Chomsky's Nativism: A Critical Review

The Analysis of Verbal Behavior 2000, 17, 39-50

Chomsky's Nativism: A Critical ReviewDavid C. PalmerSmith College

Is grammar innate? Noam Chomskyholds that it is, or, more accurately, thatthe hypothesis that it is innate is theonly coherent and plausible one thathas yet been proposed to account forthe acquisition of language. Extrapo-lating to broader issues, he has cham-pioned a retreat from behaviorism andempiricism to cognitivism and ratio-nalism, from approaches that seek todetermine the relationship between anorganism's behavior and the environ-ment to those that wish to discover theorganism's "essential nature," ofwhich behavior is an incomplete ex-pression. His arguments, which are de-tailed, polemical, and persuasive, areevidently inspiring to a thriving schoolof linguists and to many laymen withan interest in language and philosophy.It is important to assess his positioncarefully, not only because he con-cludes that little is to be gained by pur-suing the analysis of verbal behaviorwith the assumptions and methodologyof radical behaviorism but because heclaims to have achieved considerablesuccess with very different assump-tions and methodology. Success in ex-plaining complex behavior deservesour attention whatever the approach.When we examine Chomsky's posi-

tion, however, we find that, not onlyare his objections to other approachesweak, but the success of which hespeaks has been achieved by renderingother problems more difficult, if not

Reprinted, with permission, from P N. Chase& L. J. Parrott (Eds.). (1986). Psychological as-pects of language (pp. 44-60). Springfield, IL:Charles C Thomas.

I thank John Donahoe for discussions thatcontributed to the present paper.

Correspondence conceming this paper shouldbe sent to the author at Department of Psychol-ogy, Smith College, Northampton, Massachusetts01063 (E-mail: [email protected]).

completely insoluble. This paper out-lines Chomsky's position, emphasizinghis argument that the brain of the new-born infant must be organized to ex-tract rules of grammar from samples ofspeech. This position is criticized ontwo grounds. First, it places too heavya burden on evolutionary principles.Second, the putative innate mecha-nisms must respond to stimuli, to ac-tual physical events, but it appears tobe impossible in Chomsky's system tocharacterize these events. Finally,Chomsky's sophistical argumentsagainst behavioral accounts of lan-guage are rejected on grounds that hehas confused properties of his formalsystem with properties of human be-ings. The notion that language consistsof an infinite number of sentences mustbe abandoned when we move from therarefied atmosphere of formal analysisto the world of stimulus and responseclasses.

CHOMSKY'S ASSUMPTIONS

Chomsky shares a number of fun-damental assumptions with behavior-ists and other experimental psycholo-gists. He believes that organisms are ajoint product of their genetic endow-ment and individual experience andthat the experimental approach of thenatural sciences is appropriate for thestudy of language. He is tentativelymonistic; while he freely uses mental-istic terminology, he does so for thesake of convenience, believing theseterms to be abstractions of physicalstructures or processes in the body,presumably the brain (see Chomsky,1980b, for a recent review of his po-sition). His goals, however, are differ-ent from those of behaviorists. He isnot particularly interested in verbal be-havior itself, influenced as it is by the

39

40 DAVID C. PALMER

idiosyncratic history and circumstanc-es of the speaker; rather, he is interest-ed in the "essential nature" of humanbeings that enables us to acquire a lan-guage. Specifically, Chomsky wishesto discover those elements of our ner-vous systems implicated in languagethat are genetically coded, hence "uni-versal." Chomsky calls these elements"universal grammar," a name that sug-gests his view of the task accomplishedby these innate mechanisms: providinga set of rules to be used in speech pro-duction and comprehension.Chomsky is not dogmatic about the

nature of universal grammar, so de-fined. He concedes that it may proveto be some general-purpose reinforce-ment mechanism, but this strikes himas implausible. Just as cells in the vi-sual cortex are organized in specialways not characteristic of cells con-trolling, say, digestion, those elementsof the nervous system responsible forour ability to acquire language shouldnot be expected to be organized in thesame way as those implicated in learn-ing to ride a bicycle. The language sys-tem, the visual system, and no doubtother systems, he asserts, may be mod-ular.Chomsky's interest in species differ-

ences is not what makes his views con-troversial. Species differences, ofcourse, are of as much interest to be-haviorists as to linguists. A child learnsto speak, and the family dog does not.No one doubts that this is due to ge-netic differences, and any illuminationof these differences will contribute toour understanding of the behavior oforganisms. Direct evidence of the ge-netic contribution to human behavior ishard to acquire, however. Programs ofgenetic engineering, surgical intrusionsof the nervous system, and well-con-trolled behavioral experiments are, forethical reasons, seldom possible. Wemust usually make do with speculationand extrapolation from indirect evi-dence. It is Chomsky's particular spec-ulations and his rejection of alternativespeculations that are controversial.

A TERMINOLOGICAL QUIBBLE

To begin with, we may object toChomsky's terminology, particularlyhis use of the term "universal gram-mar" to refer to unspecified innateproperties of the nervous system. Awriter is free to define his terms as hechooses, but as Winograd (1977) haspointed out, the reader who has agreedto Chomsky's definition finds, in sub-sequent discussion, that he has agreedto some kind of innate grammar in thetraditional sense of the term, i.e., a setof rules. This lends a spurious cogencyto Chomsky's argument. Additionally,Chomsky freely uses mentalistic termssuch as intention, belief, purpose, will,and mind without defining them. Oc-casionally he indicates that he is mere-ly talking about properties of the ner-vous system, but he does not tell uswhere in the nervous system we willfind the mind with its intentions, be-liefs, and will. As a consequence, hisdiscussions remain abstract and meta-phorical, apparently awaiting the daythat someone can operationalize histerms without endangering the formalsystem that has been erected on thisterminological quicksand.Even within the formal system itself,

Chomsky's terms are not always clear-ly defined. A language, we are told, is"a set (finite or infinite) of sentences,each finite in length and constructedout of a finite set of elements" (Chom-sky, 1957, p. 13). Subsequently, welearn that human languages are infinitesets of sentences. But what is a sen-tence? Chomsky uses two definitionsof sentence, a formal, precise one andan informal one, and he does not con-sistently use either one. The informaldefinition is simply that which nativespeakers agree to be a sentence whenthey are not encumbered by "irrele-vant" problems of memory, motiva-tion, time, or patience. This is a rela-tively small set owing, apparently, tothe ubiquity of these encumbrances.By the formal definition, a sentence isany string of symbols characterized bythe grammatical rules devised by the

CHOMSKY'S NATIVISM 41

linguist. At the very least, these gram-matical rules must be consistent withthe set of sentences defined by consen-sus. Thus, the formal definition is de-rived from data provided by using theinformal definition, and both defini-tions depend ultimately upon the gram-matical intuitions of native speakers.However, Chomsky does not provideus with criteria for deciding when ajudgment of grammaticality is to beconsidered valid, an important omis-sion considering that these judgmentsare notoriously variable.

ARGUMENTS FOR AGENETICALLY DETERMINED

GRAMMAR

One could argue that these termi-nological issues are irrelevant to eval-uate the substance of Chomsky's posi-tion, particularly his argument thatthere is an innate language module. Solet us turn to this critical argument. Asnoted above, his unit of analysis is thesentence, and his data are his judg-ments, and presumably the judgmentsof others, that particular sentences areor are not "well-formed." (He is notconcerned with prescriptive rules ofgrammar, such as proper use of the ob-jective case, but with regularities inlanguage that are respected by nativespeakers without formal training.)Thus, (1) is a well-formed sentence,while (2) is not:

(1) Is the man who is hungry here?(2) Is the man who hungry is here?

Similarly, (3), (4), and (5) are well-formed, while (6) is not, though itsmeaning is reasonably clear:

(3) Each of the men likes the others.(4) The men like each other.(5) Each of the men expects John to

like the others.(6) The men expect John to like each

other.What is it, Chomsky asks, that pre-vents people from uttering sentencessuch as (6)? Surely no one has beentaught such things in grammar school.

It can hardly be maintained that childrenlearning English receive specific instruction

about these matters, or even that they areprovided with relevant experience that in-forms them that they should not make theobvious inductive generalization, say, that"each other" takes some plural antecedentthat precedes it. Children make many errorsin language learning, but they do not as-sume, until corrected, that "The candidateswanted me to vote for each other" is awell-formed sentence meaning that eachcandidate wanted me to vote for the other.Relevant evidence is never presented formost speakers of English, just as no peda-gogic or traditional grammar, howevercompendious, would point out these facts.Somehow this is information that childrenthemselves bring to the process of languageacquisition as part of their mode of cogni-tion. (Chomsky, 1980b, pp. 43-44)

A similar problem is raised by "ques-tion formation":

We select some noun phrase in a sentence,replace it by an appropriate question word,place the latter at the beginning of the sen-tence, and with other mechanical opera-tions, form a question. Thus on the modelof the sentence, "John saw a man," we canform "Whom did John see?" Or, to take amore complex case, on the model of thesentence, "The teacher thought that his as-sistant had told the class to study the les-son," we can question "the class" and ask:"Which class did the teacher think that hisassistant had told to study the lesson?" Butconsider the following example of roughlycomparable complexity: "The lesson washarder than the teacher had told the class itwould be." Here, if we question "theclass," we derive: "Which class was thelesson harder than the teacher had told thatit would be?" Evidently this is not a well-formed question, though its intended senseis clear enough and perfectly intelligible,with a little thought. It is difficult to imag-ine that people capable of these judgmentshave all had relevant training or experienceto block the obvious inductive generaliza-tion. Rather it seems that some specificproperty of the human language faculty-hence a general property of language-leads to these consequences, a property thatderives from our modes of cognition.(Chomsky, 1980b, p. 42)

When we analyze the structure oflanguage at a certain level of abstrac-tion, according to Chomsky, we dis-cover that there are general principlesof grammar that are violated in sen-tences such as (6), for example, theprinciple that a reciprocal expressionsuch as "each other" may not refer to

42 DAVID C. PALMER

an antecedent outside of the clause inwhich "each other" occurs unless thelatter happens to be the subject of aninfinitive (Chomsky, 1980b, p. 174), asin (7). Note that (8) is ungrammaticalaccording to this principle:

(7) The candidates expect each otherto win.

(8) The candidates expect each otherwill win.Because children quickly learn to re-spect such distinctions with little, ifany, formal instructions, and becauseno one has proposed a satisfactory ex-planation of these facts in terms of atheory of learning, Chomsky assumesthat fundamental elements of the gram-mar of human languages must be ex-pressed somehow in the genetic code.He suggests that universal grammar,triggered by relatively brief exposureto a particular language, is able to ex-tract or construct a grammar for thatlanguage. Universal grammar presum-ably contains those fundamental prin-ciples that are common to all humanlanguages and constrains the particulargrammars that can be acquired.Grammar is seen as fundamental;

"language" is an epiphenomenon in-fluenced by motivational variables,memory, nonlinguistic concept learn-ing, and other things. The task of thelinguist is to characterize abstractlygrammars of various languages as theywould be spoken by ideal speakers ina homogeneous verbal community inan attempt to discover principles ofgrammar of wide generality, if not uni-versality. Work of this sort has been inprogress for several decades and, ac-cording to Chomsky, has met with con-siderable success. He concedes thatconclusions are tentative and will un-doubtedly be refined or replaced, astate of affairs to be expected in anyempirical inquiry. Nonetheless, lin-guists are becoming increasingly ableto characterize universal grammar andhence to offer hypotheses about the ca-pacities that newborn infants bring intothe world as a product of their geneticendowments.

In the current form of Chomsky's

theory, a given sentence is presumedto be "represented in the mind" at sev-eral levels. It begins as a declarativesentence, with expressed (as opposedto "understood") subjects, verbs, di-rect objects, and so on. Elements of thesentence may then be deleted or rear-ranged subject to various constraintssuch as those governing reciprocal ex-pressions like "each other." At thislevel the sentence includes "traces" ofdeleted and rearranged elements intheir original positions as well as therearranged elements in their new po-sitions. Finally, the surface structure,the sentence as it appears at the behav-ioral level, is generated by representingphonetically all elements except traceelements. This scheme accounts for in-tuitions about relatedness among sen-tences, ambiguities, and many fine dis-tinctions about what is grammaticaland what is not.

This picture is incomplete, ofcourse, but Chomsky asserts that itstands in sharp contrast to alternativetheories which do not even attempt toexplain the kinds of grammatical judg-ments people are capable of, judg-ments which have served as grist forthe theories of linguists. Chomskywrites:

The critic's task is to show some funda-mental flaw in principle or defect in exe-cution or to provide a different and pref-erable account of how it is that what speak-ers do is in accordance with certain rulesor is described by these rules, an accountthat does not attribute to them mental rep-resentation of a system of rules (ruleswhich in fact appear to be beyond the levelof consciousness). If someone can offersuch an account of how it is that we knowwhat we do know, e.g. about reciprocals, orjudge as we do judge, etc., there will besomething to discuss. Since no such ac-count has been forthcoming, even in themost primitive or rudimentary form, therereally is nothing to discuss. (1980b, p. 130)

CHOMSKY'S CHALLENGE

One need not be disconcerted by thischallenge. Chomsky is charging hiscritics to provide an alternative expla-nation for hypothetical behavior, the


behavior of judging particular sentenc-es grammatical or ungrammatical. Heevidently believes that everyone willrespond in the same way and for thesame reason, that is, that there is anindependent variable called "grammat-icality" that controls the behavior ofanyone asked to judge utterances.Since judgments of grammaticality arein fact highly variable, he insists thatwe consider only the behavior of anideal speaker in a homogeneous verbalcommunity.

Analogous idealizations have beenadopted in other sciences. When manyvariables interact, it is common prac-tice to consider each in isolation.Hence the physicist assumes pointmasses, frictionless surfaces, and per-fectly elastic collisions. However, theseidealizations are useful only if the var-iables being omitted are unimportantfor an understanding of the phenome-non under study. If one's verbal behav-ior and judgments about utterances arein fact a function of the individual'sparticular experience within a specificcommunity, then considering the intu-itions of an ideal speaker in a hypo-thetical community will tell us nothing.Faced with disorderly data, Chomskyremoves to a hypothetical world whereorder emerges. It is not surprising thatno one has proposed an alternative ac-count, for this is a world of Chomsky'sown making. Order has not been dem-onstrated; it has been assumed.

However, even if we satisfactorilydemonstrate that instances like thoseprovided by Chomsky are not universalthe task remains of explaining why cer-tain novel expressions "sound right"to someone while others do not. Froma behavioral point of view the task isformidable, requiring that we know agreat deal about the individual's rein-forcement history-more than it isusually possible to know. We wouldhave to determine appropriate units ofbehavior and the individual's historywith respect to these units. We mightfind, for example, that "each other,"though two words, is a single operantor that the frame "X ... Y ... each

other" is a single operant, where X andY have certain prosodic, temporal,functional, and perhaps formal featuresbut are otherwise free to vary. (Wewould resist the temptation to call X"a plural noun phrase" and Y "averb," for that merely raises the ques-tion of what the formal, functional,prosodic, and temporal characteristicsare of plural noun phrases and verbs.)We might find that "each other" is acomponent of half a dozen larger unitsor that it is under intraverbal control ofa number of stimuli. In the latter casea number of different operants wouldbe formally identical.Once relevant units of behavior and

their controlling variables were identi-fied it would be possible to speculatewhether a particular utterance wouldsound natural or strange to an individ-ual. Since the value of such a predic-tion by no means justifies the effort togather the relevant data, it is unlikelythat anyone will answer Chomsky'schallenge. (We can, of course, inventcontingencies to explain any given ex-ample, but he would regard this asempty.) Nonetheless, there are alter-natives to Chomsky's account that de-pend, not on internal representations orunderlying competence, but on the in-dividual's long history with relevantverbal operants. It is hard to see howit could be otherwise, for we have nointuitions about strings of grammaticalsymbols by themselves or about sen-tence tree diagrams. Despite Chom-sky's suggestion that intuitions about afull range of uses of the term "eachother" follow from learning that it is areciprocal expression and not the nameof a tree, surely we have no intuitionsabout strings of nonsense syllablesdrawn from bins labeled "noun,""verb," "reciprocal expression," andso on. Knowing grammatical catego-ries is no help in judging utterances"by ear."As for a "flaw in principle" that

Chomsky exhorts his critics to cite,there appears to be none. There are noobjections in principle to the notion,however vague, that the nervous sys-

44 DAVID C. PALMER

tem is innately designed to extract agrammar from a sample of speech, butwe can question the extent to whichthis serves as a parsimonious explana-tion of verbal behavior. Rather thanshedding light on the problem, it ren-ders it more mysterious.Chomsky begins by characterizing

grammar in formal terms. Havingachieved some success at this he thensimply imputes the formal apparatus tothe speaker as an innate mechanism.This is a tidy solution to a complexproblem, and it might even be true, butnote that, as it stands, it is a homun-culus theory, and as such, it is unsat-isfactory until the homunculus has inturn been explained. The genetic en-dowment is a convenient source of ho-munculi for every behavioral phenom-enon we don't understand. If we asknothing further of such an explanationthen it is a universal solution. There areno limits to our invoking it. Two ques-tions immediately arise to whichChomsky has provided no satisfactoryanswer. First, how did universal gram-mar get selected, and second, how doesuniversal grammar get "triggered" bya verbal environment?

UNIVERSAL GRAMMARAND EVOLUTION

Certain grammatical conventionsserve a communicative function andmight be learned because they do so.Other principles are arbitrary and seemto have little point, such as that con-cerning reciprocal expressions. Of thisprinciple, Chomsky (1975, p. 175)writes, "It is often a difficult problemeven to discover examples that bear onthe hypothesis in question." It is prin-ciples such as this that Chomsky ar-gues must be innate: The examples arefew and the grammatical rules are ar-bitrary and unnecessary. But this is anargument that cuts both ways; the sameevidence that he adduces to support hisposition can be used in a parallel ar-gument against it. If a grammaticalprinciple is an arbitrary restrictionwithout practical consequences in the

ontogenic environment and hence can-not be accounted for in terms of com-municative contingencies (Chomsky,1980a, p. 41), then it clearly cannotconfer a selective advantage to an or-ganism endowed with it. Chomsky ac-knowledges this problem but points outthat a child has only a few years toconstruct a grammar while the princi-ple of natural selection has had manythousands of years (Chomsky, 1980c,p. 44). This will not suffice, however.If the rules are arbitrary it doesn'tmake any difference how long selec-tive forces have been at work. Naturalselection is simply not an appropriatemechanism to explain universal gram-mar.

This conclusion does not troubleChomsky. He writes:

It is, in fact, perfectly possible that the in-nate structure of mind is determined byprinciples of organization, by physical con-ditions, even by physical laws that are nowquite unknown, and that such notions as"random mutation" and "natural selec-tion" are as much a cover for ignorance asthe somewhat analogous notions of "trialand error," and "conditioning," "rein-forcement," and "association." (Chomsky,1969, p. 262)

Again, Chomsky might be right thatthere are additional principles in-volved, but this hardly offers supportfor his position. Rather, it adds a fur-ther burden of proof. In addition to ex-plaining the origins of grammar, hemust now formulate and explain theworkings of new evolutionary princi-ples or "physical laws now quite un-known."

Perhaps not wanting to depend onunknown principles, Chomsky has sug-gested another solution:

[Universal grammar] may well have arisenas a concomitant of structural properties ofthe brain that developed for other reasons.Suppose that there was selection for biggerbrains, more cortical surface, hemisphericspecialization for analytic processing, ormany other structural properties that can beimagined. The brain that evolved mighthave all sorts of special properties that arenot individually selected; there would be nomiracle in this, but only the normal work-ings of evolution. We have no idea, at pres-


ent, how physical laws apply when 1010neurons are placed in an object the size ofa basketball, under the conditions that aroseduring human evolution. (Chomsky, 1980a,p. 321)

It is certainly true that not every-thing coded by the genes must be adap-tive. Hair color, eye color, and bloodtype are all genetically determined andare not obviously adaptive, but neitherare they universal. When a trait is notactively selected for, we expect vari-ability, not universality. Moreover, tothe extent that arbitrary structures re-quire energy and resources, we wouldexpect them to be selected against.The chances that a universal gram-

mar was an accidental by-product ofother properties of the nervous system,an unexpected bonus when the humannervous system exploded in size, seemremote indeed, given how abstract andcomplex the putative innate rules are.Moreover, in the absence of sugges-tions about the structure of whichgrammar is a by-product, and what therelationship between them is, Chom-sky's answer is no answer at all. Wemight just as plausibly assert that lan-guage is an accidental by-product ofother behavior acquired in the first fewyears of life.Chomsky (1980a, 1980b) repeatedly

asserts that the problem of explainingthe genetic basis of universal grammaris no different from the problem of ex-plaining the origin of any physical or-gan, say, the liver. No one ever assertsthat we learn to have arms rather thanwings, or that we learn to have a heart,he argues. No doubt there are manythings that we don't know about theorigin and development of the physicalorgans, but to be confident that the ge-netic endowment exerts considerablecontrol, surely it is sufficient to notethat such structures are adaptive andthat they are, in fact, physical struc-tures.A hypothesis about behavior need

not specify physiological correlates orevolutionary origins to be useful, but acomplex structure with no adaptivesignificance is anomalous. In contrast,

functional analyses of verbal behavior(e.g., Skinner, 1957) require few, ifany, principles in addition to those al-ready known to apply to nonverbal be-havior; moreover, these principles areclearly adaptive (Skinner, 1966), andapply with appropriate qualificationdown the evolutionary ladder. Humanshave the necessary vocal musculatureand are particularly sensitive to sec-ondary reinforcement, social contin-gencies, and, apparently, private stim-uli generated by other behavior. Theseare quantitative differences from otherorganisms that are both adaptive andeasily accommodated by evolutionaryprinciples. These differences alonemay be sufficient to account for verbalbehavior in humans.' Chomsky, in at-tributing grammatical competence tothe newborn infant, has not solved theproblems of language acquisition; hehas simply transferred them to theshoulders of the evolutionary biologist,where they remain as intractable as be-fore.

THE STIMULUS CONTROLOF INNATE BEHAVIOR

An additional difficulty facingChomsky's position is perhaps morefundamental. Let us assume that he iscorrect, that humans are innatelyequipped with a neurological modulethat extracts an acceptable grammarfrom a small and degenerate sample ofspeech, triggered perhaps by criticalexperiences and with parameters set by

' The development of vocal musculature sen-sitive to reinforcement contingencies may be es-pecially significant. First, it is a response systemthat is free from the demands of locomotion, ori-entation, and the manipulation of objects. Pri-mates usually have plenty to do with their handsother than sign with them. Second, and perhapsmore importantly, when we speak, we stimulateourselves exactly as we stimulate others, and wedo so essentially instantaneously. This immedi-ate and faithful stimulation, which is not char-acteristic of, say, sign language, is no doubt im-portant both in maintaining somewhat uniformcontingencies throughout the community and infacilitating the acquisition of verbal operants.Under some conditions, reinforcement will be"automatic."

46 DAVID C. PALMER

developments in "other cognitive do-mains." Subsequent performance is ahodgepodge of behavior, an epiphe-nomenon, determined in part by thegrammatical module and in part bymany other factors. The problem thatnow arises is the relationship between"degenerate speech" and the devicethat extracts a grammar from thisspeech. The putative device is an in-put-output device. In go samples ofspeech, and out comes the grammar, orperhaps a set of candidate grammars,most of which will be winnowed outlater. Setting aside the improbabilitythat such a device is an accidental by-product of, say, increased cortical sur-face, we must determine the functionalrelationship between this input andoutput. This can be considered a kindof problem in stimulus control, sinceeach verbal stimulus controls a partic-ular response of the device, as in a re-flex. However, unlike the reflex, the re-lationship between stimuli and thegrammar is arbitrary. Languages varyfrom culture to culture, and within alanguage there is no relationship be-tween the sound of an utterance and itsgrammatical structure. Clearly there isno physical property of the stimulusthat suffices to identify its part ofspeech. Nothing about the word houseenables us to conclude that it is a noun,or that it might be a "subject."The input to this device, then, must

be the product of a grammatical anal-ysis rather than raw stimuli. At thevery least, words must be classifiedinto their parts of speech. But parts ofspeech have formal definitions; they donot have operational ones. Nouns areoften uttered in the presence of"things" and verbs in the presence ofactivity, but many nouns are not"things," and many verbs are not per-ceptible actions. Perhaps when a childutters a particular word in the presenceof a particular class of objects (or stateof affairs) and is reliably reinforced fordoing so, that acoustical signal is rep-resented and tagged with an "N." Thusevery grammatical distinction might betraced to a particular set of reinforce-

ment contingencies. This is unsatisfac-tory, since we still do not end up witha class of symbols that coincides withthe concept noun. We do, however, endup with a repertoire of behavior thatcoincides precisely with Skinner's con-cept of tact. Once grammatical distinc-tions are traced to contingencies of re-inforcement, the innate grammar is nolonger doing any work. On the otherhand, if they cannot be traced to rein-forcement contingencies, then the child(or the innate mechanism) has no wayof generating a grammar.

Chomsky's allusions to imprintingand fixed-action patterns as examplesof complex innate behavior (1959, pp.41-43) suggest that he fails to appre-ciate that these behaviors do not occurspontaneously but are elicited by spe-cific stimuli. Herring gull chicks do not"peck at their mothers' bills to getfood"; they peck at red spots. Duck-lings do not "follow their mothers";they are reinforced by proximity to ob-jects similar to the particular objectthat was bustling around when theyhatched. If we wish to say that a par-ticular behavior is genetically deter-mined or "wired in," it must be pos-sible to specify the environmentalevents that elicit, release, or trigger it.Not only has Chomsky failed to do thisfor his hypothetical grammar-generat-ing device, he apparently thinks it can-not be done:

Although one might propose various oper-ational tests for acceptability, it is unlikelythat a necessary and sufficient operationalcriterion might be invented for the muchmore abstract and far more important no-tion of grammaticalness. (Chomsky, 1965,p. 11)Furthermore, there is no reason to expectthat reliable operational criteria for thedeeper and more important theoretical no-tions of linguistics (such as grammatical-ness and paraphrase) will ever be forth-coming. (Chomsky, 1965, p. 19)

If there are no stimuli, objective crite-ria, or even a set of operations bywhich we (or our innate language ac-quisition devices) can identify suchtheoretical entities as grammatical sen-tence, subject, noun phrase, and so on,


then it is a mystery how we can reflex-ively generate rules characterizing per-missible relationships among these en-tities. Chomsky has been able to for-mulate precisely his theoretical ideasbecause they have remained abstract,but useful theories cannot remain ab-stract forever. If there is no way to usethem to predict, control, or describe ac-tual events, then they are empty.

THE UNIT OF ANALYSISAND THE NOTION OF

INFINITY

The choice of a unit of analysis inbehavior is critical. The orderly rela-tionship between behavior and its con-trolling variables deteriorates if weconsider units that are too broad, toolong, or too narrowly specified (Skin-ner, 1935). If one defines one's units apriori rather than empirically, it is pos-sible that behavior will appear to beinfinitely variable and to bear little re-lationship to environmental events.Chomsky commits this error by choos-ing the sentence as a unit of analysis.He does not defend this choice; he ap-pears to regard it as self-evident, de-spite the fact that people often do notspeak in sentences and in appropriatecontexts regard single words or phrasesas "well-formed." As noted above, thesentence is a formal unit, not a behav-ioral one, though Chomsky pays littleheed to this distinction. Since thespeaking of sentences, however de-fined, typically does not display thesame dynamic properties as, say, keypecking in pigeons, he concludes, notthat he has erred in his choice of units,but that principles formulated in theexperimental analysis of behavior areof only peripheral interest in the studyof language.Of special significance to Chomsky

is the notion that humans have the ca-pacity to speak and understand an in-finite number of grammatical sentenc-es, though actual performance is lim-ited by motivation, memory, time, andother resources. There is no limit to thenumber of adjectives we can insert be-

fore a noun, or to the number of timeswe can repeat the word very for em-phasis; nor is there any limit to thenumber of sentences or clauses that wecan add or insert in other sentences, asin (9) and (10):

(9) The rat the cat the dog chased killed atethe malt.(10) Anyone who feels that if so many stu-dents whom we haven't actually admittedare sitting in on the course than ones wehave that the room had to be changed, thenprobably auditors will have to be excluded,is likely to agree that the curriculum needsrevision. (from Chomsky & Miller, 1963,p. 286)

Although native speakers gape in dis-may when asked if (10) is a grammat-ical sentence, the authors assure us thatthis "is a perfectly well-formed sen-tence with a clear and unambiguousmeaning, and a grammar must be ableto account for it if the grammar is tohave any psychological relevance" (p.286). However, it is obvious that suchsentences are not behavioral units butare strings carefully constructed to beconsistent with grammatical rules. It istrue that there are an infinite number ofsuch strings, but their relevance to ver-bal behavior is doubtful. Nevertheless,Chomsky uses the notion that there arean infinite number of grammatical sen-tences to dismiss the use of the termprobability in discussions of languageand particularly to criticize Skinner'sanalysis of language as a repertoire ofverbal operants:

It is unclear what sense there would be tothe assertion that a person has "learned" asentence that takes twice as long to say ashis entire lifetime. . . On empiricalgrounds, the probability of my producingsome given sentence of English-say, thissentence or the sentence "Birds fly" or"Tuesday follows Monday," or whatever-is indistinguishable from the probability ofmy producing a given sentence of Japanese.Introducing the notion of "probability rel-ative to a situation" changes nothing, atleast if "situations" are characterized onany known objective grounds. (Chomsky,1969, p. 267)But what does it mean to say that somesentence of English that I have never heardor produced belongs to my "repertoire,"

48 DAVID C. PALMER

but not any sentence of Chinese (so that theformer has a higher "probability")?(Chomsky, 1971, p. 20)

According to Chomsky this followsfrom the fact that most sentences areunique and hence have a probabilitynear zero.Chomsky is making an extraordi-

nary leap from asserting that a gram-mar can generate an infinite number ofsentences to asserting that humanshave the competence to generate andunderstand an infinite number of sen-tences. This is clearly not an empiricalfact. It is not even a valid generaliza-tion from the empirical fact that behav-ior is variable. Let us suppose that wehave arrived at a definition of sentencethat allows us to determine when a sen-tence has been uttered. We have no jus-tification for predicting future variabil-ity until we analyze the variables ofwhich a sentence is a function. If wecan show that these are infinitely var-iable, and that human behavior tracksthe full range of this variability, thenwe are perhaps justified in predictingthe infinite variability of sentences.However, behavior and its controllingvariables are not divisible into an infi-nite number of orderly pairs, a pointmade by Skinner as early as 1935. Wecan illustrate the point by consideringthe "language" of honey bees.As is well known, a bee, having re-

turned from successful foraging, willfly in a pattern with a distinctive ori-entation, depending on the position ofthe sun and the location of the foodsource. Other bees, observing this pat-tern, will successfully locate the foodsource. As a circle has an infinite num-ber of diameters, so there are an infi-nite number of orientations of a patternof flight. Undoubtedly no two bee"'sentences" have ever been identical.However, this variability is irrelevant ifit is not functionally related to the lo-cation of the food. Clearly no honeybee can discriminate an infinite numberof patterns, either as a "speaker" or asa "'listener." Although an abstractcharacterization of bee communicationcould generate an infinite number of

"sentences," it is likely that bees gen-erate or respond appropriately to morethan a hundred or so. (Note that sincebees have other ways of locating flow-ers, this number would be more thansufficient to satisfy the contingenciesof natural selection.) To argue that beeshave the "competence" to interpret aninfinite number of patterns is to con-fuse a property of our formulation witha property of the organism.We can make a similar argument

with respect to human language. Sen-tence (1 1) is indiscriminable from (10)in normal discourse.

(11) Anyone who feels that if somany more students whom we haven'tactually admitted are sitting in on thecourse than ones we have that theroom had to be changed, that probablyauditors will have to be excluded, islikely to agree that the curriculumneeds revision.If the two sentences are in print, wecan detect a physical difference inthem, given a pencil and enough time,but we do so in a purely mechanicalway, analogous to comparing signa-tures in a forgery case. We clearly donot do so on the basis of grammar.Once again, talk of competence is mereinvention. There is no behavioral jus-tification for calling these strings dif-ferent stimuli, or, if emitted, differentresponses. Nonetheless, they are differ-ent sentences as defined by Chomsky.Evidently the sentence is an inappro-priate unit of analysis of verbal behav-ior. Dropping it in favor of an empiri-cally defined unit not only avoids theproblem that only an infinitesimal frac-tion of all sentences are discriminable,it accommodates the awkward fact thatpeople often do not speak in sentencesat all. Moreover, it obviates the need tofind a translation between the formalapparatus and actual data. That is, weno longer need to find an operationaldefinition of "sentence" to match theformal one.

But when we have abandoned thesentence as a unit of analysis and thenotion that language consists of an in-finite number of sentences, the argu-


ment against an analysis of languageas a repertoire of verbal operantsbreaks down. It now makes sense tosay of a string of phonemes thatwould take twice as long to say asone's lifetime that it simply is not aunit of behavior, and it now makessense to ask whether a particular unitis in one's repertoire.

BACK TO THEVERBAL OPERANT

By choosing the sentence as his unitof analysis, Chomsky has been led tomaintain that grammar is central to lan-guage and that grammar must be ge-netically determined. Since extractinga completely adequate grammar fromsamples of speech is an achievementthat has eluded many years of effort bylinguists, surely it could not be accom-plished by every 3-year-old unless thejob were, in important respects, genet-ically coded.As we have seen, this offers only the

illusion of an explanation, since wemust now explain the origin of thecode in the genes, a task for whichevolutionary principles are ill-suited.Moreover, any innate device must re-spond to actual physical events, notmetaphors or abstractions; unlessgrammatical terms can be definedphysically or operationally there is lit-tle reason to believe that such a deviceis possible. Chomsky and his col-leagues have analyzed formal proper-ties of language in commendable detailand have found a wealth of curiousregularities that deserve explanation.However, they have not advanced thefunctional analysis of verbal behaviorat all.When we turn from the sentence to

the verbal operant as a unit of analysis,we avoid many of the problems facedby a formal analysis. Our terms areempirically defined, and the principlesinvoked are clearly adaptive. Chom-sky's arguments notwithstanding, nov-elty and diversity are not problems fora functional analysis. The tremendousdiversity in language, like the tremen-

dous diversity of living organisms, is afunction of selecting contingencies in adiverse environment.

Methodological problems remain.Owing to ethical constraints, it maynever be possible to account for verbalbehavior to the satisfaction of the mostcautious critic. At the moment perhapsthe best we can do is to continue thework that Skinner and others have be-gun; analyze complex verbal contin-gencies informally while attacking ex-perimentally the more tractable prob-lems in verbal behavior.

REFERENCES

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

Chomsky, N. (1959). Review of Verbal Be-havior by B. F Skinner. Language, 35, 26-58.

Chomsky, N. (1965). Aspects of a theory ofsyntax. Cambridge, MA: MIT Press.

Chomsky, N. (1969). Some critical assump-tions in modem philosophy of language. InS. Morgenbesser, P. Suppes, & M. White(Eds.), Philosophy, science and method(pp. 260-285). New York: St. Martin'sPress.

Chomsky, N. (1971, December 30). The caseagainst B. F Skinner. The New York Reviewof Books, 17, 18-24.

Chomsky, N. (1975). Reflections on lan-guage. New York: Pantheon.

Chomsky, N. (1980a). On cognitive struc-tures and their development: A reply toPiaget. In M. Piattelli-Palmarini (Ed.), Lan-guage and learning: The debate betweenJean Piaget and Noam Chomsky (pp. 35-52). Cambridge, MA: Harvard UniversityPress.

Chomsky, N. (1980b). Rules and represen-tations. New York: Columbia UniversityPress.

Chomsky, N. (1980c). Rules and represen-tations (with author's responses). The Be-havioral and Brain Sciences, 3, 1-61.

Chomsky, N., & Miller, G. A. (1963). Intro-duction to the formal analysis of naturallanguages. In R. D. Luce, R. R. Bush, &E. Galanter (Eds.), Handbook of mathe-matical psychology (Vol. 2, pp. 269-321).New York: Wiley.

Skinner, B. F (1935). The generic nature ofthe concepts of stimulus and response.Journal of General Psychology, 12, 40-65.

50 DAVID C. PALMER

Skinner, B. F (1957). Verbal behavior. NewYork: Appleton-Century-Crofts.

Skinner, B. F (1966). The phylogeny and on-togeny of behavior. Science, 153, 1205-1213.

Winograd, T (1977). On some contestedsuppositions of generative linguistics aboutthe scientific study of language. Cognition,5, 151-179.

Chomsky's Nativism: A Critical Review

Documents