Could Chomsky be Wrong?


Timothy Mason

Université de Paris 8


Well, could he? If you trawl the net, you will find that the majority of material on language acquisition - whether of a first or a second language - is strongly Nativist and often simply takes it for granted that Chomsky and Fodor have, between them, swept away all possibility of opposition. In the English-speaking world - the French, for example, are far more skeptical - the Universal Grammar or the language module rules supreme.


This page is simply an attempt to redress the balance; you will find a set of links to pages that offer alternatives to what appears to be the reigning paradigm. I add material as I find it, trying to give some indication of the arguments. Although I am neither a linguist nor a neuroscientist, I have tried to exercise some judgement over what to include, but you may find you disagree with a number of my calls. So if anyone has any comments - other than to pooh-pooh the whole idea of questioning Chomsky - I'd be glad to hear them.


             Chomsky does not believe that human language evolved from any previous animal communication system, but sprang into being from nowhere. Pinker disagrees with this, but there are problems for the Nativist view in admitting a Darwinian origin for human speech. One of the foremost critics of the Chomskian model from the evolutionary perspective is Terrence Deacon, whose book 'The Symbolic Species' is an attempt to show how language must have evolved gradually, and how the underlying rules, far from being too complex to be learned without the aid of a special module, are, in fact, child's play. You may wish to read William Calvin's review of the book. James Hurford, reviewing it for the Times Literary Supplement, is critical but ultimately favourable. Deacon himself has a paper on 'meme theory' from the Semiotic Review of Books on the web which does expand on one of his arguments.


             While Calvin, particularly in his conversations with Bickerton, seems sometimes sympathetic to the idea that language is an innate and specific competence, other neurologists are less charitable ; Ralph-Axel Mueller, in a paper entitled 'Innateness, Autonomy, Universality?' makes a point that is often put by brain-specialists - the specialization of certain regions of the brain for language processing is an end-result of development rather than being programed into the genome, and, in general, it cannot be said that any of the higher brain functions are innate.

It may be that the burgeoning interest in the evolution of language will prove to be one of the most crucial proving grounds of Nativism ; see the Paris Conference on Language Evolution web-site, where you will find abstracts of interventions by both Nativists and anti-Nativists (You will also find a link to the most recent Conference which is to be held at Leipzig). The Nativist position on modularity is strongly questioned by one strand in evolutionary thinking, which claims that our ancestors' first symbolic communication system was gestural, and that language is intimately related to kinesis : they make the point that Broca's area is concerned both with linguistic and with bodily sequencing.


For an intriguing piece on evolution and meme theory which takes on Chomsky, see Vaneechoutte and Skoyles 'The memetic origin of language ; Modern humans as musical primates'. They cite Robin Allott, who has written a critique of Pinker's "The Language Instinct", arguing that one cannot construct an evolutionary account of language from within the Chomskyan tradition. Allot has other interesting papers on this page.


             For a Whorfian perspective on the evolution of language see Daniel Moonhawk Alford's pages, hosted by Don Watson. Alford provides a line-by-line critique of Chomsky's rebuttal of Whorf. For another critical view from within anthropology, see Chris Knight's "Noam Chomsky ; Politics or Science?" , who argues that that Chomsky's anarchist politics are inversely related to his linguistics paradigm, which was moulded initially by military and other corporate requirements..


             Quartz and Sejnowski, in their 'Constructivist Manifesto', argue that Chomsky and Pinker have avoided confronting the problems inherent in modeling a 'non-stationary mechanism'. They argue that :


Two themes emerge from finding a structural measure of representational complexity: (1) development is a progressive increase in the structures underlying representational complexity, and (2) this increase depends on interaction with a structured environment to guide development. These form the basis of neural constructivism, the developmental theory we present.


(N.B. Ash Asudeh, of the University of Edinburgh, disagrees that they have refuted Fodor and Chomsky).


A group of British cognitive scientists ran a 'Constructivist Workshop' for a while. An outline of their meetings, with useful bibliographical indications, is still on the web, although they announce a little ruefully that 'the workshop has been temporarily discontinued'. However, the webmaster, Gert Westermann, has several of his own papers on the constructivist approach to language available from his home page.


             Deacon sees language as being basically simple in structure - simple enough to be learned by a two-year-old ; Nativists often claim that it is, in fact, far too complex to be learned without help. One of the critical moments in any argument with a Generative Grammarian is when s/he will draw up a list of 'difficult' sentences and then lean back with a satisfied smile, asking you to show how a child could have worked them out for herself. I always feel a little nonplussed at this, for often I am unable to see why it is that the sentences are supposed to be so obscure ; their difficulty seems to be a product of the linguist's analytical tools, rather than of the sentence-structures themselves. Richard Hudson, who is Professor of Linguistics at University College London, seems to be on to something ; he believes that if you look to the words, the meaning will out. Have a look at The difficulty of (so-called) self-embedded structures - you'll have to scroll down the page and then load it down. Hudson has developed a theory of language structure which he calls 'Word Grammar', and also provides a list of links to sites representing other approaches.


             In a similar vein, Thomas Schoenneman (pdf file) argues that the structural similarities found across languages are of a very simple nature, and that they are best explained as arising from the fact that "all languages attempt to communicate the same sorts of semantic information".


Addressing the same argument, some linguists have been working with computers to see whether they can write programs that will learn rules of grammar without needing the specialist module that the Nativists believe in. Although no-one has yet produced a machine that can learn a fully-fledged language, some specialists do claim to have trained their computers to deduce rules in such a way as to falsify the Chomskian impossibility arguments. For one example, have a look at Gerry Wolff's 'Language Learning as Compression' page. (References to other, similar attempts will be found on other pages to which I shall be pointing).


A model of learning which its authors see as applying to all forms of knowledge - language as much as any other - is Anderson's ACT*. At one time, you could access an on-line tutorial from the page, but this seems to be difficult to reach at the moment.


Martin Redington and Nick Chater are working in a similar way - although they are not specifically interested in falsifying the Nativist view. Geoffrey Sampson, one of the participants in the Paris Conference, working in the same area, is far more dismissive of 'the Language Instinct', and claims to have countered every one of the Nativist arguments in his book 'Educating Eve'. His paper "There is No Language Instinct", given at the Conference, is on-line.

Both Wolff and Hudson see language as consisting of a network or set of networks. One of the most powerful of anti-modularity voices is that of Elizabeth Bates ; several of her works can be loaded down from her home page. She also takes a network view of language - you can look at "On the inseparability of grammar and the lexicon: Evidence from acquisition, aphasia and real-time processing", for example (a PDF file) - and she is particularly good at picking holes in the Nativist arguments advanced on the basis of the characteristics of 'language-savants' such as Williams syndrome children : see On language savants and the structure of the mind, another PDF file. In fact, everything on this site is worth looking at, and a happy life-time could be spent following up the bibliographical leads. You will also want to look at Annette Karmiloff-Smith's "Precis of Beyond modularity: A developmental perspective on cognitive science". (Behavioral and Brain Sciences - an unedited preprint).


Bates and Karmiloff-Smith's colleague (or coconspirator?) Jeff Elman also has a number of papers on-line, including "The emergence of language: A conspiracy theory" (in PDF format). Elman can be described as a 'connectionist'.

Another important source in this tradition is Brian MacWhinney's page, where several of his articles - in PDF format again - are available for down-load. MacWhinney sees linguistics in general, and Chomskyan linguistics in particular, as having neglected competitive models of grammar. They fail to see that "Language structure and processing are completely intertwined: There is no distinction between competence and performance"(Competition and Teachability, 1986).


             (The Stanford Encyclopedia of Philosophy provides a short article on connectionism, with some useful links). Also well worth looking at is David Plaut's page, which has a number of links to his essays, including "Language acquisition in the absence of explicit negative evidence: how important is starting small?", written with Douglas Rohde, in which the authors argue that "under a statistical model of the language environment, Gold s theorem and the possible lack of explicit negative evidence do not implicate innate, linguistic-specific mechanisms". (I found their discussion of Gold particularly useful).


One of the main voices in connectionism is that of William Bechtel ; his article entitled : "What knowledge must be in the head in order to acquire language" (available as pdf file from his Publications page) , in which he argues against Fodor that language cannot be uniquely situated in the mind, is available in PostScript format. Grounds for skepticism concerning connectivist claims are voiced by the philosopher István S. N. Berkeley in his essay "Some Myths of Connectionism".)


             Fodor sees Connectionism as a born-again Behaviourism. This characterization is enthusiastically embraced by Ullin T. Place, in 'Behaviourism as a Standpoint in Linguistics', published in the net review 'Connexions'. In another article, 'The Role of the Hand in the Evolution of Language', elsewhere on the web, Place puts forward a Behaviourist model of the evolution of language, which he sees as developing first through gesture.


             Two further critiques of Chomsky's position from a behaviourist perspective can be downloaded in pdf form from the site of the Review 'Analysis of Verbal Behavior : David Palmer's 'Chomsky's Nativism; a Critical Review' and 'Chomsky's Nativism Reconsidered' can be downloaded from the 'Select Electronic Reprints' page. Palmer sees Chomsky's review of Skinner's 'Verbal Behavior' as having been 'harmful to linguistics'.


             Bates and Karmiloff-Smith do not deny modularity as such - but see it as developing through learning. William O'Grady, at the University of Hawaii, on the other hand, may be described as taking a 'nativist' position, but with no need for such a thing as a grammar module. He writes "the phenomena that lie at the core of traditional work on syntax (the architecture of phrase structure, pronominal co-reference, control, agreement, constraints on extraction, and the like) can be explained by the right theory of sentence processing-a theory that does not make reference to a grammar at all." You can access several of his papers from the site, including 'The Radical Middle' (scroll down and load as pdf).


             Another model of interest is Ezra Van Everbroeck's CLASPnet. This is a web-version of his M.Sc dissertation, which presents, he says "an original model illustrating the relevance of connectionist simulations to theoretical linguistic research."

Another well-known critic of Chomsky's is Roger Schank, a specialist in Artificial Intelligence. In the page indexed here, he is interviewed for John Brockman's 'The Third Culture ; Beyond the Scientific Revolution' (Simon and Schuster, 1995). Chomsky, he says, represents everything that's bad about academics, (at one point he could not talk about what he considers to be Chomsky's charlatanism without getting angry). In particular, he was enraged by Chomsky's deliberate exclusion of meaning. John F. Sowa, in a detailed comparison of Schank's work with that of Richard Montague, looks at how a lexicon might be constructed.


             Chomsky's distinction between competence and performance, and his dismissal of the latter as good evidence for linguists has, it seems, condemned those interested in language to spend the rest of their lives contemplating such phrases as 'John promised Bill to leave' or 'John asked Bill for permission to leave' (both from Chomsky's 'Language and Mind', Harcourt, Brace Jovanovich, 1972). Given the richness of human expression, this is a self-denial of awesome proportions. Is it justified? Robert de Beaugrande argues that it is not - he also directs his argument at Stephen Krashen's attempt to build on Chomsky's work to elaborate a theory of second-language acquisition. In another paper, de Beaugrande holds that Chomsky has lead a 'performative campaign to replace real language with ideal language' which allows him to avoid confronting the social nature of linguistic processes.


Ngoni Chipere of the University of Cambridge, in his paper, 'Real Language Users', suggests that if we look at the way people really use language, instead of exchanging phrases between linguists, we discover that there are wide variations in the degree to which people can handle grammatical complexity (and see Huttenlocher, below). He compared skill in the comprehension of grammatically challenging sentences on the part of educated and non-educated subjects, finding that that highly educated non-native speakers may show a greater ability to understand challenging sentences than less educated native speakers. (It has to be said that the educated speakers were, in fact, linguists - which may account for their being good at the kind of games that linguists play).


             Another linguist whose main work is constructing grammar with the aid of the computer is Remko Scha of the Institute for Logic, Language and Computation of the University of Amsterdam. In "Language theory and language technology; competence and performance", first published in Dutch in 1990, he argues that : "The current generation of language processing systems is based on linguistically motivated competence models of natural languages. The problems encountered with these systems suggest the need for performance-models of language processing, which take into account the statistical properties of actual lanuage use".


             While some construct computer programs to learn languages, others are actually looking at the ways in which real flesh-and-blood children do it. (For an idea of what is going on in this field, see the Standford "Child Language Research Forum" page. You can also go to the 'Language Science Research Group" page, which has a rich set of resources, including several articles and some recordings - with transcripts - of children's speech).


             Although a number of workers in this field - such as Boysson-Bardies, whose excellent 'Comment la langue vient aux enfants' has now been translated into English - take a Nativist stance, not all of them do. One interesting approach is that of Janellen Huttenlocher, whose investigations indicate that mothers' verbal behaviour towards their children is crucial not only to their acquisition of vocabulary, but also to their mastery of grammar - children whose mothers talk to them a lot are better able to decode and encode complex sentences than are those who receive little stimulation ("With respect to syntax", she writes, "the studies show that the mastery of complex syntax forms (recursive devices) is highly related to the proportion of complex speech used by their parents". Chomsky believes that there are no significant differences in the degree to which individual native-speakers acquire competence). Although her work is not web-available, a journalistic summary may be found here - scroll down the page a little to find that part of the article.


             Other researchers into language acquisition - Melissa Bowerman, for example - can be found from the Max Planck Institute for Psycholinguistics home page.


             Michael Gasser and Eliana Colunga, in "Playpen : Towards an Architecture for Modeling the Development of Spatial Cognition", argue that "the study of the acquisition of word meaning requires taking seriously non-linguistic cognition, in particular human vision and the pre-linguistic development of concepts". "Playpen" itself is a connectionist modeling of what they characterize as a radical grounded conception of language acquisition.


             On vocabulary acquisition, Paul Bloom, a psychologist at Yale, has recently published 'How Children Learn the Meanings of Words'. Bloom, who takes neither a nativist nor a connectivist position, argues that "word learning is the product of a set of cognitive and linguistic abilities that include the ability to acquire concepts, an appreciation of syntactic cues to meaning, and a rich understanding of the mental states of other people". An article setting out the main arguments of the book will be found at the BBS Online site.


             Another specialist in child-development (and primatology), Michael Tomasello, is also skeptical of nativist claims ; he says that children are great imitators, compared with our chimpanzee cousins. They model their behaviour on that of the adults around them - they learn - for language as well as for other behaviours. Some indications of Tomasello's thought can be found in an article written with Christophe Boesch, published in Current Anthropology, although this is more concerned with primates in general than the human animal in particular. (My thanks to Tim Dougherty for the reference).


             One beef I have with Chomsky is the way he airily declares out of court any work which is not within his own domain. Working from first principles, he is able to dismiss investigations - such as those of Basil Bernstein - without having taken the trouble to read the papers. To the extent that American linguistics has come under the thrall of Nativism, other ways of working have been marginalised - so that, for example, we hear little of Functionalist Linguistics (or of its French equivalent 'enunciative grammar'). Could it be that the victory of the Nativist perspective is largely a matter of rhetorical power - both Chomsky and Fodor are formidable in argument, as Piaget discovered to his cost - rather than of substance? Konrad Koerner, of the University of Ottowa, sees the Chomskian revolution as being due to factors other than its adequacy, elegance or economy ; he offers a socio-historical outline of the rise of Generative Linguistics. Rather more gently, Keith Allan, of Monash University, argues that while Chomsky's work has value, it is not the only way of going about things.


The most often cited alternative to Chomsky's model is that of Halliday's Systemic Functional approach. The Systemic Meaning Modelling Group at Maquarie University is the primary web resource for this. Then you could see The Systemic Functional Room, maintained by Noboru Yamaguchi, of Tohoku University, Japan, (but beware of some very aggressive graphics) and you can also consult Carol Chapelle's page on Functional linguistics. (Ms. Chapelle asks her students in applied linguistics to compare Chomsky and Halliday).

Some of Chomsky's own students became dissatisfied with certain features of his model, and were instrumental in forging a distinct brand of language study - Cognitive Linguistics. An interview with George Lakoff by John Brockman gives some background into how the break came about.The International Cognitive Linguistics Association has its own site, which is not, at the moment, particularly exciting. An introductory bibliography for the field has been compiled by Dick Hudson.


There are several papers written in this tradition that you can load down from Zouhair Maalej's page - such as "Metaphoric Discourse in the Age of Cognitive Linguistics (with special reference to Tunisian Arabic).


Out of the same stable is Fillmore's Construction Grammar. There is a page dedicated to this on the Berkeley server, put together by Charles Fillmore and Paul Kay. Although it hasn't been updated since 1997, it still offers a couple of chapters of a book they were writing on the grammar, and a number of lectures. A bibliography on Construction grammars is also available. (Richard Hudson compares his own Word Grammar with Fillmore's in a recent document available from his home site).


In a similar vein, a web-page put up from Kazan University, under the aegis of the Russian Association of Artificial Intelligence, offers The Web Journal of Formal, Computational and Cognitive Linguistics. Several interesting papers are available for down-loading - including Two Paradigms of Linguistics : The Semiotic Versus Non-Semiotic Paradigm by Sebastian Shaumyan (I can't load the original page any more - this link takes you to the paper via another site - load down as pdf)). Shaumyan argues that Chomsky has ignored the relationship between sound and meaning, and advances what he calls the semiotic paradigm, continuing the tradition in linguistics as rooted in the work of Saussure. Shaumyan also posted a short statement of his case against Chomsky's psychologism to the Linguist list.


Other dissatisfied Chomskans can be found at the HPSG Server at Stanford. This has a good set of links, and interviews with Ivan Sag, Bob Carpenter, Dan Flickinger and Hans Uszkoreit (Note - this has been shifted from the University of Ohio server, losing some items in the process, it seems).


There is also a 'Beyond Chomsky' Website run by Bernard Paul Sypniewski out of Rowan University, hosting a number of papers by linguists such as Victor Yngve ('Human Linguistics - the Hard Science Alternative'). Its 'statement of purpose', written by Bruce Richman opens :"The main obstacle that we have today to clearly understanding the nature and origin of language is the overly formalistic, anti-empirical, anti-historical influence of Chomsky's paradigm for doing linguistics." Bernard Sypniewski encourages contributions. (I can't seem to access this site using Mozilla on OSX. Bruce Richman tells me that there is a site menu on the left hand side, but I can't see it. I can't see it using Opera on Linux either. Perhaps it's just for Windows people).

For an early call to go "Beyond Chomsky", see Leon James' "Prolegomena to a Theory of Communicative Competence", written in 1969, and which takes on the task of evaluating the Chomsky revolution 'ten years after'.


On a rather different tack, James Cooke Brown, who has spent many years teaching people to speak the logical language Loglan, argues that it is his experience that the extension and change in world-view and thinking-patterns his students report is good evidence for the Sapir/Whorf hypothesis - the latter is, of course, decried by Nativists, such as Pinker, who see in it a denial of the unicity of language. (There seem to be at least two languages going by the name of Loglan, one of which is a programming language developed by a Polish group of computer specialists in the late 70s (now supported by the University of Pau), while the other, developed by Cooke Brown, was, it seems, specially devised to test the Sapir/Whorf hypothesis. A breakaway version called Lobjan, also exists, and may even have more support than Brown's original version. (This page is sometimes difficult to access, but does seem to be still active).


Finally, I will mention a page by Alex Gross, who is a specialist in translation. This has lead him to question both the basic theory of the Nativists, and their use of languages which they have no real knowledge of, such as Chinese, to make their points. You will find other leads from there.


Pages en français


Le philosophe Paul Ghils (Haute Ecole de Bruxelles) voit Chomsky comme représentant d'une pensée linguistique prisonnière de la logique formelle, et suggère qu'une voie davantage 'héraclitienne' serait plus fructueuse.

L'université de Laval nous fournit un site sur l'oeuvre du linguiste Gustave Guillaume. Le site s'est enrichi récemment d'articles divers.


Ailleurs que, plutôt que contre, Chomsky - comme, à vrai dire, la majorité des linguistes francophones - Gilles Bernard, linguiste et informaticien à Paris VIII, met à notre disposition un grand nombre de ses articles dans lesquels la linguistique énonciative rencontre l'intelligence artificielle. (Depuis sa page personnelle, on peut facilement atteindre la page du Groupe CSAR, (Catégorisation Sémantique Automatique par Réseau neuromimétique).


Pierre Bourdieu s'est souvent montré critique à l'égard de la linguistique Chomskyenne - sans vraiment élaborer là-dessus. Pour ma part, je le soupçonne de ne pas avoir pris le temps d'entrer suffisamment dans le champs pour formuler une opposition intéressante, et j'ai l'impression qu'il ne saisit pas très bien ce que Chomsky veut dire. J'inclus néanmoins un lien vers son Intervention au Congrès de l'AFEF, (Limoges, 30 octobre 1977), où il expose certaines des idées qui nourrissent son livre 'Ce que parler veut dire'.



© 2011 by Timothy Mason