Printer Friendly

Has Chomsky's argument been refuted? A reply to Skinner, Cautilli, and Hantula.

In a recent paper in this Journal, Skinner, Cautillli, & Hantula (2003) analyze various aspects of Nonstandard English, and they arrive at a number of interesting conclusions. However, although there is much to say about these conclusions, I will not deal with them here. Rather, I want to concentrate on the section that deals with historical accounts of language studies and, in particular, on their criticism of Chomsky's views regarding verbal behavior. In their paper, Skinner et al. (2003) repeat the usual behavior-analytic arguments against Chomsky. However, just as most behavior-analysts, they do not realize that these arguments fail to respond to the essential component of Chomsky's claim. In addition, as long as behavior analysts have not refuted this component, it is very unlikely that a dialogue can be established between them and psycholinguists. I will therefore deal again with this issue, although I have already done so in the past (see, e.g., Stemmer, 1987a, 1987b, 1990, 1994)

Keywords: verbal behavior, Skinner Chomsky debate, psycholinguistics, functional analysis of language.


What is Chomsky's argument? Let me give the version that appears in his Aspects of Syntax (1965). This argument, which he has repeated in several places, is that behaviorists (Chomsky speaks of empiricists, but this is irrelevant here) cannot account for the speaker's ability to produce and understand new sentences that "are not similar to those previously heard in any physically defined sense ... nor obtainable from them by any sort of "generalization" known to psychology and philosophy" (1965, p. 58).(Similar statements can be found in, e.g., 1959, pp. 55-57; 1968, p. 30, pp. 51-53, p. 79; 1975, pp. 152-153, pp. 172-174). Nevertheless, children do produce and understand new sentences. Since Chomsky thinks that these behaviors are not based on generalizations from samples, he therefore speculates that humans are born with some kind of innate capacity to form verbal categories such as Noun or Subject. Moreover, a person's verbal behavior is supposed to derive from generalizations that are based on these categories.

Behavior analysts (e.g., Mabry, 1993; Palmer, 1986), as well as philosophers (e.g., Goodman, 1967; Putnam, 1967) have usually dealt with the speculative component of Chomsky's argument, with his Innateness Hypothesis. However, most of them have ignored the essential component of his argument, namely, that a child's production of new sentences requires generalizations that are not known to psychology or philosophy. This also holds for Skinner et al. (2003). They deal with the Innateness Hypothesis and they think that it assumes that "the relationship between words and the world is intrinsic, fixed and determined" (p. 14). They then cite Palmer (1986) who shows that the relationship between words and the world does not have this character:
 ... within a language there is no relationship between the sound of
 an utterance and its grammatical structure. Clearly, there is no
 physical property of the stimulus that suffices to identify its
 part in speech. Nothing about the word 'house' enables us to
 conclude that it is a noun, or that it might be a 'subject' (pp.

Palmer indeed proves that the relationship between words and the world is not intrinsic, fixed and determined. First, it is not clear whether Chomsky's hypothesis indeed requires a fixed relationship between words and the world. But most importantly, Palmer does not deal here with the essential part of Chomsky's argument, the part that motivates Chomsky's speculative hypothesis. On the contrary, Palmer's analysis seems to support the claim that the verbal behavior of children indeed requires generalizations that are not known to psychology. Consider for example plural endings. Children are able to produce the plural endings of a very great number of verbal stimuli, such as 'house-houses', 'song-songs', 'idea-ideas', etc. But there are no children who have been exposed to concrete examples of each and all of the singular-plural pairs that they are able to produce. Rather, they generalize from a particular set of examples. Some of them even overgeneralize, e.g. 'foot-foots', which is a pair that is clearly not included in the sample to which most English speaking children have been exposed.

The generalization that enables children to produce new singular-plural pairs is of course not based on physical similarity. As Palmer observes, "there is no physical property of the stimulus that suffices to identify its part in speech"--that is, there is no physical property that is shared by the singular-plural pairs that children can produce. Still, children do produce plural endings for words that are not in the sample. Consequently, unless we can show that such generalizations are of "the sort known to psychology and philosophy", Palmer's observation that the generalization is not controlled by a physical property seems to support Chomsky's claim rather than rebutting it.

We therefore have to examine the nature of the generalization process that accounts for a child's production of new plural endings. Before dealing with this process, let me notice two issues. First, it does not matter whether the set of examples to which children are exposed is small or rather extremely large. As long as there is at least one instance of generalization, we must show that it is of the sort "known to psychology", if we don't want to accept Chomsky's conclusion that children are born with some kind of innate category, a category that contains all and only the expressions for which they perform the generalization.

Second, Chomsky's argument is valid even if the children's "generalization classes" do not match the adult categories--if, for example, the class that contains the singular-plural pairs that a particular child has produced does not match the adult class Noun. The essential component of Chomsky's argument does not require such a matching. As long as there is a generalization, it has to be accounted for.


So which is the process that stands behind the children's generalizations? Clearly, it is not stimulus generalization, since stimulus generalization requires a physical property that should control the corresponding generalization class. (2) Moreover, it is not sufficient merely to give a name to the generalization process, e.g., to say that it is an abstract, or structural, or functional generalization. From a scientific point of view, name giving has very little value. Detailed information has to be provided of the generalization process in order to admit the process as an explanation of the relevant verbal behavior.

Methodological considerations suggest that in the present case, the detailed information should contain at least the following three elements. First, it should contain a detailed description of the nature of the events that transform a non-physical class--i.e., a class that is not determined by a physical feature--into a generalization class for certain children. Second, reasons must be given that support the assumption that the children who do generalize according to such non-physical classes have indeed been exposed to the relevant events. Third, efforts should be made to try to find out whether this type of generalization process can be detected in animals or whether, as Chomsky suggests, it only holds for humans.

Most behavior analysts think that the non-physical generalizations that occur in verbal behavior derive from stimulus equivalence (on stimulus equivalence, see, e.g., Dixon & Spradlin, 1976; Hayes, 1991; Sidman, 1997, 2000; Sidman & Tailby, 1982; Sidman, Wynne, Maguire, & Barnes, 1989). But highly specialized events are required for stimulus equivalence to occur, and it is unlikely that the children who engage in the non-physical generalizations have been exposed to such specialized events. Consider for example the generalizations that stand behind the production of words such as 'toy' (or 'clothes', 'tool', etc.). There is no physical similarity between different toys. Yet, already in very early stages, children are able to generalize from a limited sample to further objects, since they apply the word 'toy' to new objects, provided they have been exposed to events which given them the "knowledge" that one can play with the new objects. It is unlikely however that these young children have been exposed to the complex events that generate stimulus equivalence. One cannot blame psycholinguists for ignoring stimulus equivalence as the source of the non-physical generalizations that occur in the early stages of verbal behavior. (For a devastating criticism of the stimulus-equivalence program, see, Tonneau, 2001.)

There exists an alternative explanation, which is much more plausible. It attributes the generalization to the exposure to certain pairing events (see, e.g., Quine, 1974, p. 20; Stemmer, 1973, pp. 50-56, 1980, 1989), and it is supported by experiments, including with non-humans. To be sure, the explanation requires the exposure to the pairing of the word 'toy' with a limited number of toys--the sample--and to the pairing of the further objects with playful activities or with the observation of such activities. No one disputes that the children who do apply 'toy' to new objects have indeed been exposed to such events. Admittedly, except for a few instances (see, e.g., the comments on Stemmer [1989], by Harris, 1989; Morris, 1989), most psycholinguists have also ignored this explanation. At least this explanation assumes the role of learning processes that have been demonstrated experimentally, and it assumes events that one can reasonably assume to have occurred prior to the occurrence of the non-physical generalization.

Still, the above process is not yet sufficient for generating generalization classes containing "nouns"--more exactly, classes containing the verbal stimuli that, among other features, show the singular-plural endings. In order to account for this generalization, we must attribute to the children not only the above capacity to learn to generalize according to nonphysical classes but also the capacity to pay attention to many subtle features that are shared by the verbal stimuli, such as the possibility of being preceded by "articles" like 'the' (see, e.g., Stemmer, 1973, pp. 69-71; 1987a, 1987b). Additionally, these features plus the above capacity then confer a generalization efficacy to the relevant class. (Psycholinguists often speak in this connection of a distributional analysis, see, e.g., Maratsos, 1982.) In this way, then, we can explain generalizations such as plural endings without having to assume innate syntactic categories. Again, it is clear that stimulus equivalence cannot explain such sophisticated generalizations.

Let me summarize so far: Palmer's conclusions do not refute the essential component of Chomsky's argument. Rather, they point to a research program that should give a behavior-analytic explanation of the generalization efficacy of classes that show "noun-like" features, and that are the basis of generalizations such as plural endings.


But the most serious challenge is still to come, and it is illustrated by another argument against Chomsky that is raised by Skinner et al. (2003). They cite Moerk's (1983) research, who "discovered that children experience every major sentence type about 100.000 times / month" (p. 15). And these experiences are supposed to explain the children's grammatical performance. But the readers have by now surely realized that Moerk's discovery does not counter Chomsky's argument. The critical expression in Moerk's conclusion is 'sentence type' for it acknowledges that children have the capacity to classify sentences into types (or categories), such as passive forms, questions, or negative sentences--more exactly, it acknowledges that children can generalize from, say, a particular set of active-passive examples to further instances. Clearly, these generalizations are not instances of stimulus generalization, since they are not physically determined. There is no physical similarity between the active-passive pairs which children are able to produce, and the pairs that constitute the sample to which they have been exposed. Consequently, by acknowledging the behavioral efficacy of sentence types, Moerk's discovery actually supports Chomsky's claim, unless the three points mentioned above have been taken care of, the points that confer explanatory power to the appeal to generalizations that are not physically determined.

The present author has also dealt with this issue, and his conclusions refute the essential component of Chomsky's argument with respect to these types of generalizations as well (see, e.g., Stemmer, 1987a, 1987b, 1990, 1994, 2000). This is important because when Chomsky speaks about generalizations that are not obtainable "by any sort of generalization known to psychology and philosophy", he mainly intends the generalizations that agree with sophisticated syntactic categories such as active-passive, etc. Let me give a very brief account of an active-passive transformation which shows the complexity of the generalizations it requires, and which suggests that the simplistic explanations that have been given by behavior analysts have little scientific value. (The example is adapted from Chomsky, 1975, pp. 31-32.)

Consider the statement,

(i) The man who holds the painting holds the book.

The corresponding passive statement is,

(ii) The book is held by the man who holds the painting.

Clearly, as Chomsky has stressed, the passive statement (ii) does not derive from the physical features of the active statement (i), because the physically identical occurrences of 'holds' in (i) are treated differently in (ii). One of them becomes 'is held'. So what is the process that stands behind the transformation of (i) into (ii)? (Before continuing, the readers can try their hand at this task.)

In Stemmer (1987a, 1987b, 1990), it has been shown that the process that stands behind the transformation of (i) into (ii) can be explained in a behavior-analytic framework (see also Stemmer, 1973, pp. 69-73, 82-86). However, the explanation is far from self-evident. It must at least:

(1) appeal to generalizations based on non-physical properties;

(2) specify the events that give generalizing efficacy to the non-physical properties that stand behind these generalizations;

(3) give reasons that support the assumption that the persons who can produce the passive form (ii) have indeed been exposed to the relevant events;

(4) Specify the events by which certain stimuli come to control the relational expressions x holds y and x who y;

(5) show that relational expressions give origin to structures, i.e., structured combinations of verbal expressions;

(6) distinguish between the main and the secondary relational expression in (i), a distinction that is determined by the structures established by the relational expressions.

Once these points have been established, we can then show that a generalization, that is based on some of the non-physical properties of the structural components of (i), accounts for the transformation of (i) into (ii). In particular, the generalization treats differently the first and the second occurrence of 'holds', because the structure of (i), which is determined by the relational expressions x holds y and x who y, gives them a different status. Moreover, it establishes the clause 'who holds the painting' as an independent unit that is transposed unchanged in the passive transformation.

In a well known criticism of Chomsky, MacCorquodale (1970) attributes the learning of grammatical behavior, such as the one that enables us to transform (i) into (ii), to the child's ability "to make complex abstractions and to generalize from them to diverse new instances" (1970, p. 93). It is very likely, however, that the main reason why MacCorquodale's paper had only a minimal impact on psycholinguists is that it attributes grammatical behavior to "abstract generalizations", without specifying the nature of the events that give origin to the generalizations, the nature of these generalization processes, the properties that control the generalizations, and the empirical evidence showing the behavioral efficacy of the generalization classes. The explanation is therefore an instance of "name giving", and is subject to Chomsky's criticism that "talk of generalization in this case is entirely pointless" (1959, p. 56; see also Chomsky, 1965, p. 57-58; Chomsky, 1968, p. 30, p. 38; Chomsky, 1975, p. 142, p. 152, p. 202, and other places).

This again suggests that in the absence of a detailed account of the acquisition of grammatical behavior, Moerk's discovery does not refute the essential component of Chomsky's argument. On the contrary; since it acknowledges the child's capacity to classify sentences in types (or categories) that are not determined by physical properties, it must specify how these types acquire their behavioral efficacy, and how they then explain the acquisition of particular verbal skills. However, no such specification and explanation has been given by Moerk.


Skinner et al. (2003) also mention various authors who argue, "that no evidence exists to support the claim that children operate with adult-like categories or rules in formulating early sentences and thus these categories should not be posited" (p. 14). As I explained above, this observation does not refute the essential component of Chomsky's argument because this component does not depend on the matching of the categories. What is critical is that when acquiring verbal behavior, children indeed operate with categories; yet, these categories are not physically determined. In Palmer's words, "there is no physical property of the stimulus that suffices to identify its part in speech." Consequently, behavior analysts have to account for the children's use of categories, whether adult-like or otherwise.


A detailed and empirically supported refutation of the essential component of Chomsky's argument is a prerequisite for a serious dialogue between behavior analysts and other psychologists. The present note may contribute to make clear to behavior analysts that vague explanations of verbal behavior, and in particular of grammatical behavior, will not do the job.


Chomsky, N. (1957). Syntactic structures. The Hague: Mouton.

Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.

Chomsky, N. (1968). Language and mind. New York: Harcourt, Brace & World

Chomsky, N. (1975). Reflections on language. New York: Pantheon Books.

Dixon, M., & Spradlin, J. (1976). Establishing stimulus equivalences among retarded adolescents. Journal of Experimental Child Psychology 21, 144-164.

Goodman, N. (1967). The epistemological argument. Synthese, 17, 23-28.

Harris, M. (1989). Comprehension and production in early lexical development: A comment on Stemmer. Mind & Language, 4, 229-234

Hayes, S.H. (1991). A relational control theory of stimulus equivalence. In L.J. Hayes & P.N. Chase (Eds.) Dialogues on verbal behavior (pp. 19-40). Reno, NV: Context Press.

Mabry, J.H. (1993). Comments on Skinner's grammar. The Analysis of Verbal Behavior, 11, 77-88.

MacCorquodale, K. (1970). On Chomsky's review of Skinner's Verbal behavior. Journal of the Experimental Analysis of Behavior, 13, 83-99.

Maratsos, M. (1982). The child's construction of grammatical categories. In L. R. Gleitman & E. Wanner (Eds.), Language acquisition: The state of the art (pp. 240-266). New York: Cambridge University Press.

Moerk, E.L. (1983) A behavioral analysis of controversial topics in first language acquisition. Journal of Psycholinguistic Research 12, 129-155.

Morris, M. (1989). Empirical versus epistemological considerations. Mind & Language, 4, 222-228.

Palmer, D.C. (1986). Chomsky's nativism: A critical review. In P.N. Chase & L.J. Parrott (Eds.) Psychological aspects of language. (pp. 44-60). Springfield, IL: Thomas

Putnam, H. (1967). The 'innateness hypothesis' and explanatory models in linguistics. Synthese, 17, 12-22.

Quine, W.V. (1974). The roots of reference. La Salle, IL: Open Court.

Sidman, M. (1997). Equivalence relations. Journal of the Experimental Analysis of Behavior, 68, 258-266.

Sidman, M. (2000). Equivalence relations and the reinforcement contingency. Journal of the Experimental Analysis of Behavior, 74, 127-146.

Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching to sample: An expansion of the testing paradigm. Journal of the Experimental Analysis of Behavior, 37, 5-22.

Sidman, M., Wynne, C.K., Maguire, R.W., & Barnes, T. (1989). Functional classes and equivalence relations. Journal of the Experimental Analysis of Behavior, 52, 261-274.

Skinner, L.M., Cautillli, J.D. & Hantula. D.A. (2003). A verbal analysis of nonstandard English. The Behavior Analyst Today, 3, 12-23.

Stemmer, N. (1973). An empiricist theory of language acquisition. The Hague: Mouton.

Stemmer, N. (1980). Natural concepts and generalization classes. The Behavior Analyst, 3, 41-48.

Stemmer, N. (1987a). The learning of syntax: An empiricist approach. First Language, 7, 97-120.

Stemmer, N. (1987b). The learning of syntax: A reply. First Language 7, 137-144.

Stemmer, N. (1989). Empiricist versus prototype theories of language acquisition. Mind and Language, 4, 201-221.

Stemmer, N. (1990). Skinner's Verbal behavior, Chomsky's review, and mentalism. Journal of the Experimental Analysis of Behavior, 54, 307-319.

Stemmer, N. (1992). The behavior of the listener, generic extension, and the communicative adequacy of verbal behavior. The Analysis of Verbal Behavior, 10, 69-80.

Stemmer, N. (1994). On structure-dependent grammars: A reply to Mabry. The Analysis of Verbal Behavior, 12, 97-99.

Stemmer, N. (2000). The role of action names, action frames, and modifiers in listener behavior. The Behavior Analyst Today, 1, 21-26.

Tonneau, F. (2001), Equivalence relations: A critical analysis. European Journal of Behavior Analysis, 2, 1-33.

Nathan Stemmer

Beth David Institute, Jerusalem (1)


(1) Address for correspondence: Nathan Stemmer, 9 Diskin Street, Jerusalem, Israel. Email address:

(2) Behavior analysts generally use the term 'stimulus class'. But I prefer the term 'generalization class' because it stresses the fact that at least some elements of the class are not strictly identical to the sample. Moreover, there are generalization classes that are not determined by physical features and it seems somewhat strange to speak of a stimulus class that is not controlled by a physical property. For an exact definition of the term 'generalization class', see Stemmer (1980, 1992).
COPYRIGHT 2003 Behavior Analyst Online
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Stemmer, Nathan
Publication:The Behavior Analyst Today
Date:Sep 22, 2003
Previous Article:FAP-enhanced group therapy for chronic pain.
Next Article:The emperor's new anthropomorphism.

Related Articles
B.F. Skinner: A Reappraisal.
Yesterday, today, and tomorrow: looking back at a year of growth and forward to our new century.
Fifty years later: comments on the further development of a science of Verbal Behavior.
Dodo-Bird or dressing up the pig and the politics of psychotherapy: a reply to Vandenberghe and de Sousa (2005).
The long good-bye: why B.F. Skinner's Verbal Behavior is alive and well on the 50th anniversary of its publication.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters