Cognition, 50 (1994) 315-346 OOlO-0277/94/$07.

00 @ 1994 - Elsevier

Science

B.V. All rights reserved.

Ever since language and learning: afterthoughts on the Piaget-Chomsky debate _

Massimo Piattelli-Palmarini*
Dipartimento di Scienze Cognitive, Istituto San Raffaele, Via Olgettina 58, Milan0 20132, Italy Center for Cognitive Science, MIT, Cambridge MA 02139, USA

Abstract The central arguments and counter-arguments presented by several participants during the debate between Piaget and Chomsky at the Royaumont Abbey in October 1975 are here reconstructed in a particularly concise chronological and logical sequence. Once the essential points of this important exchange are thus clearly laid out, it is easy to witness that recent developments in generative grammar, as well as new data on language acquisition, especially in the acquisition of pronouns by the congenitally deaf child, corroborate the language specificity thesis defended by Chomsky. By the same token these data and these new theoretical refinements refute the Piagetian hypothesis that language is constructed

Correspondence to: Massimo Piattelli-Palmarini, Dipartimento di Scienze Cognitive, Istituto San Raffaele, Via Olgettina 58. Milano, 20132, Italy. I am in debt to Thomas Roeper for his invitation to give a talk on the Piaget-Chomsky debate to the undergraduates in linguistics and psychology, at the University of Massachusetts at Amherst, in April 1989. The idea of transforming it into a paper came from the good feedback I received during that talk, and from a suggestion by my friend and colleague Paul Horwich, a philosopher of science, who had attended. Steven Pinker reinforced that suggestion, assuming that such a paper could be of some use also to the undergraduates at MIT. Noam Chomsky carefully read the first draft, and made many useful suggestions in the letter from which I have quoted some passages here. Paul Horwich, Morris Halle and David Pesetsky also offered valuable comments and critiques. Jerry Fodor stressed the slack that has intervened in the meantime between his present position and Chomskys. inducing me to revise sections of the first draft (perhaps the revisions are not as extensive as he would have liked). The ideas expressed here owe a lot to a lot of people, and it shows. I wish to single out, however, my special indebtedness to Noam Chomsky, Jerry Fodor, Jacques Mehler, Jim Higginbotham, Luigi Rizzi. Ken Wexler, Laura-Ann Petitto, Lila Gleitman, Steve Gould and Dick Lewontin. The work I have done during these years has been generously supported by the Alfred P. Sloan Foundation, the Kapor Family Foundation, the MIT Center for Cognitive Science, Olivetti Italy and the Cognitive Science Society. I am especially indebted to Eric Wanner for initial funding.

SSDZ 0010-0277(93)00610-J

316

M.

Piattelli-

Palmarini

I Cognition

50 (I 004) 315-346

upon

abstractions

from

sensorimotor

schemata.

Moreover,

in the light of modern roots of cogniall this accrues learning systems. as

evolutionary tion, language a transition

theory,

Piagets

basic assumptions paradoxical

on the biological In hindsight. against conceptual argument

and learning

turn out to be unfounded. to mmore powerful

to the validity of Fodors from

seemingly

less powerful

I. Introduction This issue of Cognition offers a rare and most welcome invitation to rethink the whole field in depth, and in perspective. A fresh reassessment of the important Royaumont debate (October 1975) between Piaget and Chomsky may be of interest in this context. After all, the book has by now been published in ten languages, and it has been stated (Gardner, 1980) that the debate is certainly a strong contender. . as the initial milestone in the emergence of this field (i.e., cognitive science). It is not for the co-organizer. with Jacques Monod. of that meeting, or for the editor of the proceedings (Piattelli-Palmarini, 1980) to say how strong the contender is. It is a fact, however, that many of us have witnessed over the years many impromptu re-enactments of arguments and counter-arguments presented in that debate, and that if one still wants to raise today the same kind of objections to the central ideas of generative grammar as Piaget, Cellerier. Papert. Inhelder, and Putnam raised at the time, one cannot possibly do a better job than the one they did. Moreover, the most effective counters to those objections are still basically the same that Chomsky and Fodor offered at Royaumont. That debate also foreshadowed, for reasons that I shall come back to, much of the later debate on the foundations of connectionism (Pinker & Mehler, 1988; Fodor & Pylyshyn, 1988). What I will attempt to do here is support the Chomsky-Fodor line with further evidence that has become available in the meantime. In fact, as time goes by, it is increasingly clear that the pendulum is presently swinging towards the innatist research program in linguistics presented at Royaumont by Chomsky (and endorsed by Mehler with data on acquisition). and away from even the basic, and allegedly most innocent, assumptions of the constructivist Piagetian program. Lifting, at long last, the self-imposed neutrality I considered it my duty to adopt while editing the book, I say here explicitly, and at times forcefully, what I studiously avoided to say there and then. I also wish to highlight some recent developments in linguistics and language acquisition that bear clear consequences on the main issues raised during the debate.

2. The debates within the debate In hindsight, it is important Royaumont debates eventually to realize that there were at least four distinct collapsing into one, a bit like a swarm of virtual

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

317

particles

collapsing

into a single visible

track in modern

high-energy

laboratories: thought one that

the event that would happen;

actually happened; the one which we, the organizers, the one Jean Piaget hoped would happen, and the

Chomsky urged everyone not to let happen. Let me digress for a moment and sketch also these other virtual debates. Piaget assumed that he and Chomsky were bound to agree in all important matters. It was his original wording that there had to be a compromis between him and Chomsky. the preparatory desire to meet In fact, this term is recurrent throughout the debate. During phase, Piaget made it clear that it had been with Chomsky at great length, and witness his long-standing the inevitable

convergence of their respective views. As Piaget states in his invitation paper, he thought there were powerful reasons supporting his assumption. I will outline these reasons in a simple sketch. Reasons for the compromise Piagets assessment of the main points of convergence between him and Chomsky - Anti-empiricism (in particular anti-behaviorism) - Rationalism and uncompromising mentalism - Constructivism and/or generativism (both assigning a central role to the subjects own internal activity) - Emphasis on rules, principles and formal constraints - Emphasis on logic and deductive algorithms - Emphasis on actual experimentation (vs. armchair theorizing) - A dynamic perspective (development and acquisition studied in real time, with real children)

Piagets proposal was one of a division of labour, he being mostly concerned with conceptual contents and semantics, Chomsky being (allegedly) mostly concerned with content-independent rules of syntactic well-formedness across different languages. Piaget considered that the potentially divisive issue of innatism was, at bottom, a non-issue (or at least not a divisive one) because he also agreed that there is a fixed nucleus (noyaux fixe) underlying all mental activities, language included, and that this nucleus is accounted for by human biology. The only issue, therefore was to assess the exact nature of this fixed nucleus and the degree of its specificity. voiced by CellCrier and Toulmin, was to consider two The suggestion, complementary strategies: the Piagetian one, which consisted of a minimization
In Language and Learning: abbreviated as LL), pp. 23-24. The Debate between Jean Piagef and Noam Chomsky (hereinafter

31x

M. Piattelli-Palmarini

I Cognition

50 (1994) 315-346

of the role of innate of these factors-once

factors, and the Chomskian more, a sort of division

one, consisting of labor.

of a maximization

It was interesting for all participants, and certainly unexpected to Piaget, to witness that, during the debate proper, the constant focus of the discussions was on what Piaget considered perfectly obvious (allant de soi): the nature and origin of this fixed nucleus. He was heading for severe criticism from the molecular concerning major biologists present at the debate (especially from Jacob and Changeux) his views on the origins of the fixed nucleus. And he was heading for with Chomsky concerning the specificity of this nucleus.

disagreements

It can be safely stated that, while Piaget hoped for a reconciliatory settlement with the Massachusetts Institute of Technology (MIT) contingent about particular hypotheses and particular mechanisms concerning language and learning (and, in particular, the learning of language), he found himself, unexpectedly, facing insuperable disagreement about those very assumptions he hardly considered worth discussing, and which he believed were the common starting point - more on these in a moment. Piagets imperception of these fundamental differences was, in essence, responsible for the vast gap between the debate he actually participated in, and the virtual debate he expected to be able to mastermind. One had the impression that, to the very end, Piaget was still convinced he had been misunderstood by Chomsky and Fodor. In Piagets opinion, had they really understood his position, then it would have been unthinkable that the disagreement could still persist. One of Piagets secrets was his deep reliance on the intuitive, unshakeable truth of his hypothbes directrices (guiding hypotheses). These were such that no reasonable person could possibly reject them - not if he or she actually understood what they meant. One could single out the most fundamental of Piagets assumptions (Piaget. 1974) m . words that are not his own. but which may well reflect the essence of what he believed:

Piugets guiding hypothesis (hypothbe - Life is a continuum -Cognition is an aspect of life therefore - Cognition is a continuum

directrice)

This is a somewhat blunt rendition, but it is close enough to Piagets core message. Some of his former collaborators in the Geneva group, in 1985, of Piagets hypothkse expressed basic agreement that this was a fair rendition directrice (as expressed, for instance in his 1967 book Biologie et Connuissunce).
Bkbel Inhelder, personal communication

M. Piattelli-Palmarini

I Cognition 50 (1994) 315-346

319

As any historian of medieval logic could testify, if literally well-known logical fallacy (compare with the following):

taken

this version

is a

-New York is a major metropolis -Central Park is part of New York therefore - Central Park is a major metropolis

Decidedly, one does not want to impute to Piaget and his co-workers assent to a logical fallacy. Thus stated, it cannot pass as a fair reformulation. That would be too devious. A better reformulation, one that passes the logical test, would be the following:

A better heuristic version of Piagets core hypothesis - Life is (basically) auto-organization and self-stabilization in the presence of novelty - Cognition is one of lifes signal devices to attain auto-organization and self-stabilization therefore - Cognition is best understood as auto-organization and self-stabilization in the presence of novelty

This much seemed to Piaget to be untendentious and uncontroversial, but also very important. He declared, in fact, that this central hypothesis had guided almost everything he had done in psychology. In order better to understand where the force of the hypothesis lies, one must remember that he unreservedly embraced other complementary hypotheses and other strictly related assumptions. Here they are (again in a succinct and clear-cut reformulation):

Piagets additional assumptions I Auto organization and self-stabilization are not just empty metaphors, but deep universal scientific principles captured by precise logico-mathematical schemes. universal and invariable sequence of stepwise II There is a necessary, transitions between qualitatively different, fixed stages of increasing selfstabilization. III The logic of these stages is captured by a progressive hierarchy of

320

M. Piattrlli-Palmarini

I Cognition

.50 (I YY4) 31.5-346

inclusion

between

ascending

levels of abstraction one as a sub-set).

and generalization

(each

stage contains

the previous

IV The necessary and invariant nature of these transitions cannot by the Darwinian process of random mutation plus selection. Corollury V Another differing theory of biological evolution is needed (Piagets

be captured

third

way,

both from Darwins

and Lamarcks).

Piaget believed that there is a kind of evolution that is unique to man, and of the mental maturational stages.3 These are what which grants the necessity they are. and could not be anything else; moreover they follow one another in a strict unalterable sequence. The random process of standard Darwinian evolution is unable in principle (not just as a temporary matter of fact. due to the present state of biology) to explain this strict logical necessity. One the last two points the biologists. obviously. had their say, as we will see in a moment. Within this grand framework, it is useful to emphasize what were Piagets specific assumptions concerning learning and language:

Piagets

crucial assumptions

about learning

The transitions (between one stage and the next) are formally constrained by logical necessity (fermeture logique) and actually, dynamically, take place through the subjects active effort to generalize, equilibrate, unify and systematize a wide variety of different problem-solving activities. The transition is epitomized by the acquisition of more powerful concepts and schemes, which subsume as particular instances the concepts and schemes of the previous stage. Piagets crucial assumptions about language

The basic structure of language is continuous with, and is a generalizationabstraction from, various sensorimotor schemata. The sensorimotor schemata are a developmental precondition for the emergence of language, and also constitute the logical premise of linguistic
LL, p, SC).

M. Piattelli-Palmarini I Cognition 50 (1994) 31.5-346

321

structures (word order, the subject/verb/object tient/instrument relation, and so on).

construction,

the

agent/pa-

Conceptual links and semantic relations are the prime movers acquisition. Syntax is derivative from (and a mirror of) these. It was inevitable that Piaget should meet strong opposition

of language

on each of these of his with

assumptions, on their alleged joint force and on the overall structure argument. In a sense, the whole debate turned only on these assumptions, Piaget growing increasingly impatient to pass onto more important technical matters, but failing to do so, on account of the insurmountable

and more problems

presented by his core tenets. Chomsky and Fodor kept mercilessly shooting down even the most obvious and the most innocent reformulations of the basic assumptions of the Piagetian scheme, notably in their many spirited exchanges with Seymour Papert, who boldly undertook the task of systematically defending Piaget against the onslaught. The debate was not the one Piaget had anticipated, and it became clear to everyone, except possibly to Piaget himself (see his Afterthoughts),4 that no compromise could possibly be found.

3. Another

virtual debate: the one the organizers

thought they were organizing

There was, as I said, another virtual debate, the one which the organizers molecular biologists with a mere superficial acquaintance of cognitive psychology and linguistics - believed they were organizing. It was closer to what Piaget had in mind than to the debate that actually took place, because they too anticipated some kind of convergence. How could that be? How could we, the biologists in the group, believe for a moment that some form of compromise could be reached? The simple answer to this, in retrospect, is: ignorance. What we thought we knew about the two systems was simple and basic. I think I can faithfully reconstruct it in a few sentences:

What we (the biologists) thought we knew About Piaget:

-There is a stepwise development of human thought, from infancy to adulthood, through fixed, qualitatively different stages that are common to all cultures, though some cultures may fail to attain the top stages. - Not everything that appears logical and necessarily true to us adults is so
LL, pp. 278-284.

322

M. Piattelli-Palmarini

I Cognitwn

50 (lYY4)

31.5-346

judged by a child, differences lie. - Constructivism,

and

vice versa.

Suitable

experiments

show

where

the

a variant

of structuralism,

is the best theoretical

framework

to explain the precise haviorism, constructivism role of logical deduction. - Set theory and propositional the theory. About Chomsky:

patterns of cognitive development. Unlike bestresses the active participation of the child and the calculus are (somehow) central components of

-There are linguistic universals, common to all the different languages the world over. - These are not superficial, but constitute a deep structure.+ -This deep structure is innate, not learned, and is unique to our species. - Formal logic and species-specific computational rules are (somehow) involved in determining deep syntactic structures. - Syntax is autonomous (independent of semantics and of generic conceptual contents). - There are syntactic transformations (from active to passive, from declarative to interrogative, etc.) that preserve the deep structure of related sentences. Semantics links up with syntax essentially at this deep level. - Behaviorism is bad, while innatism and mentalism are OK. is OK. Linguistics and psychology are, at - The expression mind/brain bottom, part of biology.

The organizers, in fact, knew very little, but they liked what they knew, on both sides. There was every reason (in our opinion) to expect that these two schools of thought should find a compromise, and that this grand unified metatheory would fit well within modern molecular biology and the neurosciences. structures, on universals, on precise Both systems relied heavily on deeper logico-mathematical schemes, on general biological assumptions. This was music to a biologists ears. All in all, it was assumed that the debate would catalyze a natural scientific merger, one potentially rich in interesting convergences and compromises.

4. Chomskys

plea for an exchange,

not a debate

Commenting on a previous version of the present paper, Chomsky has insisted that he, for one, had always been adamant in not wanting a debate, but rather an
There was at the time some confusion and universal grammar. among non-experts between the terms deep structure

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

323

open and frank discussion, devoid frontiers: I am a little uneasy about Piaget Piaget

of pre-determined positions and pre-set presenting the whole thing as a Chomsky-

debate. Thats not the way I understood it, at least, and I thought that didnt either, though I may be wrong. As far as I understood, and the only to participate, there was a conference (not debate)

way I would have even agreed

on a range of controversial issues, which was opened by two papers, a paper by Piaget and my reaction to it, simply in order to put forward issues and to open the discussion.5 Chomsky then adds: Debates are an utterly shouldnt exist in a reasonable world. In a debate, irrational institution, which the assumption is that each

participant has a position, and must keep to this position whatever eventuates in the interchange. In a debate, it is an institutional impossibility (i.e., if it happened, it would no longer be a debate) for one person to say to the other: thats a good argument, I will have to change my views accordingly. But the latter option is the essence of any interchange among rational people. So calling it a debate is wrong to start with and contributes to ways of thinking and behaving that should be abandoned. After pointing out that, as is to be expected in any ongoing scientific activity, his views are constantly changing and are not frozen into any immutable position, Chomsky insists that neither he, nor Fodor, nor the enterprise of generative grammar as a whole, are in any sense an institution, in the sense in which in Europe Marxism, Freudianism, and to some extent Piagetism, are institutions. The following also deserves to be quoted verbatim from his letter: There is, thank God, no Chomskyan view of the world, or of psychology, or of language. Somehow, I think it should be made clear that as far as I was concerned at least, I was participating by helping open the discussion, not representing a world view. These excerpts from Chomskys letter should make it very clear what his attitude was. But it is well beyond anyones powers now to un-debate that debate, partly because it is the very subtitle of the book (The debate between Jean Piaget and Noam Chomsky), and partly because the community at large has been referring to the event in exactly those terms for almost two decades. So, after having made clear which kind of virtual non-debate Chomsky assumed one should have organized, let us finally return to what actually happened.

5. The real debate From records, now on, lets faithfully from the recorded tapes, attempt to reconstruct, from the published and from the vivid memory of some of those

With Chomskys permission, this, and the following, Piattelli-Palmarini, dated May 8, 1989.

are verbatim

quotes

from

a letter

to M.

323

M,

Pia~telli-Palmarini

I Cognition

SO ( 1994) 3 IS.346

who were present, into the real one. the

how all these imaginary,

unlikely,

virtual

debates

precipitated

Chomskys written reply to Piaget, made available a couple of months before debate. rightly stressed, among other things, the untenability of Piagets of evolution. Not until the first session of the debate proper had realized that Piaget was (Heaven forbid!) a Lamarckian. It was, however. clear from his distributed invitation paper that he had a curious idea of

conception anyone already

how genes are assembled and of how evolution acts on gene assemblies. Chomsky clearly had got it right and Piaget had got it wrong. This was the first important point in favor of Chomsky. Moreover, Chomsky stressed the need for specificity, while Piaget stressed the need for generality. The concrete linguistic examples offered by Chomsky seemed indeed very, very remote from any generalization of sensorimotor schemata. Some participants already felt sympathetic to Chomskys suggestion that one should not establish any dualism between body and mind, and that one should approach the study of mental organs exactly in the way we approach the study of the heart, the limbs, the kidneys, etc. Everything he said made perfect sense and the concrete linguistic examples (which Piaget and the others never even began to attempt to deal with) made it vastly implausible that syntactic rules could be accounted for in terms of sensorimotor schemata. Chomskys arguments against learning by trial and error were compelling - very compelling. One clearly saw the case for syntax, but one may still have failed to see the far-reaching import of his arguments for learning in general. For this. the participants had to wait until Fodor made his big splash at the meeting. But lets proceed in chronological order. Most important, to some of the biologists. was the feeling. at first confused, but then more and more vivid, that the style of Chomskys argumentation, his whole way of thinking, was so deeply germane to the one we were accustomed to in molecular biology. On the contrary, Piagets biology sounded very much like the old nineteenth-century biology; it was the return of a nightmare, with his appeal to grand unifying theories. according to which life was basically this or that, instead of being what it, in fact. is. Chomskys call for specificity and his reliance on concrete instances of language were infinitely more appealing. It became increasingly clear to the biologists at Royaumont that Chomsky was our true cotzfrPre in biology and that the case for syntax (perhaps Ott/y for syntax) was already lost by Piaget. As the debate unfolded. the participants were in for further surprises and much more startling revelations. In order not to rcpcat needlessly what is already in full length in the book itself, lets recapitulate only the main turning points of the debate.

LL. pp. T-52

M. Piattelli-Palmarini

I Cognition

50 (1994) 315-346

325

5.1.

The mishaps of phenocopies deeper turned probing into his rather peculiar He actually idea of phenocopy, Piaget

Upon indeed

out to be a Lamarckian.

believed

in some feedback,

however devious and indirect, from individual experience to the genetic make-up of the species. The biologists were aghast! Jacob made a marvelous job of politely and respectfully (Monod setting the record straight on phenocopies, have been aided carried by Changeux away by the was not present, and maybe he would

discussion, behaving slightly less courteously to Piaget than Jacob and Changeux did. Monod, haunted by the memory of the Lyssenko affair, always reacted to Lamarckism by drawing his gun!) Well, believe it or not, Piaget was unruffled. He had the stamina to declare himself tr& surpris by the reactions of the biologists, and reject Jacobs rectifications, quoting a handful of pathetic heretics, obscure Lamarckian biologists who happened to agree with him. The alienation of Piaget from mainstream biology was consummated there and then; patently, he did not know what he was talking about. (The young molecular biologist Antoine Danchin undertook, after the meeting, the task of making this as evident as it had to be made).X Subsequent exchanges with CellCrier and Inhelder showed that they had no alternative explanation to provide for the linguistic material brought in by Chomsky. When they mentioned linguistic examples, these were of a very peculiar generic kind, nowhere near the level of specificity of Chomskys material. They pleaded for an attenuation of the innateness hypothesis, so as to open the way to the desired compromise. But Chomskys counter was characteristically uncompromising: first of all, the high specificity of the language organ, and, therefore its innateness, is not a hypothesis, it is a fact, and there is no way one may even try to maximize or minimize the role of the innate components, because the task of science is to discover what this role actually is, not to pre-judge in advance how much of it we are ready to countenance in our theories. Second, it is not true that Chomsky is only interested in syntax, he is interested in every scientifically approachable aspect of language, semantics and conceptual systems included. These too have their specificity and there are also numerous and crucial aspects of semantics that owe nothing to sensorimotor schemata, or to generic logical necessity - no division of labor along these lines, and again no compromise _ The salient moments of this point in the debate can be summarized as follows:

LL, pp. 61-64. LL, pp. 356-360.

326

M. Piattelli-Palmarini

/ Cognition SO (1944) 315-346

Counters to Piaget from the biologists Jacobs counter: is made only by structures which are there already and which

- Autoregulation

regulate minor variations within a heavily pre-determined - Regulation cannot precede the constitution of genetically tory structures. - (Gentle reminder) genes. Individual experience cannot

range of possibles. determined regulainto the

be incorporated

Piaget simply did not see the devastating effect of Jacobs counters on his private and idiosyncratic conception of evolution by means of autoregulation. Cellerier was visibly embarrassed by Piagets anti-Darwinism and tried, I think unsuccessfully, to disentangle the personal attitudes of Piaget in matters of biological evolution from the objective implications of the Darwinian theory for psychology proper.

5.2.

The mishaps of precursors came another major

During the next session, when Monod was also present, counter, on which Fodor quickly and aptly capitalized:

Monods counter _ If sensorimotor schemata are crucial for language development, then children who are severely handicapped in motor control (quadriplegics, for instance) should be unable to develop language, but this is not the case. - Znhelders answer: Very little movement is needed, even just moving the eyes. - Monods and Fodors punch-line: Then what is needed is a triggering experience and not a bona fide structured precursor. Once again, it was the impression of several participants that the weight of this counter was not properly registered by the Piagetians. Yet the Monod-Fodor argument was impeccable, and its conclusion inevitable. One thing is a triggering input, quite another a structured precursor that has to be assimilated as such, and

LL, pp. 70-72 LL. p. 140.

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

327

on the basis of which a higher structure is actually built. A trigger need not be isomorphic with, and not even analogous to, the structure it sets in motion. Admitting that this precursor can be just anything you please (just moving your eyes once) is tantamount to admitting that it is nothing more than a releasing factor, in accordance with the innatist model of growth and maturation and against the literal notion of learning. Papert, for instance, went on at great length in offering primitives. the virtues of indirect, implicit These, he insisted, and only these, proposed learning and of the search for can be said to be innate, not the

highly specific structures

more fundamental, simpler up his sleeve, as we will see in a moment. (Healthy correctives to Paperts, and Piagets notion of implicit learning in the specific domain of lexical acquisition are to be found in Atkins, Kegl, & Levin, 1986; Berwick, 1985; Grimshaw, 1990; Jackendoff, 1983, 1990, 1992; Lederer, Gleitman & Gleitman, 1989; Lightfoot, 1989; Piatelli-Palmarini, 1990a; Pinker, 1989.) Before Fodors cold shower a lot of the discussion turned, rather idly, around the existence, in language, of components which are not specific to it, but are also common to other mental activities and processes. Again, a division of labour was proposed along these lines. Chomsky had no hesitation in admitting that there are also language factors that are common to other intelligent activities, but rightly insisted that there are many besides which are unique to language, and which cannot be explained on the basis of general intelligence, sensorimotor schemes, communicative efficacy, the laws of logic, problem-solving, etc. These languagespecific traits, Chomsky insisted, are the most interesting ones, and those most amenable to a serious scientific inquiry.

by Chomsky. These clearly are derived from For this illusion, Fodor had a radical cure primitives.

5.3.

Chomskys plea for specificity is an essential summary of the line he defended:

Here

Chomskys The simplest interrogatives

argument for specificity2 and therefore (allegedly) most plausible rule for the formation of

The man is here. Is the man here?

LL, pp. 90-105. lZLL, pp. 39-43.

328

M. Piattelli-Palmarim

I Cognition

.W (1994) 315-346

is the following But look at

(a structure-independent

rule):

Move

is to the front.

The man who is tall is here. Is the man who tall is here? language) Is the man who is tall here? The simple
rule is never

(bad sentence, (good sentence)

never

occurring

in the childs

even

tried out by the child. Why? acquired by the child is not simple (in this

The correct

rule,

uniformly

transparent and shallow sense of the word) and involves abstract. specifically linguistic notions such as noun phrase. Therefore it is not learned by trial and error and is not derivative on sensorimotor schemata. (What could the motor equivalent of a noun phrase conceivably be?)

This is, somewhat bluntly put. the core of the argument. If the process were one of induction. of hypothesis formation and confirmation, we should expect to see the simplest and least language-specific rules being tried out first. But this is not what we observe. More specific data on language acquisition in a variety of languages and dialects (Berwick & Wexler. 1987; Chien & Wexler. 1990; Guasti. 1993; Jusczyk & Bertoncini. 1988; Lightfoot, 1989; Manzini 8r Wexler, 1987; Wexler, 1987; Wexler, 1990; Wcxler &r Manzini, 1987) by now make the case against learning syntax by induction truly definitive. We will come back to this point.

Chomskys argument against any constraints j

derivation of syntactic rules from

generic

We like each other = each of us likes the others WC expect each other to win = each of us expects Near-synonymous each other BUT *We expect
ILL. pp. 113-117.

the others

to win

expressions: = each. the others

John

to like each other

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

329

is NOT Each WHY?

well formed of us expects There

and is NOT John

synonymous

with

to like the others logical or communication-theoretical explanation. structures of the

is no obvious

(There arent even non-obvious ones, at that). The linguistic rule is of the following kind. In embedded form . . . X...[...Y...]

where X and Y are explicit or understood components (names, pronouns, anaphoric elements etc.) no rule can apply to X and Y if the phrase between brackets contains a subject distinct from Y. The nature of this rule is specifically linguistic: the rule has no conceivable sensorimotor counterpart, nor any justification in terms of general Further confirming evidence (just apply the rule): The men heard stories about each other. *The men expect John to like each other. Who did the men hear stories about? *Who did the men hear Johns stories about? John seems to each of the men to like the others. *John seems to the men to like each other. Evidence from another language: intelligence.

(OK) (bad) (OK) (bad) (OK) (bad)

Jai laisse Jean manger X. Jai laisse manger X a Jean.

(both

OK)

These are apparently freely interchangeable broken in the next example:

constructions,

but the symmetry

is

Jai tout laisse manger a Jean. *Jai tout laisd Jean manger.

(OK) (bad)

NB: Update. These phenomena have received much better and deeper explanations in recent linguistic work, in terms of complete functional complexes (for a summary, see Giorgi & Longobardi, 1991; Haegeman, 1991). The overall thrust of Chomskys argument for specificity comes out further reinforced.

330

M. Piattelli-Palmarini

I Cognition SO (1994) 315-346

Conclusion: These rules are tacitly known by the speaker, but they are neither learned some even

(by induction, prcblem-solving or trial-and-error), nor determined by genera1 necessity. Genera1 intelligence and sensorimotor schemata cannot begin to explain what is happening.

Chomskys point failed to impress Piaget and the Piagetians. A lot of their counter-arguments turned on the possibility of explaining these facts in some other way. One could not fail, I think, to be impressed, there and then. by the fact that no other way was actually proposed, but that it all turned around the sheer possibility that some other rule, at some other level, might explain all of the above. (Anthony Wilden even tried out Russellian logical typeslJ to no avail.) Wisely, and unflinchingly, Chomsky kept replying that this might well be the case, but that he did not expect it to be the case. (In fact, many years have gone by, and these alternative explanations are still sorely missing - for a precise account, firmly grounded in generative grammar, but altogether charitable to the Piagetian viewpoint. see Jackendoff, 1992.) And finally came Fodors jeu de massacre, one of the truly high points in the whole debate.
5.4. Fodors demise of learning

His argument was learning by means of core of the Piagetian argument at first (not let us try to simplify
Fodors argument

not limited to language, but applicable to any theory of conceptual enrichment. He went squarely against the very system. Nobody, in the other camp, really understood his even Putnam, in his critique, written after the debate),15 so it drastically, still preserving its force:
against learning by enrichment

The typical

situation

of belief

fixation

or learning:

is an exemplar
ILL. See LL. by Fodor Harvard

is not an exemplar

pp. 117~121. his exchange with Fodor in Part II of LL. pp. 143-149 and the ensuing discussion. The argument had been developed in greater detail in his 1976 essay The Language of Thought, Hassocks: Harvester. (Reprinted in 1979 by University Press.)

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

331

Inductive

hypothesis: if and only if X is Y process (for example): come to the conclusion that

X is an exemplar Target

of the learning

Y = miv = square (Alleged) stage

and gray 1: the subject has access, separately, to the concept

square, to the concept gray, but not to the conjunct square-and-gray. (Alleged) stage 2: the subject constructs a tentative new concept Y, and tries it out on the experimental materials. Y is not yet miv Stage 3: the subject miv = square-and-gray Fodors argument: If this is the case, that is, if the language at stage 3 is really more powerful than the language at stages 1 and 2, then this transition cannot be the result of learning, it cannot come from induction. Mini-proof: At some point the subject must formulate the hypothesis that Y is true of (applies to, is satisfied by) all and only those things which are mivs, that is, which are square and gray. But this cannot happen unless the subject has the concept miv. Unless Y is to all intents and purposes indistinguishable from miv (from being square and gray), we have no idea whatsoever of what Y could be. Therefore: the language of stage 1 is not weaker (less powerful, more limited) than the language of stage 3. You always start with a language which is at least as powerful as any language which you can acquire. Where do all these concepts come from? Fodors three hypotheses: correctly comes to the conclusion that Y must be:

(a) they are innate (b) God whispers them to you on Tuesdays

332

M. Piattelli-Palmarini

I Cognition

SO (I 994) 315-346

(c) you acquire

them

by falling

on your head

Dismissing hypothesis (b), the only plausible conclusion is that they are innate (and/or arise for totally endogenous reasons, due - for instance - to a stepwise hypothesis brain (c). maturation. This latter is just a slightly less fancy version of

The

audience

was really

impressed,

Monod

most of all. There

was a distinct

flavor of paradox in Fodors position and he did not try to hide this fact. The whole argument sounds paradoxical, yet it is perfectly compelling. Fodor, to his own regret, cited, just as one concrete, indubitable, patented, example of a logical system which is more powerful than another, the case of propositional logic (a provably weaker system) versus first-order quantificational logic (a provably stronger system which contains the former as a sub-system). This created a lot of misunderstanding and endless discussions about the history of the discipline of logic. Fodors thesis remained essentially unchallenged (in the book, it will be Hilary Putnam to accept the challenge, but that was after the debate). In the aftermath of Fodors onslaught on induction and learning, most of the ensuing debate revolved, on the one hand, around plain misunderstandings of his and Chomskys position (leading to clarifications and reformulations) and on the other, around a rather idle insistence that their position appears vastly implausible and even paradoxical. This apparent implausibility was never denied by Fodor and Chomsky, but with the crucial proviso that the appearance of paradox persists only if we maintain the traditional assumptions of the domain. Their main point, however, was to subvert these very assumptions, not to maintain them. I surmise that, then as well as now, those who have concluded that the debate was won by Piaget did so solely on the grounds that the theses defended by Piaget sounded intuitively very plausible, while the theses presented on the other side sounded preposterous. It does not seem to occur to them that, in science, even preposterous hypotheses often turn out to be true (Piattelli-Palmarini, 1986, 1989).

5.5.

Sequels

to the debate of explaining why these construct an attenuated Fodors positions (even Piagets).17 This line of

After the debate, Putnam, at least, took the trouble theses sound preposterous, and made an attempt to version of what he thought was right in Chomskys and suggesting that they come closer than they believe to

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

333

resistance against strong innatism and meaning atomism in the domain of lexicalconceptual semantics has been vastly expanded in subsequent philosophical works by Putnam, Fodor, Dennett, Millikan, Loar, Burge and others. I must leave it out, for reasons of space. I must also leave out the interesting Kripkes essay on Wittgenstein (Kripke, 1982) revamping skepticism rejoinders sequels elicited a special brand met interesting by of

about the notion of following a rule, which have by Chomsky (1986), Horwich (1984), and others.

The whole recent debate on connectionism has revamped several of the anti-specificity theses already present in Piaget (the most explicit link between the theses of connectionism and the Piaget-Chomsky debate has been made, in the domain of lexical-conceptual learning, by Ray Jackendoff (1992). Connectionist architectures are, in fact, a concrete embodiment of the idea of order-fromnoise. At the time of the debate Piaget could only summon in defense of the order-from-noise paradigm the physicochemical theories of Prigogine and his school, and some rather confused speculations by Heinz von Foerster (1960). Had he lived long enough to see present-day connectionism, I am persuaded that he would have endorsed it wholeheartedly. In fact, an implicit alliance between Piagetism and connectionism is amply consummated (Elman, 1989). This is not the place and time to re-examine the controversy on connectionism (Pinker & Mehler, 1988), but I wish to stress that many of the recent polemics do find their roots, ante litteram, in the Royaumont debate.

6. What happened

ever since in linguistics

and language

acquisition

In the rest of this paper, I will briefly present a number of further developments which support the positions defended by Chomsky and Fodor. And I will do it precisely by showing that their theses, preposterous as they may have seemed, are presently the only plausible explanation for a variety of facts concerning language, language acquisition and cognitive development. The swing of the pendulum in their favor has not only continued, but has gained further momentum.

6.1.

The new turn in linguistic

theory

From October 1975 to the present day, the brand of linguistic theory called generative grammar has undergone an unprecedented growth. As a consequence of previous partial success (and, of course, past errors), Chomsky and others have developed the so-called government-and-binding theory (Chomsky, 1981, with important antecedents in work published in 1979 and 1980) (Chomsky, 1980), and more recently the minimalist framework (Chomsky,

334

M. Piattelli-Palmarini

I Cognition

50 (1994) 315-346

1993). Referred to also as the principles and parameters approach (Giorgi & Longobardi, 1991; Haegeman, 1991; Lasnik & Uriagereka, 1988; Rizzi, 1990; van Riemsdijk & Williams, 1986), the turn of the 1980s has been a revolution within the revolution, bringing both picture presented by Chomsky genuinely novel features and extensions to the at Royaumont. The line of argument developed

orally at Royaumont, and then sharpened in the written proceedings, has not only been preserved in the new theory, but even made more radical. Just to cite a few examples, some of the oldest and most central notions of traditional grammars (subject, object, grammatical construction, phrase-structure rules, etc.) are now demoted to epiphenomena of much deeper and much more abstract notions of generative grammar. The importance of the lexicon has grown explosively, to the point that some linguists (certainly Chomsky himself, especially after 1992, in the minimalist framework) now claim that acquiring the lexicon is almost all a child has to do, in order to acquire a language. Everything else is generated by a strictly invariant, language-specific computational system, and by the several output conditions arising at the many interfaces of this computational system with other internal mental systems. Inevitably the inner structure of each lexical item (particularly the inner structure of verbs, adverbs and adjectivals) has become much richer and much more abstract than it ever was in the traditional grammars (Giorgi & Longobardi, 1991; Grimshaw, 1990; Hale & Keyser, 1993; Higginbotham, 1983; Jackendoff, 1990; Keyser & Roeper, 1992; Tenny, 1988). None of the highly abstract and tightly knit principles and parameters of universal grammar, nor any of the specific rules of the core grammar of a particular language, bear any resemblance whatsoever to derivations from non-linguistic principles (even less from sensorimotor schemata). The irrelevance of abstractions from motor schemata even in the development of sign languages is particularly striking (a point to which I will return). The exploration of possible phonemes and syllables by the congenitally deaf child, in the course of babbling in the manual mode, shows a marked linguistic specificity in its ontogenesis, not a continuity with generic manual gesturing (Petitto & Marentette, 1991). The case against any derivation of linguistic structures from nonlinguistic gestures, and/or from the perception of generic movements, appears clear-cut even for the most obvious candidates. namely time relations (Hornstein, 1990), agent/patient relations (Grimshaw, 1990; Pinker, 1989), aspectual semantics (Tenny, 1988), the geometry of events (Pustejovsky, 1988), verbs of movement and change of possession (Jackendoff, 1983, 1990, 1992), reference (Higginbotham, 1985, 1988), and even co-reference in sign languages (Keg], 1987). The richness and depth of these recent developments as a whole is a proof that the program presented by Chomsky at Royaumont was sound and productive, while the very idea of a continuum between language and non-linguistic precursors was doomed. Many of the objections raised during that debate seem to

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

335

voided by the tumultuous progress made in generative specificity of the language grammar in the last 15 years or so. The extreme system, indeed, is a fact, not just a working hypothesis, even less a heuristically convenient postulation. Doubting that there are language-specific, innate computational capacities today is a bit like being still dubious about the very existence of molecules, in spite of the awesome progress of molecular biology. There are, nonetheless, also particular data and particular developments that seem to me to refute the most fundamental tenets of Piagetian psychology (there linguistics). Limitations of never really was, nor will there ever be, a Piagetian space demand that I concentrate on these, and only these. I will do it briefly, and I will do it bluntly, because I think that these recent developments are blunt refutations of the most central classical assumptions then entertained by the Piagetian school.

me to be automatically

6.2.

The knock-down

case of pronouns

in the congenitally deaf child

First and foremost, I will sketch a perfect case against the dependence of language on motor schemes. It is based on a clear-cut, elegant and devastating piece of data from the acquisition of sign language by the congenitally deaf child. I give special privilege to this case, because it looks a lot like those decisive experiments in physics and in biology which refute, by a single stroke, longentertained hypotheses. It is due to the psycholinguist Laura-Ann Petitto of McGill, probably the only researcher who has worked in depth both with chimps and with congenitally deaf children. (By the way, if language indeed were supervenient on motor schemata and on general intelligence, then chimps ought to have language, which is clearly not the case, as definitely shown, after the Bever, Seidenberg and by Petitto Royaumont debate, by Premack, Terrace, herself) (For reviews, see Premack, 1986; Roitblat, Bever, & Terrace, 1984).

The counter-case

of pronouns

in the congenitally deaf (Petitto,

1987)

If language were continuous with (prompted by, isomorphic with, supervenient on) motor schemata, then sign languages should show this causal dependency in a particularly clear and transparent way. Within sign languages, the case of constructing personal pronouns (which superficially look a lot like pointing) out of generic pointing ought to be even more transparent. It turns out that this is not the case. Data from the acquisition of pronouns in the congenitally deaf, in fact, show that:

336

M. Piattelli-Palmurini

/ Cognition pointing appear

50 (1 WI)

31.5-.?ilh

(1) Non-linguistic are used. (2) Yet, pronouns

is present suddenly

and widely (within

used

long before

pronouns

a couple

of weeks)

and at exactly

the same age in the hearing and in the congenitally deaf child. (3) In the few weeks preceding the appearance of linguistic pointing (i.e., pronouns) in the congenitally deaf child, generic pointing temporarily disappears. (4) Paradoxical as it may seem, child make the same mistakes you, you to mean the hearing child and the congenitally deaf in their initial use of pronouns (me to mean

me, etc.)

at oneself to refer to someone else, and vice NB: This means pointing versa. This kind of mistake is uniquely linguistic (as demonstrated by the exact parallelism with the hearing child) and never happens with generic pointing. If motor skills and general intelligence were involved at all, then these mistakes would be a disturbing symptom of deep-seated motor troubles and/or of an intolerably low level of general intelligence (of course, neither is the case).

(5)

Another piece of conclusive evidence (Petitto & Marentette, 1991) is the autonomy and the specificity of babbling in the ontogeny of sign languages. There is no hope any more of deriving from motor schemata even the form of syllables in the manual mode, let alone the meaning of words and the structure of the sentence. And, since there is such a clear discontinuity even between generic gesturing and the basic elements of sign language, one can safely dismiss the alleged continuity between gestural schemata and the structure of the sentence in spoken languages.

In my opinion at least. this is precisely the way a scientific hypothesis is definitively refuted: you make it as clear, as specific and as predictive as possible, then look for an ideal experiment, one in which the phenomenon stands out unambiguously, in its purest form, then see whether the experiment confirms or refutes the hypothesis. The rest is idle discussion on vague metaphors.

6.3.

On the inexistence

of horizontal

stages

There are numerous other recent (and some not so recent) data which militate powerfully against the very existence of Piagets horizontal stages. Let us be reminded that a horizontal stage a la Piaget is one in which a concept or a logical operation is either present or absent in toto; its presence or absence

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

337

will show up no matter

which problem

you present

to the child, from mathematics

to gardening, from geometry to economics from language to moral judgement. Well, there are overwhelming reasons to believe that these horizontal capacities simply do not exist. Every direct or indirect proof in favor of the modularity of mind (Fodor, is ipso facto most salient 1983; Garfield, a disproof ones: 1987), of the content-drivenness of horizontal stages. of problem-solving, Lets see some of the of the existence

Evidence Domain

for the non-existence specificity (modularity)

of Piagets

horizontal

stages

of problem-solving:

(4 (b)

At a given age, the child who applies conservation to one kind of problem (involving volume and weight) does not apply it to other kinds of problems (involving speed, temperature, concentration, etc.). The conservation of identity is qualitatively different for different conceptual kinds (animals, artifacts, nominal kinds, etc.) and the conceptual transition to conservation of identity or kind takes place at different ages for each of the different kinds. (For reviews, see: Carey, 1985; Keil, 1979, 1986; Markman, 1989) Piagetian illogicalities can be elicited also in the adult:

Typical

(4

The power of what Amos Tversky and Daniel Kahneman have called typicality, anchoring and ease of representation (Kahneman, Slavic, & Tversky, 1982) has nothing to do with having versus not having a concept. It has to do with the domain to which our intuitions of what is typical apply. Cognitive strategies, and the underlying heuristics and biases often do not generalize from one domain to the next, not even in the adult. There would be no end to the succession of stages, well up to, and beyond, the level of Nobel laureates (Piattelli-Palmarini, 1991, 1993). Some of our intuitions about typicality are the opposite of what we should derive from actual experience (the signal case is offered by intuitions about probability). In many cases, there is no abstractive assimilation at all of objective external structures by the mind, and even less so a mandatory, logically determined one.

(b) (4

Many of the typical experiments a la Tversky and Kahneman the qualitative results obtained by Piaget and his collaborators

replicate exactly on children. The

33x

M. Piattelli-Palmarini

I Cognition SO (1994) 31.5-346

simple

secret

is to change

the domain.

The typical

experiment

a la Piaget

was to

elicit from, say, a Syear-old the judgement that there are more girls than children in a photograph, more roses than flowers in a bouquet, more cars than vehicles in a drawing, etc. The explanation concepts of set/subset/super-set. that there ----ing (where ----i-They judge, just like the Syear-old child, that the subset has more members than the super-set. These same educated adults will judge that there are more words of English which begin with an r than there are which have r in the third-to-thelast position (the reality is that there are vastly more words of the second kind than of the first). Do they lack the concept of subset? No! The correct explanation has to do with our intuitions of what is most typical of a kind, and what we can easily represent mentally. The easier it is for us to mentally generate typical members of the set, the larger the set appears to us. The same applies to the child: it is easier to mentally generate typical instances of a simple set (boys, roses, flowers, etc.) than to generate instances of a disjunctive set (children = boys or girls; flowers = roses or carnations or. . .; vehicles = cars or trucks or .). What counts is familiarity with, ease of representation of, and typicality of, the standard exemplar of a specific set, in a specific domain. There is no horizontal and biases are never lack of a certain concept everywhere. These heuristics horizontal, but always vertical, domain-specific, in a word, modular. each - stands for a letter whatsoever) than there are of the form are more seven-letter was given by Piaget as: Lack of possession of the But the majority of highly educated adults judge words of the form

6.4. Further counters from linguistics and from biology Let me now come back to language proper, and to evolutionary conclude with a drastic simplification of recent progress in these bears direct, and rather final, negative consequences for the hypotheses, as presented at Royaumont. biology. I will domains which core Piagetian

Recent developments (I chose schemata,

in linguistic theory exclude learning any continuity with an empty metaphor.) sensorimotor

those which further and make language

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

339

- The essentials of syntactic structures are derived (projected) directly from to the lexicon). the lexicon (and, of course, there are no motor equivalents - Rules are replaced by principles and parameters. Most (maybe all) parameters have only two possible positions. Learning a given language means acquiring the lexicon and (in the most recent minimalist framework (Chomsky, 1993), one should rather say thereby) setting the correct Lightfoot, values 1989; for all the parameters. (Language acquisition

is not an induction,

but a selection:

Piattelli-Palmarini, 1989.) - There are in every natural language (sign languages included) silent elements, phonetically inexpressed particles called empty categories, and these cannot be learned, because they are not part of the sensory, explicit, input to the learner. (Language acquisition cannot be based on imitation, generalization and assimilation.) _ Linguistic principles are highly specific, they bear no resemblance to general and have no explanation in terms of communicative laws of thought, efficacy. (Self-regulation, adaptation and pragmatic expediency explain nothing at all in this domain.) (The best, yet still unconvincing, adaptationist reconstruction is to be found in Pinker & Bloom, 1990, see also the peer commentaries to that paper.) -The form of linguistic principles is very specific, mostly stating what cannot be done to highly abstract and uniquely linguistic elements, categories and constructs (based on notions such as c-command, X-bar, PRO, projection of a lexical head, trace of a nounphrase, specifier of an inflectional phrase, etc.). The typical principle of universal grammar sounds a bit like the following: do whatever you please, but never do such-and-such to so-and-so.

(There is no hope, not even the dimmest one, of translating these entities, these principles and these constraints into generic notions that apply to language as a particular case. Nothing in motor control even remotely resembles these kinds of notions.) (For a clear, global presentation of this theory, see Haegeman, 1991. For the recent minimalist framework, see Chomsky, 1993; for the parametric approach to language acquisition, see Lightfoot, 1989; Manzini & Wexler, 1987; Piattelli-Palmarini, 1989; Roeper & Williams, 1987; Wexler, 1982; Wexler & Manzini, 1987; and the vast literature cited in these works. For the existence of empty categories in sign languages, see Kegl, 1986, 1987.)

Last, but not least, modern biology and evolutionary and more radical reasons to refute Piagets basic tenets tion:
Knock-down arguments from modern evolutionary

theory offer further about life and evolu-

theory

- No inheritable

feedback

is even remotely

possible

from individual

experience

to the genes. - The metaphor of problem-solving as a driving force in evolution (in particular in speciation; Schull. 1990) 1s wrong: each species creates its own specific problems 1990b, 1990~). (Lewontin. 1982, 1983, 1990b. Piattelli-Palmarini, 1989,

-Novelty and complexification do not logically, nor even factually, imply an enrichment. since they often arise as a consequence of impoverishment and specialization. (The possibility of evolutionary complexification by means of impoverishment was first demonstrated in bacteria by Boris Ephrussi; see Jacob, 1977, 1987.) - Life is basically what it is: the old grand theories (auto-equilibration, minimization of disturbance. increasing autonomization, increasing adaptation, increasing order from noise, etc.) have never explained anything. It proved impossible to deduce biological structures and functions from first principles. - Even the most transparent (one would have said) instances of adaptations of organs-cum-behaviors to environmental conditions arc sometimes fallacious. (Ten different elaborate kinds of mouthpiece organs for cutting, crunching, searing and syphoning have evolved in insects one hundred million years before there were any flowers on earth (Labandeira & Sepkoski, 1993). Until very recently (July 1993) these organs had universally and obviously been judged to offer examples of exquisite and fine-tuned selective adaptations to the environment and to the mode-of-life of their bearers.) _ Biological evolution is not (at least not always) gradualistic (Eldredge & Gould. 1972: Gould, 1984; Gould & Eldredge, 1977. 1993; Gould Br Lewontin, 1984; Gould & Vrba, 1982) and does not (at least not always) proceed through a stepwise combinatorial enrichment out of pre-existing more primitive structures. The brain, for one, did not evolve by piling up new structures on top of older units (Changeux. Heidmann, & Patte, 1984; Edelman, 1987). _ Selection out of a vast innate repertoire is the only mechanism of growth, acquisition and complexification which we can scientifically understand (Piattelli-Palmarini, 1986. 1989, 19YOa). (The theory of. and data on, language acquisition in the principles-and-parameters framework confirm the success of selective theories in the domain of linguistics - as rightly foreshadowed by Chomsky. Fodor. and Mchler at Royaumont.)

M. Piattelli-Palmarini

I Cognition

50 (1994) 315-346

341

7. Conclusion I may seem to have been saying rather school. This was not my intention. I think harsh things about Piaget and his there are overwhelming reasons to

conclude that his approach was fundamentally wrong, but this is no judgement on his personal merits. It took a great mind to draw such a vast and coherent picture, one that is still attractive to many developmental psychologists the world over, one that appeared as deep, novel of fields, from the philosophy and important to many researchers in a variety of science to anthropology, from ethics to

sociology, from mathematics to feminist studies. He certainly introduced, or rather reintroduced, into psychology a much-welcome rationalistic and antiempiricist stance, combined with an unerring flair for experimentation. I am told by the best present-day experimentalists in cognitive development that, even if his interpretations of the data are often wrong, the reproducibility of his original data is always next to perfect. In hindsight, and judging from a different theoretical frame, we see that often he did not perform the next inevitable check, or the decisive counter-experiment, but he never erred in what he actually did, or in telling what he actually found. Much of present-day experimentation on the or indirectly, from his classic childs cognitive development stems, directly experiments and those of his collaborators. Piaget was truly a universal thinker, with an insatiable curiosity for facts and theories well beyond his profession. He had an encyclopaedic mind, and was, alas, one of the last global intellectuals. Most of all, he brought to perfection, and elaborated down to the most minute details, a theory which was intuitively very appealing. This, as I have endeavored to show, was his strong point, and also his great weakness. The very basic intuitions, to which Piaget brought order and depth, and between which he established unprecedented systematic interconnections, have turned out to be wrong, misleading, or empty. They were, indeed, prima facie very plausible - no one would want to deny that - but often in science the implausible must triumph over the plausible, if the truth lies on the side of the implausible. This is what Piaget refused to accept, to the point that, in spite of his towering intelligence, he could not understand the message brought to him by Chomsky and Fodor at Royaumont. For the ideological reasons so well explained by Chomsky at the Royaumont debate and elsewhere, in the domain of psychology and linguistics (at odds with physics, chemistry and molecular biology) hypotheses that appear, at first blush, preposterous are often simply assumed to be wrong, without even listening to reason, proof or experiment. With these notes, I hope to contribute just a little to the demise of this strange and irrational attitude in cognitive science. Also in linguistics, in psychology and in cognitive science the prima facie implausible can turn out to be true, or close enough to the truth. In fact, my main point here, as in previous articles (Piattelli-Palmarini, 1986, 1989, 1990a), is that it already has.

342

M. Piattelli-Palmurini

I Cognition

SO (I 994) .?l.C-346

I wholeheartedly agree with what Chomsky said at the very end of the debate. Little of what we hypothesize today will survive in the long run. Twenty or fifty years from now we will probably have gained much deeper and much better insights into these matters, and not much of present-day theorizing will still be valid. But what is important is that we may look back and ascertain that those hypotheses, those explanations, the right kind. As I endeavored were at least on the right track, that they were of to show, at least this much is already happening only, I have allowed myself the liberty of

now, with respect to the debate. In this sense and in this sense

speaking of winners and losers. The race is mostly still ahead of us, and all I have offered here are arguments in favour of a certain choice for the kind of competition still to come. A final, very personal touch: I have fond memories of my conversations with Jean Piaget. I was always impressed by his bonhomie, his wit, his eager search for better understanding, his serene attitude towards life. He has run a long, difficult race, and has left a highly talented multitude behind him. No one could have led that race with greater aplomb, and no one ever will. It is no paradox, I believe, to admire him for his great achievements, but also feel sorry for the path he insisted on choosing. It was a bit painful, at least for some of us at Royaumont, to see him lose an important confrontation, one which he had eagerly sought, without fully realizing what was happening to him, and to his most cherished ideas, and why. His search for a compromise was unsuccessful, simply because the compromise was neither possible nor desirable. I heard Gregory Bateson, after the meeting, define Piaget as a lay saint. He was implying, I believe, that Chomsky and Fodor had fulfilled the ungracious role of executioners. But it would be a paradox to admire Piaget as much as Bateson did, and still wish he had been lulled by the false conclusion of a possible compromise. Not even the saints appreciate such forms of inordinate devotion.

Bibliography
B.T., Kegl, J.. & Levin, B. (1986). Explicit and implicit information in dictionaries. Lexicon Project Working Paper No. 12. Cambridge, MA: Center for Cognitive Science. MIT. Berwick. R.C. (1985). The acquisifion of syntacfic knowledge. Cambridge, MA: MIT Press. Berwick. R.C., & Weinberg, A. (1985). The grammatical basis of linguistic performance: Language asc and ucquisition. Cambridge, MA: Bradford Books/MIT Press. Berwick. R.C., & Wexler, K. (1987). Parsing efficiency, binding, c-command and learnability. In B. Lust (Ed.), Studies in fhe acquisilion of anaphora. Dordrecht: Reidel. Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: Bradford Books/MIT Press. Changeux, J.-P.. Heidmann, T., & Patte, P. (1984). Learning by selection. In P. Marler & H.S. Terrace (Eds.), The biology of learning. New York: Springer-Verlag. Chien, Y.-C., & Wexler. K. (1990). Childrens knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition, I, 225295. Chomsky. N. (1980). On binding. Linguistic Inquiry. Il. l-46. Atkins,

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

343

Chomsky. N. (1981). Lectures on government and binding (the Pisa lectures). Dordrecht: Foris. Chomsky, N. (1986). Knowledge of language: Its nature, origin, and use. New York: Praeger. Chomsky, N. (1989). Some notes on economy of derivation and representation. Cambridge, MA: MIT, Department of Linguistics and Philosophy, Working Paper no. 1. Chomsky, N. (1993). A minimalist program for linguistic theory. In K. Hale & S.J. Keyser (Eds.), The view from building 20: Essays in honor of S. Bromberger. Cambridge, MA: MIT Press. Dennett, D.C. (1987). The Intentional Stance. Cambridge, MA: Bradford Books/MIT Press. Dretske, F. (1988). Explaining behavior: Reasons in a world of causes. Cambridge, MA: Bradford Books/MIT Press. Edelman, G.M. (1987). Neural Darwinism: The theory of neuronal group selection. New York: Basic Books. Eldredge, N., & Gould, S.J. (1972). Punctuated equilibria, an alternative to phyletic gradualism. In T.J.M. Schopf (Ed.), Models in paleobiology. San Francisco: Freeman. Elman, J.L. (1989). Representation and structure in connectionist models. In Center for Research in Language Technical Report 8903. La Jolla, CA: University of California, San Diego. Fodor, J.A. (1975). The language of thought. New York: Crowell. Fodor, J.A. (1981). Representations: Philosophical essays on the foundations of cognitive science. Cambridge, MA: Bradford Books/MIT Press. Fodor, J.A. (1983). The modularity of mind. Cambridge, MA: Bradford Books/MIT Press. Fodor, J.A. (1987). Psychosemantics. Cambridge, MA: Bradford Books/MIT Press. Fodor, J.A., & Pylyshyn, Z. (1988). Connectionism and cognitive architecture: A critical analysis. In S. Pinker & J. Mehler (Eds.), Connections and symbols. Cambridge, MA, Bradford Books/ MIT Press. Gardner, H. (1980). Cognition comes of age. In M. Piattelli-Palmarini (Ed.) Language and learning: The debate between Jean Piaget and Noam Chomsky (pp. xix-xxxvi). Cambridge, MA: Harvard University Press. Garfield, J.L. (Ed.) (1987). Modularity in knowledge representation and natural-language understanding. Cambridge, MA: Bradford Books/MIT Press. Giorgi, A., & Longobardi, G. (1991). The syntax of noun phrases: Configuration, parameters and empty categories. Cambridge, UK: Cambridge University Press. Gould, S.J. (1984). Toward the vindication of punctuational change. In W.A. Berggren & J.A. Van Couvering (Eds.), Catastrophes and earth history: The new uniformitarianism. Princeton, NJ: Princeton University Press. Gould, S.J., & Eldredge, N. (1977). Punctuated equilibria: the tempo and mode of evolution reconsidered. Paleobiology, 3, 115-151. Gould, S.J., & Eldredge, N. (1993). Punctuated equilibrium comes of age. Nature, 366, 223-227. Gould, S.J., & Lewontin, R.C. (1984). The spandrels of San Marco and the Panglossian paradigm: A critique of the adaptationist programme. In E. Sober (Ed.), Conceptual issues in evolutionary biology. Cambridge, MA: Bradford Books/MIT Press. Gould, S.J., & Vrba, E.S. (1982). Exaptation: A missing term in the science of form. Paleobiology, 8, 4-15. Grimshaw, J. (1990). Argument structure. Cambridge, MA: MIT Press. Guasti, M.-T. (1993). Verb syntax in Italian child grammar: Finite and nonfinite verbs. Paper presented at the workshop on parametric language acquisition, Trieste Encounters in Cognitive Science, SISSA, Trieste (Italy), July 1993. Haegeman. L. (1991). Introduction to government and binding theory. Oxford, UK, and Cambridge, MA: Blackwell. Hale. L., & Keyser, S.J. (1987). A view from the middle. Lexicon Project Working Papers no. 10, Center for Cognitive Science, MIT. Hale, K., & Keyser, S.J. (1993). On argument structure and the lexical expression of syntactic relations. In K. Hale & S.J. Keyser (Eds.), The view from building 20: Essays in honor of S. Bromberger. Cambridge, MA: MIT Press. Higginbotham, J.T. (1983). Logical form, binding and nominals. Linguistic Inquiry, 14, 395-420. Higginbotham, J.T. (1985). On semantics. Linguistic Inquiry, 16, 547-593.

344

M. Pia~felli-Pulmarini

I Cognirion

511 (1994) .115-346

Higginbotham. J.T. (1986). Elucidations of meaning. Linguistics und Philosophy, 12, 365-517. Higginbotham. J.T. (1088). Knowledge of reference. In A. George (Ed.), Reflections on Chornsky. Oxford. Basil Blackwell. Hornstein, N. (1990). As /ime goes by: The syntux offense. Cambridge. MA: Bradford Books/MIT Press. Horwich, P. (1984). Critical notice: Saul Kripke: Wittgenstein on rules and private language. Philosophy of Science, 51, l63- I7 1. Jackendoff, R.S. (lYX3). Semaniics and cognifion. Cambridge, MA: Bradford Books/MIT Press. Jackcndoff, R.S. (1990). Semuniic .v/ructure.t. Cambridge, MA: MIT Press. Jackendoff, R.S. (19Y2). Word meanings and what it would take to learn them: ReRcctions on the Piaget-Chomsky debate. In R.S. Jackendoff (Ed.). Lan~uuges of the mind. Cambridge. MA: Bradford Books/MIT Press. Jncoh. F. (1970) Ln lo~ique dlc vivunt: Line hisloire de IhPridirP. Paris: Gallimand. American Ed. Jacob. F. (1973). The logic of life: CI history of heredity. New York: Pantheon. Jacob. F. (1977). Evolution and tinkering. Science, 19h. llhl-1166. Jacob. F. (1087). La stutue infkrieure. Paris: Odile JacobiLe Seuil. Jusczyk. P.W.. & Bertoncini. J. (1988). Viewing the development of speech perception as an innately guided learning process. Language and Speech, 32, 217-238. Kahneman, D., Slavic, P. & Tversky, A. (Eds.) (IYXZ). Judgment under uncertuin/y: Heuristics and biases, Cambridge. UK: Cambridge University Press. Kegl. J. (19X6). Clitics in American Sign Language. In H. Borer (Ed.). Tire .s~ntux of pronominal Clitics. Syntax and Semantics Series. New York: Academic Press. Kepl. J. (19X7). Corcfcrcncc relations in American Sign Language. In B. Lust (Ed.), The rtcq~ci.sition of unuphoru. Vol. 2: Applying rhe constrminfn. Dordrecht: Rcidel. Keil. F.C. ( 1979). Semuntic rend &ceprual development: At7 ontological per.~pec~~~e.Cambridge. MA: Harvard University Press. Keil, F.C. ( IYXh). The acquisition of natural kinds and artifact terms. In W. Dcmopoulos & A. Marras ( Eds.), Lunguqe lecrrning und concept ucyuisikm: Folmdufional issuev. Norwood. NJ: Ablex. Keyser, S.J.. & Rocper. 1. (lY92). Rc: The abstract clitic. Linguisric Inquiry. 23. XYS125. Klimn. E., & Bcllugi. U. (1979). The .sign.v of lunguc~ge. Cambridge. MA: Harvard University Press. Kripkc, S. (19X2). Wittgensfein on rules urld private languqe. Cambridge, MA: Harvard llniversity Press. Labandeira, C.C.. & Sepkoski. J.J. (1993). Insect diversity in the fossil record. kience. Zhl. 310~313. Lasnik. H., & Uriagereka. J. ( IYXX). A course in GB syr~tcrx: Lecrures o,z hir~ding und empry crr/egor-icr. Cambridge. MA: MIT Press. Lederer. A., Gleitman. H.. & Gleitman. L.R. (IYXY). Svnftrcric hoorstruppin~: Are rhe du/n for rr deducrire. principle-driven Lzerh learning procedwr m?ailahle to childrerz? Paper presented at the 14th Annual Conference on Language Development, Boston University (October 1089). Lewontin. R.C. (lY70). The units of selection. Annurrl ReL,iew of Ecohgy urd Sysfemaric~s. I, l-16. Lcwontin. R.C. ( 1982). Organism and environment. In H. Plotkin (Ed.), Letrrning. developnwnt. und culture. New York: Wiley. Lewontin. R.C. (IYX.?). The organism as the subject and object of evolution. Scienrirr. I/h. 6382. Now in Lcvins. R.. & Lewontin. R.C. (1985). The dialecfical hiologisf. Camhridge. MA: Harvard University Pres\. Lcwontin. R.C. ( IYYOa). The evolution of cognition. In D.N. Oshcrson & E.E. Smith (Eds.). An invircrtion 10 cognitive .ccience. Vol. 111: Thinking. Cambridge, MA: MIT Press. Lewontin. R.C. (lY9Ob). How much did the brain have to change for speech! Commentary of (Pinker bi Bloom 1990). Reltuvioral and Brain Sciences. 1.3. 740~741. Lightfoot. D. ( IYXY). The childs trigger experience: Degree-O learnability. ~ehuviorcr/ cmd Brrruz SciPrKr.s. II. 32 l-375. Loar, B. (19X1). Mind cmd meaning. Cambridge, UK: Cambridge University Prcsa. Manzini. M.R.. & Wexler. K. (lYX7). Parameters, binding theory and learnability. Lin,quistic Inyuiry, 18. 413-344. Markman. E.M. ( IYXY). Cafegorizutiotl cmd naming iti children: Problems of rnductiorr. Cambridge, MA: Bradford Books/MIT Prcsa.

M. Piattelli-Palmarini I Cognition 50 (1994) 315-346

345

McCloskey, M. (1983). Naive theories of motion. In D. Gentner & A.L. Stevens (Eds.), Mental models. Hillsdale, NJ: Erlbaum. Mehler. J. (1990). Language at the initial state. In A.M. Galaburda (Ed.), From neurons to reading. Cambridge, MA: Bradford Books/MIT Press. Mehler, J.. & Fox, R. (Eds.) (1985). Neonafe cognition: Beyond the blooming buzzing confusion. Hillsdale, NJ: Erlbaum. Mehler, J. et al. (1987). An investigation of young infants perceptual representation of speech sounds. Comptes Rendus de IAcademie des Sciences, Paris, 303 III (IS), 637-640. Millikan, R.G. (1984). Language, thought, and other biological categories. Cambridge, MA: Bradford Books/MIT Press. Millikan, R.G. (1986). Thoughts without laws: Cognitive science with content. Philosophical Review, 95, 47-80. Petitto, L.A. (1987). On the autonomy of language and gesture: Evidence from the acquisition of personal pronouns in American sign language. Cognition, 27, I-52. Petitto, L.A. (1988). Language in the pre-linguistic child. In F.S. Kessel (Eds.), The Development of language and language researchers: Essays in honor of Roger Brown. Hillsdale, NJ: Erlbaum. Petitto, L.A., & Marentette, P.F. (1991) Babbling in the manual mode: Evidence for the ontogeny of language. Science, 251, 1493-1496. Piaget, J. (1974). Biology and knowledge, Chicago: Chicago University Press (translation of 1967, Biologie et Connaissance, Paris: Gallimard). Piatelli-Palmarini, M. (Ed.) (1980). Language and learning: The debate between Jean Piaget and Noam Chomsky. Cambridge, MA: Harvard University Press. Piattelli-Palmarini, M. (1986). The rise of selective theories: A case study and some lessons from immunology. In W. Demopoulos & A. Marras (Eds.), Language Learning and Concept Acquisition: Foundational Issues. Norwood, NJ: Ablex. Piattelli-Palmarini, M. (1989). Evolution, selection and cognition: From learning to parameter setting in biology and in the study of language. Cognition, 31, 1-44. Piattelli-Palmarini. M. (1YYOa). Selection semantique et selection naturelle: Le role causal du lexique. Revue de Synthese, IV (1 - 2), 57-94. Piattelli-Palmarini, M. (1990b). Which came first, the egg-problem or the hen-solution? Commentary of (Schull 1990). Behavioral and Brain Sciences, 13, 84-86. Piattelli-Palmarini, M. (1990~). An ideological battle over modals and quantifiers: Commentary of (Pinker & Bloom 1990). Behavioral and Brain Sciences. 13. 752-754. Piattclli-Palmarini. M. (1991). Probability blindness: Neither rational nor capricious. Bostonia. March/April, 28-35. Piattelli-Palmarini, M. (1993). Lillusione di sapere: Che cosa si nasconde dierro i nostri errori, Milan: Mondadori [English translation in press for Wiley, New York]. Pinker, S. (1979). Formal models of language learning. Cognition, 7. 217-283. Pinker, S. (1984). Language learnability and language development. Cambridge, MA: Harvard University Press. Pinker, S. (19811). Learnability and cognition: The acquisition of argumenr structure. Cambridge, MA: MIT Press. Pinker, S.. & Bloom, P. (1990). Natural language and natural selection. Behavioral and Brain Sciences, 13, 707-784. Pinker. S.. & Mehler, J. (Eds.) (1988). Connecfions and symbols. Cambridge, MA: Bradford Books/MIT Press. Premack. D. (1986). Gavagai, or the future history of rhe animal language conboversy. Cambridge, MA: Bradford Books/MIT Press. Pustejovsky, J. (1988). The geometry of events. In C. Tenny (Ed.), Srudies in Generative Approaches to Aspect. Lexicon Project Working Papers, no. 24. Cambridge. MA: MIT Center for Cognitive Science. Rizzi, L. (1990). Relativized minimaliry. Cambridge, MA: MIT Press. Roeper, T.. & Williams. E. (Eds.) (1987). Parameter setting. Dordrecht: Reidel.

346

M. Piattelli-Palmarini I Cognition -50 (1994) 315-346

H.L.. Bever. T.G.. & Terrace, H.S. (Eds.) (1984). Animal cognition. Hillsdale, NJ: Erlbaum. Schull. J. (1990). Are species intelligent? Behavioral and Brain Sciences. 13, 63-108. Smith. E.E.. & Medin. D.L. (1981). Ccrtegories and concepts. Cambridge MA: Harvard University Press. Spelkc. E.S. (1985). Perception of unity. persistence. and identity: Thoughts on infants conception of objects. In J. Mehler & R. Fox (Eds.). Neonate cognition: Beyond the blooming buzzing confusion Hillsdale, NJ: Erlbaum. Spelkc. E.S. (1988). Where perceiving ends and thinking begins: The apprehension of objects in infancy. In A. Yonas (Ed.), Perceptual development in infancy: Minnesota symposium on child psychology (Vol. 20). Hillsdale. NJ: Erlbaum. Stanley, S.M. (1975). A theory of evolution above the species level. Proceedings of the Nutional Academy of Sciences, USA. 72. 646-650. Tanenhaus. M.K., Garnsey, S.M., & Boland, J. (1989). Combinatory lexical information and language comprehension. In G. Altman (Ed.). Cognifive models of speech processing: Psycholinguistic curd computational perspectives. Cambridge, MA: MIT Press. Tenny. C. (Ed.) (1988). Studies in generutive approaches to aspect. Lexicon Project Working Papers. no. 24. Cambridge. MA: MIT Center for Cognitive Science. van Riemsdijk. H.. & Williams. E. (lY86). Introduction to the theory of grcrmmur. Cambridge. MA: MIT Press. von Foerstcr, H. (1060). On self-organizing systems and their environments. In M. Yovitz & S.E. Cameron (Eds.), Self-organizing systems. Elmsford. NY: Pergamon. Wanner. E.. & Gleitman, L.R. (1982). Language acquisition: The .state of the art. Cambridge. UK: Cambridge University Press. Wexler, K. (1982). A principle theory for language acquisition. In E. Wanner & L.R. Gleitman (Eds.). Languuge ucquisition: The state of the art. Cambridge, UK: Cambridge University Press. Wexler. K. (1987). On the nonconcrete relation between evidence and acquired language. In B. Lust (Eds.), Studies in the acquisition of unaphora. Dordrecht: Reidel. Wexler. K. (1990). On unparsablc input in language acquisition. In L. Frazier & J. de Villiers (Eds.), Languuge processing and language acquisition. Dordrecht: KluweriAcademic Press. Wexler. K.. & Culicover, P. (1980). Formul principles of language acquisition. Cambridge, MA: MIT Press. Wexler, K., & Manzini, M.R. (1087). Parameters and learnability in binding theory. In T. Roeper & E. Williams (Eds.), Parameters setring. Dordrccht: Reidcl. Roitblat,