is decided on the basis of its own claim; it becomes a necessity whether it is “useful” in analyzing an instance of casual conversation or not.

Chomsky seems convinced that Skinner claims that “slow and careful” reinforcement applied with “meticulous care” is necessary for the acquisition and maintenance of verbal behavior (Chomsky, 1959, pp. 39, 42 [twice], 43). Chomsky does not cite Verbal Behavior in this context, and the fact is that Skinner does not say or imply that the reinforcement for verbal behavior must be carefully arranged or that differential reinforcement must be “careful”, applied with “meticulous care”, and “slow and careful” (Chomsky, 1959, p. 42). The idea is preposterous and the implication that Skinner said it is both careless and false.

Skinner does not, in fact, explicitly claim that any reinforcement is necessary for verbal behavior, although Chomsky supposes he does (Chornsky, 1959, pp. 36, 37, 38). His references are to statements in Verbal Behavior which say no such thing, and to Miller and Dollard (1941), who may. Skinner does claim that reinforcement is a potent influence upon verbal behavior, and, in fairness, he specifies no other strengthening operation for it. Nothing whatever is at stake in excluding from the hypothesis such alternative response-strengthening mechanisms as learning by imitation or by latent (non-reinforced) learning, if these should become demonstrable. The system would not then be destroyed or disproved; it would simply be supplemented by laws which specify the conditions under which these processes occur. Chomsky suggests that it is well-known that much language learning in children proceeds by imitation (Chomsky, 1959, p. 43). So, in fact, does Skinner (1957, pp. 5565) but he further specifies that the imitative repertoire (which he calls echoic in the verbal case) is itself a product of reinforcement. The evidence for an innate imitative tendency is very weak, so that the problem as Skinner saw it was to explain echoism when it does occur, and to account for the facts that the imitative tendency gradually restricts itself to the small segment of the vocal spectrum which the parent language uses, that its flexibility disappears with age, and that the echoic repertoire contains quite different dimensions in different speech communities (such as pitch in some, and not in others). These are all consistent with a reinforcement interpretation of the echoic's origins.

As for latent (unreinforced) learning, it is certainly incorrect to conclude that “Few investigators still doubt the existence of the phenomenon (Chomsky, 1959, p. 39).” The many studies which Chomsky cites in support of the existence of latent learning revealed mostly that the methodological problems involved in a crucial experiment on that question are overwhelming. The matter was not resolved. It was dropped.

Probability.

Chomsky criticizes Skinner’s “'extrapolation' of the notion [sic] of probability” as being, “in effect, nothing more than a decision to use the word 'probability' (Chomsky, 1959, p. 35).” This is the same objection that has been made to “stimulus” and “reinforcer”, i.e., the word occurs in a hypothesis, and therefore we need not reconstruct the argument on either side. Chomsky says, in addition, that “The term 'probability' has some rather obscure meaning for Skinner in this book (Choinsky, 1959, p. 34).” Small wonder, since he cites (Chomsky, 1959, pp. 29, 34) Hull's definition of probability (resistance to extinction) as Skinner’s basic indicator of probability or “strength” rather than Skillner's, which is simply the likelihood of occurrence of a response, measured as a rate where possible, but as a relative frequency in any case. Skinner thus defines probability quite as any other natural scientist does. Much more ominously for Skinner’s purposes, Chomsky seems not to grasp the difference between the overall probability of occurrence of an item in a speaker's verbal repertoire, which is the frequency with which it occurs in his speech over time without regard to his momentary circumstances, and the momentary probability of a given response in some specified set of circumstances. (See, for example, Chomsky, 1959, p. 34.) The two probabilities are very different. The overall probability that any speaker will say, for example, “mulct”, is very low; it occurs rarely in comparison with such responses as “the” or “of”. The probability that he will say “mulct” may become momentarily extremely high, as when he sees the printed word. Of the two, overall probability is a typically linguistic concern, while momentary probability shifts are, in a sense, the very heart of the psychologists' problem, since they reflect the relation between speech and its controlling variables. Under what conditions does an organism speak an item from his repertoire? Simply knowing the repertoire tells us precisely nothing about that. If Chomsky really did not, in fact, see this difference it is impossible to imagine what the rest of Verbal Behavior could have meant to him, and no wonder that he regarded it with such astonishment and dismay.

Verbal behavior's momentary probabilities are difficult to assess in practice because the most sensitive experimental indicator in nonverbal research, rate, is not useful: strong verbal responses are not normally repeated several times. Skinner mentions some production effects which on occasion may reflect the strength of a non-repeated, single utterance, such as loudness, speed of production, or repetition if it does occur. Skinner says quickly and explicitly of these, however, that they are untrustworthy: “It is easy to overestimate the significance of these indicators (Skinner, 1959, p. 25; additional warnings are given on pp. 27 and 141).” It is somewhat shocking, therefore, that in spite of Skinner’s disclaimers Chonisky imputes to his hypothesis an entailment that a strong response must be “shrieked (Chomsky, 1959, p. 35)” or shouted “frequently and in a high-pitched voice (Chomsky, 1959, p. 52).”

So much for the emigration of the system out of the laboratory. Chomsky faults the argument because it did so.

Originally published in the Journal of the Experimental Analysis of Behavior, 1970, pp. 83-99. Reproduced here by permission of the Society for the Experimental Analysis of Behavior.

Return Home