Wittgensteinians and Chomskyans: Mentalism in Linguistics

An extended defence of Chomskyan linguistics and the Cognitive Paradigm in general against the Wittgensteinian critiques of Saul Kripke, Baker & Hacker, Esa Itkonen and others. Consideration is given to the Private Language arguments; to the distinction between public (outer) and private (inner); to the distinction between the social and the individual; to ascription and causality; to rule-normativity; to creativity arguments. Comments from Noam Chomsky and Peter Hacker are included in the extensive Endnotes. Five lines of Wittgensteinian critique are identified in each of the essay's sub-section headings

Baker, G (1985) Review of Kenny (1984). Times Literary Supplement., 5 July, p 754

Baker, G and Hacker, P (1984a) Language, Sense, and Nonsense.. Oxford: Basil Blackwell

Baker, G and Hacker, P (1984b) Scepticism, Rules and Language . Oxford: Basil Blackwell

Bhaskar, R (1975) A Realist Theory of Science. Leeds: Books (2nd edn Brighton: Harvester Press 1978)

Bickerton, D (1977) 'Some Problems of Acceptability and Grammaticality in Pidgins and Creoles' In S Greenbaum, ed, Acceptability in Language. pp 27 - 37 The Hague: Mouton

Bickerton, D (1981) Roots of Language.. Ann Arbor, Michigan: Karoma

Bickerton, D (1984) 'The Language Bioprogram Hypothesis' The Behavioral and Brain Sciences., 7, pp 173 - 221

Burge, T (1978) 'Individualism and the Mental' in P French et al, eds, Midwest Studies in Philosophy ,. IV, pp 173 - 221 Minneapolis, Minn: University of Minnesota Press

Caplan, D et al (1984) Biological Perspectives on Language.. Cambridge, Mass: MIT Press

Cavell, S (1958) 'Must We Mean What We Say?'Inquiry., 1, pp 172 - 212

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

Chomsky, N (1975a) Reflections on Language. London: Fontana

Chomsky, N (1975b) The Logical Structure of Linguistic Theory.. New York: Plenum

Chomsky, N (1980) Rules and Representations.. New York: Columbia University Press.

Chomsky, N (1982b) 'A Note on the Creative Aspect of Language Use' Philosophical Review., 91, pp 423 - 434

Chomsky, N (1986) Knowledge of Language: Its Nature, Origin and Use. New York: Praeger

Chomsky, N (2000) New Horizons in the Study of Language and Mind. Cambridge: Cambridge University Press

Coulter, J (1983) The Social Construction of Mind. London: Macmillan

Dennett, D (1979) Brainstorms. Hassocks: Harvester Press

Dennett, D (1981) 'True Believers: The Intentional Strategy and Why it Works' in A F Heath, ed, Scientific Explanations., pp 53 - 75 Oxford University Press

Dennett ,D ( 1982) 'Beyond Belief' in AWoodfield, ed, Thought and Object, pp 1 - 95. Oxford: Clarendon Press

Dennett, D (1982-83) 'Styles of Mental Representation'.Proceedings of the Aristotelian Society., pp 213 - 26

Dennett, D (1983) 'Intentional Systems in Cognitive Ethology' The Behavioral and Brain Sciences, . 6, pp 343 - 90

Dreyfus, H (1979) What Computers Can't Do. . 2nd edn. New York: Harper and Row

Evans, G ( 1978) 'Semantic Theory and Tacit Knowledge' in S Holtzman and C Leich, eds,Wittgenstein: To Follow a Rule . pp 118 - 37 London: Routledge and Kegan Paul

Feldman, H et al (1978) 'Beyond Herodotus: The Creation of Language by Linguistically Deprived Deaf Children' in A Lock, ed, Action, Gesture and Symbol: The Emergence of Language. pp 351 - 414 London: Academic Press

Field, H (1978) 'Mental Representation', Erkenntnis., 13, pp 9 - 61

Fodor, J (1976) The Language of Thought.. Hassocks: Harvester Press

Fodor, J ( 1981) Representations. Brighton: Harvester Press

Fodor, J (1983) The Modularity of Mind. Cambridge, Mass MIT Press

Forbes, G (1983 -84) 'Scepticism and Semantic Knowledge' Proceedings of the Aristotelian Society., pp 223 - 37

Goldin-Meadow, S and Mylander, C (1983) 'Gestrual Communication in Deaf Children: Noneffect of Parental Input on Language Development', Science., 221, pp 372 - 74

Hare, R ( 1960) 'Philosophical Discoveries' Mind., 69

Hockett, C (1968) The State of the Art.. The Hague: Mouton

Itkonen, E (1978a) Grammatical Theory and Metascience . Amsterdam: John Benjamins

Itkonen, E (1981b) 'The Concept of Linguistic Intuition' in F Colmas, ed, A Festschrift for Native Speaker. pp 127 - 40 The Hague: Mouton

Itkonen, E (1984) 'Concerning the Ontological Question in Linguistics', Language and Communication., 4, pp 241 - 46

Jenkins, L (1979) 'The Genetics of Language' Linguistics and Philosophy., 3, pp 105 - 19

Katz, J (1981) Language and Other Abstract Objects. Oxford: Basil Blackwell

Kenny, A (1973a) 'The Origin of the Soul' in A Kenny et al The Development of Mind. Edinburgh: Edinburgh University Press

Kenny, A (1984) The Legacy of Wittgenstein. . Oxford: Basil Blackwell

Kripke, S (1982) Wittgenstein on Rules and Private Language. Oxford: Basil Blackwell

Lenneberg, E (1967) Biological Foundations of Language.. New York: Wiley

Lyons, J (1981) 'Language and Speech' in H Longuet-Higgins and and J Lyons, eds, The Psychological Mechanisms of Language., pp 1 - 8 London: Royal Society

McDowell, J (1984) 'Wittgenstein on Following a Rule'Synthese., 58, pp 325 - 63

McGinn, C (1984) Wittgenstein on Meaning.. Oxford: Basil Blackwell

Malson, L (1971) Wolf Children.. London: New Left Books

Marr, D (1982) Vision.. San Francisco: W H Freeman

Matthews, P (1979) Generative Grammar and Linguistic Competence.. London: George Allen and Unwin

Mey, M de (1982) The Cognitive Paradigm.. Dordrecht: D Reidel

Muhlhausler, P (1985) 'Patterns of Contact, Mixture, Creation and Nativization: Their Contribution to a General Theory of Language' in C-J Bailey and R Harris, eds, Developmental Mechanisms of Language., pp 51 - 87. Oxford: Pergamon Press

Muhlhausler, P (1986) Pidgin and Creole Linguistics.. Oxford: Basil Blackwell

Newport, E et al (1977) 'Mother, I'd rather do it myself: Some effects and non-effects of matgernal speech style' in C Snow and C Ferguson, eds, Talking to Children., pp 109 - 149. Cambridge: Cambridge University Press

Pateman, T (1982b) 'What I tell My Students about Noam Chomsky and Seymour Papert' www.selectedworks.co.uk

Pateman, T (1987) Language in Mind and Language in Society.. Oxford: Oxford University Press

Pulman, S (1983) Word Meaning and Belief.. London: Croom Helm

Pylyshyn, Z (1984) Computation and Cognition . Cambridge, Mass: MIT Press

Quine, W (1960) Word and Object. . Cambridge, Mass: MIT Press

Quine, W (1972) 'Methodological Reflections on Current Linguistic Theory' in D Davidson and G Harman, eds,Semantics of Natural Language., pp 442 - 54 Dordrecht: D Reidel

Rosen, M (1983) 'Critical Theory: Between Ideology and Philosophy' in S Mitchell and M Rosen, eds, The Need for Interpretation., pp 90 - 117 London: Athlone

Sampson, G (1979) Liberty and Languag. e Oxford: Oxford University Press

Sampson, G (1980a) Making Sense. Oxford: Oxford University Press

Schatzki, T (1983) 'The Prescription is Description: Wittgnestein's View of the Human Sciences' in S Mitchell and M Rosen, eds, The Need for Interpretation., pp 118 - 40. London: Athlone

Searle, J (1984) Minds, Brains and Science.. London: British Broadcasting Corporation

Stabler, E (1983) 'How are Grammars represented ?' inBehavioral and Brain Sciences., 6, pp 391 - 421

Steinberg, D (1975) 'Chomsky: From Formalism to Mentalism and Psychological Invalidity', Glossa., 9, pp 218 -52

Stich, S (1983) Beyond Belief: From Folk Psychology to Cognitive Science.. Cambridge, Mass: MIT Press

Strawson, P (1985) Skepticism and Naturalism: Some Varieties.. London: Methuen

Walker, E ed (1978) Explorations in the Biology of Language.. Hassocks: Harvester Press

Winston, M (1982) Explanation in Linguistics: A Critique of Generative Grammar.. Bloomington, Indiana: Indiana University Linguistics Club

Wittgenstein, L (1958) Philosophical Investigations.. Oxford: Basil Blackwell

Wright, C (1980) Wittgenstein on the Foundations of Mathematics.. London: Duckworth

Young, R (1976) Seriation by Children: An Artificial Intelligence Analysis of a Piagetian Task. Basel: Birkhauser

Website version 2004. Lightly revised from Chapter 6 of my book, Language in Mind and Language in Society (Oxford University Press 1987), pp. 120 - 146. Copyright 1987 with me as author.

A few hardbound copies of the book from which this essay is extracted are still available, directly from the author. Price £45 post free. UK, US French and German personal checks acceptable. Calculate dollar or euro equivalent using the mid-point rate (i.e., add Buy and Sell rates and divide by two).

Mailing address:

Unit 10,
91 Western Road,
Brighton,
BN1 2NW,
United Kingdom

1. Introduction

Ludwig Wittgenstein died in 1951, 6 years before the publication of the book which inaugurated the Chomskyan revolution in linguistics, Syntactic Structures(Chomsky 1957), and a dozen years before the cognitive paradigm (to use Mey 1982's expression) became clearly visible as the successor to behaviourism in both linguistics and psychology. Nonetheless, it is probably Wittgenstein's posthumously published writings which have in one way or another inspired the majority of critiques of Chomskyan linguistics in particular and the cognitive paradigm -cognitivism - in general. Although this is some evidence that Wittgenstein himself would have regarded cognitivism as a bewitchment of the intelligence, it is by no means certain. Wittgenstein's style of thinking and writing do not make it transparent what he thought, and of the posthumously published writings only the Philosophical Investigations(Wittgenstein 1958, first published 1953) were in a form which Wittgenstein himself contemplated publishing. There has been continuous argument over what Wittgenstein meant and what its implications are: in the recent literature at least one writer (McGinn 1984) ends up with an understanding of Wittgenstein's later work which makes it reasonably if not entirely congenial to cognitivism. In contrast, Baker and Hacker 1984a draw confidence from Wittgenstein's writings for an implacable hostility to the cognitive paradigm and all its works, Chomsky's in particular. And they could certainly claim that in terms of a spectrum of views, many more Wittgensteinians are at their end than McGinn's. Baker and Hacker claim that Wittgenstein would find `cognitivism' utterly misconceived, no doubt about it.

This conflict of interpretations is one reason why I refer to Wittgensteinians rather than Wittgenstein in my title. I do not wish to be involved in exegetical questions, but to concentrate on arguments brought against the cognitive paradigm, and specifically Chomskyan linguistics.

A second reason is that it strikes me that some of the Wittgensteinian writing is less indebted to Wittgenstein than its authors think, and that many of the arguments deployed - particularly those having to do with doubts about the existence and knowability of unobservables - arise from a wider empiricist, nominalist, and sceptical ambience. This seems to me the case with Kripke's influential Wittgenstein on Rules and Private Language(Kripke 1982) which reads Wittgenstein through spectacles provided by Hume, Quine, and Goodman. (See Endnote 1)

But if this explains `Wittgensteinians' rather than `Wittgenstein', in my title, it does not explain `Chomskyans' rather than `Chomsky'. Chomsky is alive and publishing; he does not write hermetically; he has no hesitations in interpreting his own past writings; and he has addressed some of the Wittgensteinian writing directly, for example Kripke 1982 (Chomsky 1986, ch. 4. See now also Chomsky 2000).

But there are drawbacks in pitting Chomsky directly against the Wittgensteinians. First, it would then appear that exegetical problems around Chomsky's writings do not exist. Yet there is, in fact, lively controversy, in which Chomsky has participated, over the relationship of his work to Cartesianism, Quine, Goodman, and computational linguistics. A long bibliography of such discussions and controversies can be constructed; I provide a few relevant references but do not wish to be drawn in here . (See Endnote 2).

Secondly, Chomsky's work now finds its place within the cognitive paradigm which has been built up, mainly in the USA, by researchers in linguistics, psychology, artificial intelligence, and philosophy over several decades. Those researchers have also been concerned with the philosophical foundations of cognitive science and with the critiques directed at them. So both because I do not wish to be involved in Chomsky exegesis and because I wish to consider a number of areas of dispute at a level of generality which abstracts from many local nuances, I have put `Chomskyans' rather than `Chomsky' in my title, nonetheless acknowledging thereby Chomsky's pre-eminence among workers in the cognitive paradigm. And while I succeed in not quoting from Wittgenstein himself, I find it impossible not to quote from Chomsky's writings.

I have endeavoured in the remainder of this essay to parcel out Wittgensteinian critiques under a number of headings representing some very general accusations levelled at cognitivism. A range of different positions and arguments from different authors is thus considered. In each section I consider materials which can be parcelled up into a general claim about one aspect of Chomskyan thought. The criterion for selecting the items for discussion out of a vast mass of available material is very much on the basis that I should have something to say in response. The chapter is by no means expository of either Wittgensteinian or Chomskyan positions and some general familiarity with the area is presupposed.

2. `Chomskyans treat something essentially social as if it were essentially individual'

In Chomskyan linguistics, language is studied as a property of the individual human mind, both as the innate universal grammar, which makes the development of language possible in the individual, and as an attained mentally represented grammar - what Chomsky now calls an `I - language', (an internalized language as opposed to an externalized language, an `E - language) (Chomsky 1986, ch.2). This study is individualistic in the sense that properties are ascribed to individuals independently of any reference to other individuals; reference to other individuals is unnecessary either to the ontology of language in the individual or the epistemology of coming to know what linguistic properties an individual instantiates. It is not denied that social interaction is a necessary condition of the triggering of linguistic development in the child, but it is denied that reference to the language of other individuals is necessary in characterizing that linguistic development; it is also denied that specifically linguistic (as opposed to social) interaction is necessary to linguistic development (Feldman, Goldin - Meadow, and Gleitman 1978). This distinction, between social and linguistic interaction, is often ignored; for example, by those (e.g. Malson 1971) who think that the cases of wolf children are relevant to the issues now under consideration. But their relevance is only to the claim which has already been granted, that social interaction is necessary to linguistic development.

The deepest philosophical challenge to the view that language is an individual property and can be ascribed individualistically comes from a deployment of Wittgenstein's arguments against the possibility of a private language, to develop a `community view' of language. This is what Itkonen 1978a and Kripke 1982 do. I shall consider some aspects of Kripke's arguments shortly. But it does seem to me that a community view, even if not Kripke's, seems plausible to philosophers because they start out with an empirically falsifiable, and false, picture of language generalized from their local (usually Oxford) experience. I think it is worth pausing to correct this false picture.

It seems obvious to many philosophers that language is `essentially a social phenomenon'; such is Wittgenstein's view, according to Strawson (1985, p. 76).(See Endnote 3). For Anthony Kenny it is a `truism that language is a social, conventional and rule - governed activity' (Kenny I973a, p. 93) and that `However far we go with Chomsky in postulating innate mechanisms we cannot deny that what children acquire, when they learn language, is the language of their parents: it is a language spoken in their environment' (p. 97). For Baker and Hacker, `A language is a normative practice, a practice of using signs according to rules. It is also a social practice. It exists in the activities of language users in a community' (Baker and Hacker 1984a, p. 284) and, again, `The normativity of the language the child is acquiring lies with the adult who trains (and the linguistic community to which he belongs), not with the child who learns' (p. 256). It would be easy to go on assembling quotations which have in common the family resemblance of assuming that (1) linguistic development is always necessarily targetted on the pre - existing linguistic practices of a community; (2) linguistic development always necessarily consists in acquisition of such a target language. Neither claim has any philosophical necessity or empirical universality about it . Both claims are empirical, not "philosophical", and they are false.

There are at least two clear cases where linguistic development is not targetted, and consequently does not involve the acquisition of a target language. First, in creolization the children of pidgin speakers have to expand the antecedent pidgin in order to make it communicatively adequate. The adult community, far from training and being the source of normativity, is left behind in this development, which has now been studied quite extensively from a number of theoretical perspectives (Bickerton 1981 and 1984 provides a strong nativist account. For extensive empirical evidence, see Muhlhausler 1986).

Secondly, and more radically, the deaf children of hearing parents develop idiosyncratic signing systems of their own, which are clearly linguistic though in a different medium from the normal one. (See Endnote 4). These signs are used consistently and rationally in pursuit of the child's goals, but though they are passively understood by the child's parents, they were not learnt from them nor, in general, are they subsequently produced by the parents (Feldman, Goldin - Meadow, and Gleitman 1978; Goldin - Meadow and Mylander 1983). These signs are public, checkable for consistent use and potentially shareable; but they are not social, if by that is meant that they pre - exist the child's use of them, or are used by convention, or are shared with others. There are other cases of linguistic systems which are the properties of single individuals, for example, jargons which represent individual attempts at cross - linguistic communication (Muhlhausler 1985, p. 62), although these cut less philosophical ice in so far as they are properties of individuals who already have another language.

In relation to creolization and `home sign' systems, claim (2) above falls because claim (1) falls: a target language is not acquired because there is no language or only an inadequate language for development to be targetted upon. (Endnote 5). It is also worth considering whether claim (2) could be true while claim (1) is false: that there is a target language which ends up being spoken, without its being the case that linguistic development is (wholly) targetted on it. Thus it is evident that children's speech is not wholly imitative even in familiar kinds of language environment and it is by no means self - evident even in the standard language situation that `language use begins as imitative, reactive and habitual behaviour ... the roots of the mastery of language lie in training' (Baker and Hacker 1984a, p. 256): Chomsky says this is just plain false (personal communication 1985), and I agree. It is, for example, perfectly possible that children's reception of and attention to linguistic input is highly selective. In terms of Bickerton's approach, children surrounded by an adequate target language could nonetheless degrade the linguistic input they receive to a pidgin and then, using innate cognitive or grammatical resources, creolize it. Only later on do they try to bring their creolized output ("child language") more closely into line with the language being used by adults around them (Bickerton 1981, pp. 134 - 5 and 1984; cf. Newport, Gleitman, and Gleitman 1977, pp. 134 - 5). As Bickerton puts it, `The role of the bioprogram for children acquiring a "ready - made" language (rather than creating a creole) is to furnish elementary forms and structures from which (guided by input from the target language) they can develop other and more complex forms and structures' (1984, p. 185). And referring to Dan Slobin's work on basic child grammars, Bickerton observes, `The significance of Slobin's findings is that they show . . . structures consistently used by young children that violate the grammatical rules of their target languages but are consistent both with the rules hypothesised here for the bioprogram and with surface forms found in creole languages' (p. 185). In sum, even in the normal language learning situation, it is not true that for purposes of early language acquisition, children are built the way Jesuits, behaviourists, and Baker & Hacker have imagined them: trainable imitating devices. They ain't. (Endnote 6)

To return now to my initial counter - examples to the philosophers' paradigm, I suppose you cannot really blame philosophers for not imagining the cases of deaf and creolizing children, since reality is always more imaginative than philosophical fictions. In the present instance, these reduce to the overworked character of Robinson Crusoe. For the past half century, Crusoe has played a bit part in the debates around Wittgenstein's private language arguments, notably in Kripke 1982. The trouble is he is too weak a figure to do anything against the over - socialized picture of language opposed in the previous paragraphs, which is why I paused to introduce the real corrective provided by deaf and creolizing children. But I have still to consider whether these children or Robinson Crusoe are in any way relevant to or decisive against deeper philosophical arguments in favour of the community view of language, such as are advanced in Itkonen 1978a and Kripke 1982.

Itkonen is clear that the deaf and creolizing children cut no ice. Replying to my use of these examples in Pateman 1983 against Itkonen 1978a, he reiterates his view that `Because rules (qua constituents of a language) exist at the level of, or as objects of, common knowledge, they are social in character. I use Wittgenstein's private - language argument to prove that this is a necessary, and not just a contingent, property of rules' (Itkonen 1984, p. 241).(Endnote 7). Itkonen claims that there can only intelligibly be said to be a rule where there is the possibility of mistake and of the correction of mistakes in applying the rule and that self - correction is necessarily parasitic on the possibility of correction by others: my own memory of what rule I am following is inadequate as a basis for self - correction (p. 242). It follows that for him the deaf child can only be said to be following a rule if others (e.g. its parents) can correct its application of its idiosyncratic signs (p. 243). He thinks this can happen and that therefore `home sign' systems are not counter - examples to his claim about the necessarily social character of rules.

I find this response revealing on four counts, all of which as we shall see, bear on Kripke 1982.

First, it indicates just how thin the ice is on which Wittgensteinians skate. Itkonen has to claim that the deaf child could not alone correct a misapplication of a sign; its own memory of what rule it is following (how it means or intends to apply the sign) cannot be adequate. It cannot correct itself if it calls a dog a cat. At this point, my intuitions are that this is not remotely plausible. Though one may have no local motive to ascribe a rule rather than a regularity until self - corrective behaviour is observed, within a broader context of beliefs about the nature of mental operations one would want to see the child's behaviour as instancing rule - following. (Endnote 8).

Secondly, and more importantly, if scepticism about memory disbars the child from using its memory, it ought equally to disbar the parents from using their memories about how the child previously used a sign. From general scepticism about memory, there is no safety in numbers (cf. McGinn 1984, Pp. 188 - 89). The obvious move is not to start the sceptical argument off; once started it cannot be stopped.

Thirdly, and even more importantly, Itkonen's argument only shows that if the deaf child is following a rule in its public behaviour, then what rule it is following is recoverable (knowable), checkable, and corrigible by others. But this is not in dispute, except in the sense discussed in the next paragraph. What is in dispute is whether the child can follow a rule which no one else is following. i.e. whether rules are or are not necessarily social, `objects of common knowledge' in Itkonen's terms. My point is that the deaf child follows a rule (Itkonen grants that) which no one else is following (which he cannot consistently grant). (Endnote 9). The rule is potentially social and potentially the object of common knowledge, but not actually so.

Fourthly, Itkonen differs from Kripke in considering it unproblematic that if the deaf child is following a rule then it will be the sort of rule which we can recover, or at least we simply assume that the child will not be following rules which have a Goodmanesque `grue' - like or Kripkean `quus' - like character. Though the child is not following our rules, it is assumed to be following a rule of the same general character as those we habitually follow, hence recoverable by reference to our usual ideas about rationality, the nature of concepts, etc. This seems to me a straightforwardly naturalist move with much to recommend it. The child is assumed to point to the same kind of middle - sized objects, to form the same concepts, as we do just because it is con - specific with us, sharing a species - specific endowment. This naturalism, which can be read in both Chomsky and Wittgenstein, and in Kripke 1972, is one response to the scepticism of Kripke 1982 (see also Pulman 1983 on nativist semantics (Endnote 10) and Strawson 1985 on naturalism and scepticism).

For Kripke 1982 proposes - or appears to propose - not a naturalistic (straight) but a sceptical solution to the sceptical paradox he sets up as at the heart of Wittgenstein's thinking on rulefollowing and private language. This paradox is that there is no fact about me that constitutes my meaning one thing rather than another by a word (Kripke 1982, p. 21), hence no rule that I can be said to be following, hence, apparently, no such thing as meaning anything by any word (p. 55). The sceptical solution `does not allow us to speak of a single individual, considered by himself and in isolation, as ever meaning anything' (p. 68) but proposes that `Wittgenstein finds a useful role in our lives for a "language game" that licences, under certain conditions, assertions that someone "means such - and - such" and that his present application of a word "accords" with what he "meant" in the past. It turns out this role, and these conditions, involve reference to a community. They are inapplicable to a single person considered in isolation' (p. 79). What sort of reference to a community is involved? `We say of someone else that he follows a certain rule when his responses agree with our own and deny it when they do not' (p. 91); `The solution turns on the idea that each person who claims to be following a rule can be checked by others. Others in the community can check whether the putative rule follower is or is not giving particular responses that they endorse, that agree with their own' (p.101) and `If there was no general agreement in the community responses, the game of attributing concepts to individuals could not exist' (p. 96). Kripke thinks that his interpretation of Wittgenstein shows that `Modern transformational linguistics, inasmuch as it explains all my specific utterances by my "grasp" of syntactic and semantic rules ... seems to give an explanation of the type Wittgenstein would not permit' (p. 97, n. 77).(Endnote 11) It is an account of the language user in isolation.

This is the barest summary of Kripke's line of argument; further aspects of Kripke's book are considered in subsequent sections. In this section, I am concerned with it in so far as it opposes a social account of the ascription of meaning to an individualist account of its ontology. My main contention is this. As a result of a verificationist, `warranted assertibility' approach which subordinates ontology to epistemology, Kripke is led to conflate several distinct questions:

  • (A) What is it for S to follow a rule X?
  • (B) What is it for S to follow our rule Y?
  • (C) How do we tell when S is following rule X?
  • (D) How do we tell when S is following our rule Y?

In addition, he has a naturalist answer to a deeper question (E) which is a straight and not at all a sceptical solution to his sceptical paradox:

  • (E) How is it possible to see others as following any rules at all?

In outline, Kripke slides from saying that question (A) is unanswerable, because we can have sceptical doubts about what rule S is following, to saying that there is nothing which would make one answer to question (A) rather than another true; there is no fact of the matter. Thus while insisting that the sceptical problem is not merely epistemic -`The sceptic argues that there is no fact as to what I meant' (p. 38) - his arguments for this conclusion are merely epistemic. Thus, his argument against a mechanical model of language or a functionalist psychology is: `Any concrete object can be viewed as an imperfect realization of many machine programs. Taking a human organism as a concrete object, what is to tell us which program he should be regarded as instantiating? In particular, does he compute "plus" or "quus" [Kripke's illustrative example]' (pp. 367, n. 24). But this kind of epistemological scepticism is simply an objection to all theory, not just cognitive theory. More decisively, any argument from the unknowability of something to its non-existence is simply fallacious. Not least it supposes that there is a way of determining what is and what is not knowable- in- principle ( which some will recognise as an old chestnut problem with old-fashioned verificationism).

In giving his sceptical solution Kripke sometimes suggests that we can only ascribe rule - following to S if we can ascribe a rule which we ourselves follow. I have already argued that this is false. In Kripke the idea is encouraged by his treatment of language in terms of arithmetic, where there clearly is one right answer which provides a target for the learner. But even in arithmetic there is more scope for idiosyncratic rule - following than Kripke allows. Thus, Kripke writes of Smith, as judge of Jones' s rule - following, that unless he can bring Jones's responses in line with his own he `will be inclined to judge that Jones is not really following any rule at all' (p. 91) and that `A deviant individual whose responses do not accord with those of the community in enough cases will not be judged, by the community, to be following its rule; he may even be judged to be a madman, following no coherent rule at all' (p. 93). What Kripke elides in these equivocating quotations (why `inclined' not `obliged', `may even' not `must be'?) is the obvious possibility that the individual is simply following a different rule, as children clearly do when learning arithmetic . At this stage, they produce systematic (not incoherent) errors which can, for example, be modelled in production systems (Young 1976) and, more simply, understood by teachers.

Elsewhere, Kripke seems to allow for this possibility, in the same way as Baker and Hacker 1984b (n. 9); this is one way of reading his remarks on Robinson Crusoe, where he writes of bringing Crusoe into our community in the sense of applying `our criteria for rule following to him' (Kripke 1982, p 110). The major point to notice is that Kripke has absolutely nothing to say about what our criteria for rule following are, only about what are the criteria for following our rules . So, for example, Kripke does not discuss whether it is necessary or sufficient for a creature to be following rules that it be engaged in intentional activity, that it engage in self - corrective behaviour, that it use a language of discrete symbols, that it recognize the type - token distinction, and so on. In other words, he does not discuss the concept of rationality - which is treated in quite another body of recent philosophical writing, for the most part cut off from the recent rule - following and private language literature. What he does have to say is that our ascription of rule - following to others ultimately depends on the assumption that others are like us. It is because we agree (distributively) in our responses, that is, share a form of life that the game of attributing concepts to (other) people can be played. `On Wittgenstein's conception, such agreement is essential for our game of ascribing rules and concepts to each other... Wittgenstein stresses the importance of agreement, and of a shared form of life, for his solution to his sceptical problem in the concluding paragraphs of the central section ofPhilosophical Investigations (paragraphs 240 - 42; see also the discussion of agreement on pp. 225 - 27)' (p. 96).

I think this is true and important. But it is not part of a sceptical solution to the sceptical paradox. It is a straight solution which cuts the ground from under the sceptic's feet. We can get to see other people as following other rules from ours just because they are our conspecifics and respond to the world in fundamentally the same way as us. This means that even when they have different concepts from us, the alleged problems of radical translation or radical interpretation do not arise. We do not even encounter `grue' - like or `quus' - like behaviour; these are not learnable, naturally occurring concepts any more than `gavagai' meaning `undetached rabbit part' (cf. Quine1960; Pulman 1983). It is because this is the way things are that it is possible for us to see others as following rules which are not ours, and see what those rules are.

This naturalistic construal of `form of life' and `(distributive) agreement' is highly congenial to the Chomskyan. It is resisted by John McDowell. Writing of the problem of acquisition, how the child grasps the rule intended, McDowell writes, `If we picture an interpretation that would precisely bridge the gap between instruction and competent use, it seems that it can only be one which each person hits on for himself - so that it is at best a fortunate contingency if his interpretation coincides with the one arrived at by someone else' (McDowell 1984, Pp. 332 -33). But why should one want anything more than fortunate contingency? For fortunate contingencies we are fortunate: `I see no reasonable alternative to the position that . . . the basic reason why knowledge of language comes to be shared [distributively - TP] in a suitable idealised population (and partially shared in actual populations) is that its members share a rich initial state, hence develop similar states of knowledge' (Chomsky 1980, pp. 86 - 87).

I conclude that there is no principled case against treating some dimensions of language as properties of the individual; language phenomena are not necessarily social. As McGinn sums up his discussion of the community view, `The proposed community conceptions fall into two groups: those that are reasonably plausible but are not properly described as community conceptions, and those that are genuine community conceptions but are intrinsically implausible' (McGinn 1984, p. 200). The former allow that the `home - signing' deaf child is following rules regardless of what its parents do; the latter, implausibly, deny it. There is simply no obstacle to regarding the `home - signing' deaf child as a rational, consistent, goaldirected symbol - user. What more could be involved in rule - following?

At this point, another line of Wittgensteinian criticism can be discerned behind the sociologist manque enthusiasm for the social:

3 `Chomskyans treat something essentially public (outer) as if it were essentially private (inner)'

We can see the deaf child signing, and the rules we attribute (`By the sign . . . the child means a dog') are attributed on the basis of public, checkable, shareable behaviour. But - it is said - Chomskyan linguistics proper is really all about things which are not public: innate universal grammars and mentally represented grammars. The postulation of these private or inner entities is unwarranted and unnecessary: `The study of a language is an investigation into a "social whole" that exists in the practices of a community, not a psychological object that exists between the ears of a speaker' (Baker and Hacker 1984a, p. 273). Well, assume that after my section 2 and consistently with their second book, Baker and Hacker 1984b, Baker and Hacker would settle for `public' rather than `social', there still remains the question of this hostility to studying what goes on between the ears. This section examines aspects of this hostility and suggests some lines of defence.

Wittgensteinians undoubtedly share with behaviourists a suspicion of the inner, the mental. There is even more suspicion of anything mental which is not introspectible, such as the mental representations of cognitive science. Hostility to non - introspectible, tacit mentally represented knowledge is partly motivated by a general empiricist scepticism about theoretical entities and about realist forms of explanation which characteristically explain phenomena in terms of the operation of non - observable or non - observed causal mechanisms.

In philosophy of mind, this might be regarded as the legacy of Gilbert Ryle. But there is a more specifically Wittgensteinian case against tacit knowledge and all the rest, namely,the argument that says behaviour is criterial for the ascription of mental states and that mental states are internally (logically) and not causally related to such behaviour. Sometimes, it is even said that they do not exist independently or autonomously of it. That our ascriptions of mental states, events, etc., must be based on observed behaviour is not just faute de mieux; it is because the outer is not related to the inner as effect to cause but actually gives sense to our ascriptions of the inner. (Pain behaviour is not a causally -related symptom of pain, but criterial for being in pain, etc.) Expressed differently, we simply misconstrue at least some of our mental vocabulary, thinking that words refer to states and processes when really their grammar is quite different. Understanding, to take a paradigm case (much deployed in Coulter 1983), is not a state or process, but an achievement. It is therefore completely misguided to look for or hypothesize states or processes of understanding.

The obvious reply to this last claim is that `to understand' may well be an achievement verb, but that one can legitimately theorize the states and processes which subserve understanding, which make understanding possible. And to this reply Wittgensteinians will retort that the study of those states and processes is the province of neurology, not psychology. This position is most clearly articulated by Dreyfus 1979 (see also Searle 1984, ch. III) who thinks that it is a fundamental error to try to create any science between neurology and a social phenomenology which uses the evidence of introspection and an intentionalistic vocabulary. Cognitive science, in all its forms, commits this fundamental error for it is precisely the postulation of an intermediate level of reality inhabited by mental representations and described in a vocabulary which is usually, though not necessarily, intentional.

Certainly Chomskyans propose that between brain science and social phenomenology there is space and necessity for a study of mind, specifically of non - introspectible states and processes. It is not entirely clear, however, what ontological commitments are involved. Chomsky writes that `mind' and `mental representation' are an `abstract characterization of the properties of certain physical mechanisms, as yet almost entirely unknown. There is no further ontological import to such references to mind or mental representations and acts' (Chomsky 1980, p. 5). This idea of reducibility - in - principle, so that psychology is `that part of human biology that is concerned at its deepest level with the second - order capacity to construct cognitive structures that enter into first - order capacities to act and to interpret the world' (Chomsky 1975a, p. 38) allows Jenkins 1979 to compare the status of Chomskyan linguistics with Mendelian genetics, and also allows the assimilation of Chomskyan linguistics into what Dennett characterizes as the `intentional stance' in cognitive science (Dennett 1979), the stance which works `topdown', studying the mind as if intentional vocabulary was applicable all the way down.

In the context of the present discussion it is therefore unnecessary to consider stronger ontological claims for mental representations. (Endnote 12) What still requires discussion is the kind of evidential basis on which claims about non - introspectible mental representations can be made. How does one study the inner? For someone who takes the reducibility stance, there is no reason in principle why `bottom - up' work in brain science should not furnish evidence which will organize and constrain `top - down' hypotheses. But as it happens the main source of evidence will be provided by the intuitions of subjects, either those of the linguist (or cognitive scientist) or of the subjects of psycholinguistic (or psychological) experiments. It is therefore important to understand what intuitions are and what they are not.

For Itkonen, introspection pertains to subjective sensations whereas intuition pertains to concepts or rules existing in an intersubjective normative reality. Intuition cannot tell us what happens in space and time, nor can it tell us what happens below the level of self - consciousness: `My linguistic intuition operates only at the level of my (self)consciousness and does not tell me what psycholinguistic processes go on under this level' (Itkonen 1981b, p. 139). `The scope of (the justifiable use of) intuition is certainty, and the limits of intuition are the limits of certainty' (ibid., pp. 138 - 39). Itkonen's view echoes ideas to be found in earlier writings, for example Cavell 1958 and Hare 1960.

Against this view, I use the word `introspection' to cover the activities distinguished by Itkonen as introspection and intuition, since I think they are in important respects alike. I use the word `intuition' to designate that which gives us causally related indexical or symptomatic evidence for the character of underlying psycholinguistic (or, more generally, psychological) processes. Intuitions are not exercises of judgement which claim certainty or any kind of objectivity for the content of judgement and hence which claim the assent of all those implicated by the judgement. Rather, intuitions are reports of appearances, hence subjective expressions which make no judgement about how others will or should respond. In intuition we tell how something strikes us, how it appears to us and thereby (unintentionally) provide causal evidence about our minds. In Wittgenstein's terms, intuitions provide symptoms rather than criteria of what underlies them. This is particularly clear in vision experiments where subjects are asked how things appear to them. It can be made clearer in relation to the elicitation of intuitions of grammaticality if the distinction between claims of forms (1) and (2) below is clearly drawn, and if it is realized and insisted that intuition is used to establish claims of form (1), not form (2), which involves the exercise of judgement (or introspection in my terms):

  • (1) Sentence P seems grammatical to subject S.
  • (2) Sentence P is grammatical in language L, according to subject S.
  • Evidence of form (1) in turn provides the basis for the linguist's claim (3); and evidence of form (2) provides the basis for the linguist's claim (4):
  • (3) Sentence P is grammatical in subject S's internally represented language.
  • (4) Sentence P is grammatical in language L.

In Chomsky's terms (Chomsky 1986, ch. 2), intuition provides evidence for the character of I - languages (internalized languages), whereas introspective judgement - exercised, for example, when a foreigner asks me whether you can say P in English - provides evidence for the character of E - languages (externalized languages). The relationship between the I - language and the E - language, between claim (3) and claim (4), is extremely complex and variable. On this website, my essay "What is English if not a language?" discusses some of that complexity.

In this section I have dipped into a very few aspects of a very large debate concerning the foundations of cognitive science. My sampling of the Wittgensteinian literature makes me think that this places unnecessary restrictions on the forms of legitimate science, motivated by a distrust of psychology which the ontological abstemiousness of Chomsky and Dennett ought to lessen. If Wittgensteinian prescriptions were adhered to, I think we would be obliged to give up on the effort to explain many cognitive phenomena; we would be persuaded that they did not need explanation, merely (re)description (cf. Schatzki 1982 for this view; Schatzki acknowledges Dreyfus's influence). We would persuade ourselves that for all it matters, there might as well be sawdust between our ears, just as Wittgenstein appears to think that there might be no explanatory connection between the form of the acorn and the form of the oak tree. For me, this is not much better than mysticism. There is a lot going on between our ears explanatorily relevant to our phenomenal life. And there is more than one way of approaching its study. Wittgensteinians disallow the topdown approach, but do not seem any more interested in their favoured bottom - up neurological approach. Not one Wittgensteinian has been inspired by findings in neurology in the way that other philosophers of mind have taken inspiration from bottom - up work on cortical specialization and commisurotomy, and not one Wittgensteinian is aware of neurological work which has taken inspiration from top - down cognitive science (see, e.g., Lenneberg 1967; Caplan et al. 1984; Walker 1978). I think this justifies the view that much Wittgensteinianism is simply an expression of the anti - scientific attitudes which sometimes goes with the port in Oxbridge Senior Common Rooms.

4 `Chomskyans treat something we ascribe on the basis of successful practice as the cause of that success'

Wittgensteinians believe that Chomskyans make claims of the form of (P):

(P) Subject S is able to speak grammatically in L because S has an internalized linguistic competence in L

and they believe that this is to put the cart before the horse, even where (P) is thought of as stating only an unnecessary and insufficient condition of grammatical speech in L.

In reality, say the Wittgensteinians, our language game is one in which we ascribe linguistic competence to S on the basis of performance which meets the criteria we lay down. Abilities are manifest in performances and not causally related to them (Baker and Hacker 1984a, pp. 282 - 83). Consequently, (Q) represents the kind of claim we actually make and the only kind we are entitled to make:

(Q) We say S has internalized linguistic competence in L because S is able to speak grammatically in L.

To allow for cases where L is idiosyncratic, it is possible to weaken (Q) to something like (R) without losing the essential difference with claim (P):

(R) We say S has internalized linguistic competence in L because S uses language regularly/consistently in accordance with patterns which can be modelled in a rule - system L

- this claim would also represent Quine's view that grammars can be fitted to behaviour but do not guide it (Quine 1972).

In this section I wish to discuss the counterposing of claims like (Q) and (R) to claims like (P) in relation to Kripke 1982. For Kripke is among those Wittgensteinians who think that Wittgenstein argued against the propriety of claims like (P) and in favour of claims like (Q) and (R). This view is consistent with the general idea of Wittgenstein as some kind of antirealist and verificationist who treats ontological questions in epistemological terms. So to someone who asserts (P), the verificationist will ask , rhetorically,whether this is verifiable and, answering in the negative, the Wittgensteinian verificationist argues that all that we can verify is the satisfaction of criteria for successful performance. All that ascription of competence or mastery or possession of an ability does is mark the satisfaction of these criteria by an individual. It does not designate something standing behind and causally responsible for the successful performance. It is not an explanation.

The most general arguments against this position must be those against verificationism and in favour of realism, motivated by the sense that however much the Wittgensteinian denies it, successful performance can and ought to be explained. I do not propose to enter into those arguments here.(Endnote 13) What I shall do is show how Kripke's way of stating claims like (Q) and (R) involves him in consequences he is obliged to regard as fatal to his argument.

Claims like (Q) and (R) make it a necessary condition of the ascription of competence - of mastery of concepts and meanings - that the speaker or subject deploy those concepts or meanings correctly. Thus Kripke invites us to consider the example of a small child learning addition. `It is obvious that his teacher will not accept just any response from the child. On the contrary, the child must fulfill various conditions if the teacher is to ascribe to him mastery of the concept of addition. First, for small enough examples, the child must produce, almost all the time, the "right" answer . . . . For larger computations, the child can make more mistakes than for "small" problems, but it must get a certain number right and, when it is wrong, it must recognizably be "trying to follow" the proper procedure ... even though it makes mistakes' (Kripke 1982, pp. 89 - 90). Kripke proposes that we specify what is involved in ascribing competence as a contrapositive or reversed conditional. (Endnote 14). Thus, in the case of arithmetical competence this means that `If Jones does not come out with "125" when asked about "68 + 57", we cannot assert that he means addition by "+"' (Kripke 1982, p. 95). Kripke goes on, `Actually, of course, this is not strictly true because our formulation of the conditional is overly loose; other conditions must be added to the antecedent to make it true. As the conditional is stated, not even the possibility of computational error is taken into account, and there are many complications not easily spelled out' (p. 95).

Why are they not easily spelled out.? The answer is surely that to spell them out would show the proposed analysis to be circular, since some primitive notion ofJones meaning addition by `+' would have to be assumed in order to determine what to spell out. Yet the whole object of the analysis is to get rid of such a primitive. In other words, how we complicate the conditional depends on and does not analyse the idea of Jones meaning addition by `+'. If we think Jones really does mean addition by `+' we will go on complicating the conditional until it comes out that he does. In terms of the earlier example: How many sums and which sums the child may get wrong and still be ascribed mastery of addition depends on the view we take as to whether or not the child is adding. It does not, contrary to what is claimed, provide that view. We end up showing the need for claims with the form of claim (P) stated at the outset; in the case of Jones this gives us a claim like (S):

(S) Jones is able to give `125' when asked for `68 + 57' because Jones means addition by `+'.

Kripke's example has to do with arithmetic, but his argument and my counter - argument generalize to language.

Interestingly, my counter- argument is exactly parallel to Kripke's Wittgenstein - inspired argument against dispositionalist accounts of competence, which analyse linguistic or arithmetical competences as dispositions to speak grammatically or give right answers. Thus Kripke observes that dispositionalism is only remotely plausible if dispositionalist analyses are idealized by means ofceteris paribus clauses, since it is clearly false that meaning addition by `+' is simply the disposition to add up correctly. But if the analysis is amended to the formulation that "Meaning addition by `+' is a disposition to add up correctly, ceteris paribus", then, argues Kripke, such `idealized dispositions are determinate only because it is already settled what function I meant' (p. 28). In other words, the analysis is circular; we can only fill in the ceteris paribus clause because we already know or assume that the subject means addition by `+' and consequently we do not achieve the reduction which our analysis was aimed at. Kripke regards the argument that dispositional analyses are inevitably circular as decisive against dispositionalism (Endnote 15) and proceeds to a consideration of the mechanical model which will occupy me in the next section. As far as I can see, his inverse conditional or contrapositive construal of ascriptions of competence is vulnerable to a charge of circularity exactly parallel to that he brings against dispositionalism. He has therefore not provided an alternative way of looking at competence claims to those which take the form of (Q) above, and which it is the stated object of Kripke and other anti-cognitivists to get rid of.

5 `Chomskyans treat something rule - like and normative as something law - like and predictive'

The heading to this section embodies what is probably the commonest Wittgensteinian complaint against Chomskyan linguistics. On this issue, there is near unanimity among the critics. Baker and Hacker write of a `mythology of rules' in Chomskyan linguistics: `Rules cannot "act at a distance"; there is no such thing as a rule determining behaviour unbeknowst to anyone who might cite it, consult it and invoke it in explanation, criticism or justification. To think that rules await discovery by linguistic theories is to confuse the appropriate forms of explanation of normative phenomena with forms of explanation appropriate only to the physical sciences .... A rule is not a predictive device, and whether a speaker does or does not guide his linguistic behaviour by reference to a given rule [i.e., act normatively] is not decided by a prediction that he will behave in such - and - such a manner' (Baker and Hacker 1984a, p. 334). (Endnote 16). Baker reiterates this position in a criticism of Kenny: `More central to Kenny's argument [in Kenny 1984, especially ch. 10] is the claim that speaking a language may be investigated by appeal to tacit knowledge of rules and principles that speakers cannot bring to conscious formulation. To allow Chomsky this latitude is not to grant something indubitable, but rather to fall into radical confusion about rule - governed behaviour and the differences between causal and normative explanations' (Baker 1985, p 754) Kripke concurs that the notion of linguistic competence must be normative not descriptive (Kripke 1982, p. 31, n. 22) by which he means that, `A candidate for what constitutes the state of my meaning one function, rather than another, by a given function sign, ought to be such that, whatever in fact I (am disposed to) do, there is a unique thing that I should do' (p. 24). For Itkonen, language is a matter of rule - following and `The notion of a rule is inseparable from the notion of incorrect action or mistake' (Itkonen 1984, p. 242). Hence, the study of language is a study of normative phenomena, rules, which are not objects in space and time subject to law - like explanation but objects of common knowledge, known (according to Itkonen but not Baker and Hacker) by intuition; their study is consequently not an empirical study (Itkonen 1978a; cf. Katz 1981's Platonism). In all this there are few deviant voices, though Schatzki manages to find a passage in The Blue and Brown Books to support the view that `When Wittgenstein suggests that all behaviour can be described as "rule - governed", he uses "rule" in the sense of a "natural law . . . describing the behaviour of . . ." people' (Schatzki 1983,p 126). I shall ignore this exegesis.

Rules, then, can guide action. They can be followed or not followed. What rule someone is following depends on what rule they think they are following. Inquiry into rule - following must use the hermeneutic procedures of the Geisteswissenschaften. It does not yield scientific predictions, but descriptions of practices. In contrast, laws govern events in nature. They cannot be broken.They act deterministically. What laws govern a domain is independent of what anyone thinks are the laws of that domain. Inquiry into lawful domains uses the empirical procedures of natural science. It yields explanations and predictions of events. In other words, Wittgensteinians posit an entirely orthodox distinction between the moral and the natural sciences, the latter generally conceived positivistically.(Endnote 17). And they assign the study of language wholly to the moral sciences, leaving aside the possibility of `scientific investigations into what parts of the brain are causally related to linguistic abilities' (Baker and Hacker 1984a, p. 300, n. 66) . Such investigations, it is emphaised, have no bearing on the study of the normative phenomena of speaking and understanding.

Now it seems to me that this contrast between two kinds of science has become an unhelpful straitjacket in thinking about the nature of linguistic inquiry. What I want to do in the remainder of this section is try to show that though speaking is, indeed, an intentional activity, the explanation of which will make necessary reference to the goals, purposes, norms, rules, etc., being followed by the speaker, none the less speaking is not an intentional activity under all its relevant descriptions. Under some descriptions, it is properly explained (non - predictively: see again Endnote 17) in causal - mechanical terms. The nature of what is being explained will dictate that the causal - mechanical explanation include reference to inaccessible rules or representations - and this is exactly the position for which Baker and Hacker reserve the most delicate weapons in their armoury of adjectives: `nonsensical' and `absurd'.

Let me come at the argument this way. If speaking is an intentional activity, in which rules are freely followed and which can be freely broken, it follows that `any English speaker can produce at will as many ungrammatical sentences as the theorist wishes. We are no more causally constrained to speak grammatically than we are causally constrained to play chess correctly' (Baker and Hacker 1984a, p. 299). In response I say, first, that it is doubtful that this is true. It is not clear to me that speakers can readily speak ungrammatically. If you can do it, you can become a TV act as did Stanley ..... Can speakers mangle their intonation pattern, for example? And if they can do that, it could be that all they are doing is operating some superficial high - level modification of what they would otherwise have said. Secondly, and much more significantly, Baker and Hacker's argument is certainly not symmetrical as between speechproduction and speech perception and understanding. This is surely damaging to their position: it cannot be that we move from the normative to the causal domain every time we move from being speaker to being hearer.It is simply not true that I can freely mishear and misunderstand speech addressed to me in a language I know. Hearing correctly and understanding is a largely involuntary activity - one which is highly constrained. How else but causally?

Thirdly, the argument invites the response that it does not actually address the issues with which the Chomskyan linguist is characteristically concerned. This is a point which requires a rather longer development.

Undoubtedly, speaking is an intentional activity, stimulus - free or creative. Chomsky sees this as one of the mysteries of language which we simply do not know how to account for (Chomsky 1975a, 1980, and 1982b for formulations of this view). The linguist is not interested, qua linguist studying linguistic competence, in free speech, but rather in the properties of the language with which or in which we speak freely. If you like, the linguist's question is not the question, `What is left over if I subtract the fact that I produce certain linguistically structured sounds from the fact that I perform a speech act?' In this way it does not, at one level, matter if a causal - mechanical theory of speech production fails to distinguish speaking from talking in one's sleep or in a delirium, contrary to what Baker and Hacker claim (Baker and Hacker 1984a, p. 298). Indeed, it is, I think, a good argument in favour of a causal - mechanical theory of language production that we do talk in our sleep and in delirium.

For consider. In sleep or delirium we produce linguistically structured sounds. Ex hypothesi of Baker and Hacker, we do not do this intentionally or voluntarily. Nor is it, presumably, the product of a random or fortuitous process of the sort which may allow a monkey banging away at a typewriter to produce a Shakespeare sonnet or the rustling of leaves on a tree to simulate the Lord's Prayer. It happens too often for that to be plausible, and unlike monkeys or trees, sleeping or delirious humans presumably (still) possess linguistic abilities or capacities. What else can sleep talk or delirium talk be, then, than the product of the operation of mechanisms of speech production operating unconsciously and in an ordinary causal fashion? Clearly, in this case it is not I, as subject of self - consciousness, who speaks when asleep but my mind (or brain) which produces linguistically structured sounds. But, then, when I wake up or recover do the mechanisms which produced my sleep talk or delirium talk suddenly cease to operate? Do I suddenly move from the realm of necessity (laws and causes) to the realm of freedom (rules and reasons)? Surely not. The obvious hypothesis is that the self - same mechanisms which caused my sleep talk or delirium talk are causally involved (with other mechanisms and my activity as a subject) in my waking, self - conscious production of linguistically structured sound as a stratum of the activity of talking. Furthermore, how else can we (currently) describe these mechanisms than in computational or representational (rule - like) terms? It is only undermotivated prescriptive definitions of `rule' and `representation' which pose any obstacle to the cognitivist approach to causal explanation in terms of the operation of mechanisms which operate on representations (signs, symbols) below the level of introspectible awareness. In computational terms, the procedures involved are hardwired or compiled. They are unlike the rules of chess in so far as I follow those in a (largely) non - compiled way (from which follows a hypothesis: sleep chess is rarer than sleep talk).

There remains the question of how, if at all, the causal - mechanical is related to the normative - prescriptive. How do the causal mechanisms implicated in the production of linguistically structured sound interact with the activity of the speaker? At least an analogy for this is provided by the example of what happens in experiments on visual illusions. Shown the Muller -Lyer lines, a causal - mechanical process ensures that I see them as unequal in length and I can truly report this appearance. This is the basis on which I may (erroneously) judge the lines to be of unequal length, a claim which is normative in the sense that it (and I) demands the assent of others. When it is demonstrated to me that the lines are actually of equal length, I must withdraw the judgement. But I can't withdraw the intuition: nothing happens to the appearance the lines present to me. I suggest that speaking is like seeing in that it is a stratified activity, in which strata both influence what occurs in other strata and enjoy relative independence (modularity, encapsulation) (Endnote 18) from those strata. Along these lines the great divide between normative phenomena and the brain comes to appear simply unhelpful. It is a distinction which ought to be revised in the context of the research programme of cognitive science.(Endnote 19)

6 `Chomskyans treat something open - ended (creative) as something closed'

Generative grammars define infinite classes of sentences, since there is no upper bound on sentence length. Though this allows enough space for speakers to produce novel sentences - grammatical combinations which have never been produced before - it does not allow, say Wittgensteinians and others (Hockett 1968; Sampson 1979 and1980a), for essential aspects of creativity. In particular, in defining every sentence as either grammatical or ungrammatical, a generative grammar does not model the essential fuzziness or open - endedness of a language which allows speakers room to innovate creatively in ways which do not involve either rule - breaking or the replacement of one rule by another, but which involve what one might call an extension of the resources of the language (cf. also Rosen 1983 and Wright 1980). The lexicon of a language is where this kind of creativity - by - extension is most obviously possible. Thus Matthews 1979 considers the issue of lexical productivity and argues that a generative grammar cannot handle it: a generative grammar could not specify which colour words of English can take endings in -ness : `blueness' - that's fine; `orangeness' - well?; `magentaness' - you must be joking.

It seems to me that there are two lines of argument along which an answer to Matthews might be constructed.

First, it is worth observing that the persuasiveness of Matthews' arguments in paragraphs 1 - 35 of his book depends at least in part on his (and our) assumption that there is a communal language, English (an E - language in Chomsky's sense) about which each of us who are parties to it ought to share judgements, such that we could, for example, give confident advice to a learner of English. But this begs too many questions. From the fact that I am not sure whether one can say 'pinkness' it does not follow that I am not sure whether I can say it; that is to say, my acceptability judgements (what one can say in English) may well be less firm than my grammaticality intuitions. (See, again, on this website, "What is English if not a Language?"

Secondly, one might start an answer to Matthews by asking the question, `How is it that I know what "magentaness" would mean yet know that one cannot say it or intuit that I would not use it?' One answer to this question would be to propose that mentally represented grammars are organized in terms of a core and a periphery. The core of a grammar massively overgenerates forms; for example, it allows `-ness' to be added to any colour word. The periphery of the grammar functions to reject many of these forms. The resultant grammar remains categorical and either - or rather than fuzzy, though the simultaneous operation of conflicting principles in core and periphery may show up symptomatically as subjectively experienced uncertainty (`I am not sure about whether "orangeness" is grammatical for me or not'). Creative extension of a language would then involve, as one possible mechanism, the assertion of the rights of the core against the periphery and because of this could be subjectively experienced as an extension rather than a rule - violation or rule - creation. One might think of child grammars as having strong cores and weak peripheries with the consequent production of forms which adults find charming and perfectly meaningful. So in answer to my daughter's query, I tell her that we say `Turn the handle clockwise' because that is the way clock hands turn; when we turn it the other way, that is `anti - clockwise'. Her response is immediate, `And that is because anti - clocks turn that way.' QED. This core - periphery suggestion for handling at least some aspects of the creative and apparently open - ended use of language is basically an application of a Chomskyan proposal to handle irregularity. Chomsky writes that irregular morphology and idioms `do not fall naturally under the principles - and - parameters conception of UG [universal grammar - TP]' (Chomsky 1986, p. 147) and suggests that they be handled in a periphery to core grammar.

It is another question whether any progress can be made in explaining the open - endedness of speaking, the fact that what we say and when and how we say it can be both appropriate and unpredictable. Realist doubts about the role of prediction in any scientific explanation might encourage the attempts reconstructively to explain speaking now going on in pragmatics and ethnomethodology. But it is beyond the scope of this essay to consider that question.

ENDNOTES

1 I realize that Kripke does not commit himself to the positions he attributes to Wittgenstein and, in terms of his own previous work, there is good reason why he should not. This needs to be kept in mind when I write about `Kripke' rather than `Kripke's Wittgenstein'.

2 Here is a short bibliography: Chomsky 1975b, Pateman 1982b, Steinberg 1975, Winston 1982.

3 At the other end of the spectrum we find this: `Language is not, for Wittgenstein, "essentially social", as this phrase is usually intended' (McGinn 1984, p. 193, n. 75)

4 Kenny explicitly accepts the idea of the medium independence of language, writing that a language must have `an abstract structure capable of translation into diverse behavioural modalities' (Kenny 1973a, p. 100). Cf. Lyons 1981.

5 Though I have sought to illustrate claim (2) with quotations from Baker and Hacker 1984a, Peter Hacker points out that in their critique of Kripke 1982, Baker and Hacker 1984b, the claim is explicitly denied: `The concept of a language . . . is not internally related to the concept of a social practice but only to the concept of a practice' is, he says, the position defended at length in that second book. There is, he writes, no inconsistency between 1984a and 1984b in as much as the former was explicitly concerned with shared natural languages such as English or German (personal communication 1985).

6 Peter Hacker comments, `You argue against our claim . . . that "language use begins as imitative, reactive and habitual behaviour". . . Your counter - argument to us is that children could "degrade the linguistic input to a pidgin and then, using innate cognitive or grammatical resources, creolise it, only later on trying to bring their creolised output.. . more closely into line with the language being used by adults around them". This begs the question. We argued at some length that the conception of innate cognitive or grammatical resources is incoherent save in the (imaginary) case of someone born with the ability to speak, who can therefore articulate in some form . . . the rules he follows, explain in some way the meanings of the expressions he uses. This is evidently not what you have in mind. The supposition that children prior to having mastered the skills of speech have "innate cognitive or grammatical resources" is not a possibility you can use to refute our case unless you can show that it is intelligible and that the arguments we muster against it are defective' (personal communication 1985). In response, it seems to me that question - begging is always in the eye of the beholder; it is not something theory independent. Baker and Hacker want an a priori answer to questions which in my view can only be answered a posteriori (cf. Kripke 1972). In my view, it is a strong argument in favour of some conception of innate cognitive or grammatical resources that linguists of different persuasions find it necessary to invoke them in order to account for facts of child language development. Later in this chapter, I address more directly some of the arguments Peter Hacker is referring to.

7 This is probably the strongest community view thesis to be found in the literature, (cf. McGinn's formulations of absurdly strong views in McGinn 1984, pp. 194 - 5), though it is hedged about with qualifications in Itkonen i978a.

8 But see Noam Chomsky's interesting comment on this, `It is not so obvious that rule - following is always corrigible, for the reasons you touch on later when you point out that it is extremely difficult to produce ungrammatical intonations. In fact, large parts of language behavior, clearly rule - governed if anything is, are below the threshold where error would even be detected except by trained linguists or instruments, and is not under voluntary control except under extensive training. This is true, for example, of much phonetic detail. It is, in fact, a rather curious fact that children learning a language pick up extremely fine details of phonetic practice which accord with the linguistic behavior of their speech community but are far finer than adults can detect, without extensive training. This fact, incidentally, is rather hard to explain on the assumption that training or even error correction or reinforcement are crucial for language acquisition. In such cases, there is clearly rule - following but no correction under normal conditions, or even a possibility of it. I have no doubt that the same is true in syntax and semantics, as becomes clear as we make progress in understanding the properties of these domains' (personal communication 1985).

9 Baker and Hacker 1984b offer a critique of Kripke which says only that `With us social creatures, rule - following is generally a social practice' (p. 21) and which would allow the deaf child to be a rule - follower on the same grounds as Robinson Crusoe: `Whether Robinson Crusoe is following a rule is independent of whether anyone else is actually doing so too' (p. 40). Cf. Endnote 5.

10 Pulman's leading idea is to combat Quinean scepticism about the very possibility of a theory of word meaning by means of a Chomskyan - realist semantic metatheory which would define `natural' or `accessible' semantics for languages. As Pulman puts it, his book is primarily `intended as a contribution to a theory of word meaning which bears the same relation to the lexicon of an individual language as does the theory of universal grammar to the grammar of an individual language: in fact on the wider interpretation of "grammar" it is intended as a contribution to universal grammar' (p. 168). For example, where an epistemologically oriented meaning theorist would adjust translations and paraphrases to maximize the coherence of his or her interpretations, Pulman proposes an ontologically oriented approach constrained by a metatheory which makes claims about the kind of ontology natural to humans and hence will constrain the class of possible interpretations by reference to a view of the beliefs about the organization of space, time, and objects which human beings are so constituted as to hold. So we motivate the translation (in Quine's famous example) of `gavagai' as 'rabbit' rather than `undetached rabbit part' or `rabbiting' by reference to a theory of nameables which says that humans are so constituted as to view the world in terms of objects rather than object - parts or processes; hence, anything which could be the name of an object is likely to be. This approach is ontological in that it is a theory of how human beings are constituted to see the world. It is not an attempt to solve a hermeneutic puzzle.

11 Cf. Itkonen's claim that transformational grammar `maintains a conception of language which is demonstrably equivalent to the private - language conception' (1978a, p.113).

12 For relevant discussion, see Field 1978; Fodor 1976 and 1981; Marr 1982; Pylyshyn 1984; Stabler 1983; and especially Dennett's more recent papers: 1981, 1982, 1982 - 83, 1983. In my view, the concept of representation deployed by Baker and Hacker 1984a is, at the very least, misleadingly restricted to `signs used by goal - directed creatures with certain purposes and in accord with certain conventions' (p. 296). For attempts to motivate the applicability of a wider concept of representation in the context of a narrow psychology, see e.g. Evans 1978 and Stich 1983. The wider the concept of representation one settles upon, the more likely it is, of course, that the concept ceases to be used fictionally in cognitive science and comes rather to designate some real causal mechanisms.

13 The starting point for them might simply be the following analogy: Suppose someone said, `We say someone is suffering from consumption when they grow pale, weak, thin, clear headed, cough blood and die - and that's what consumption is.' This attitude would rightly be regarded as anti - scientific; had people thought to take it seriously, we should all be dead of tuberculosis.

14 Kripke attributes this view to Wittgenstein, develops it as I indicate but in his n. 76 (pp 93 - 94) declares himself `suspicious' of the procedure.

15 Forbes's (1983 - 84) defence of dispositionalism against Kripke strikes me as unsatisfactory. He treats dispositionalism as a theory about the evidence for saying that S means addition by `+' (p. 232), but Kripke's dispositionalism is considered as an analysis of what it is to mean addition by `+', the whole object of which is to eliminate primitive states of meaning. Further, Forbes's recourse to what is basically Kripke 1972 essentialism about the properties of salt and water (which makes sense of the fact that we can say that salt is water soluble even when it does not dissolve in water) is of no help in confronting Kripke's Wittgenstein's anti - essentialism about meaning, for which it is a deep (and I argue insoluble) problem how we can say S means addition by `+' even when S gets the sums wrong. (Compare Tyler Burge's well-known discussion of such problems as, How can someone mean arthritis by "arthritis" when they have false beliefs about arthritis.)

16 This may be taken as inconsistent even with the modest claim, `All that the rules of grammar tell us is that a person will (ideally) understand and analyse a sentence in certain ways, not others' (Chomsky 1980, p. 89).

17 Note the emphasis on prediction, which realist philosophers of science have argued is not a necessary feature of scientific explanation. See e.g. Bhaskar 1975 and the discussion of these issues in chapter 2 of my book Language in Mind and Language in Society. (Pateman 1987)

18 See Fodor 1983.

19 Peter Hacker comments on the arguments of the preceding four paragraphs as follows, `[1] Your first point [doubting whether we can at will speak ungrammatically - TP] is an expression of opinion, not an established fact. Read out the sentences back to front on this page and you will produce a host of ungrammatical strings. List all the objects on your desk, interpolating "but" between each, and you will produce more, etc.

`[2] Your second argument misfires. The claim that we are not causally constrained to speak grammatically is not refuted by pointing out that one cannot freely "mishear and misunderstand speech addressed to me in a language I know", any more than the claim that I am not causally constrained to move chess pieces correctly is refuted by the fact that I cannot freely misunderstand the (correct) chess moves I observe you to make. Incidentally, that I do, in normal circumstances, understand your request "Pass the sugar, please", that I am not as you put it "free to misunderstand" does not imply that I am causally constrained. (Hence your theoretical question is out of place). It only seems so to you because you persist in thinking of understanding as an act or activity which, as we argued, it is not.

`Your third argument misses its target. Our argument on p. 295 is that rule - following behaviour is a species of intentional behaviour. Behaviour during sleep is not intentional, therefore not a case of following a rule, but merely (if it is!) a case of acting in accord with a rule.

`[3] You rhetorically query whether when "moving" from a delirium or talking in one's sleep, to waking up and speaking "I suddenly move from the realm of necessity (laws and causes) to the realm of freedom (rules and reasons)"and you reply "Surely not." This is hardly an argument. In the same vein I should reply "Surely we do", i.e., we move from what is non - conscious, non - intentional, non - rational, caused behaviour, to conscious, intentional, rational behaviour, done for reasons and within the control of the agent. And that can, bombastically, be described as moving from the realm of causal necessity to the realm of freedom!' (personal communication 1985).

In response, a comment on each of the three points I have numbered: [1] The examples rely on props, written sentences and objects used as a sort of topos, and do not invalidate a claim about limits on spontaneous speech. [2] The argument from the hearer is not presented as a refutation of what Baker and Hacker say about the speaker, but as showing that their argument will not extend to the hearer who remains unaccounted for in what they say. [3] I have left the bombast in. But I also talk of several mechanisms and strata of talking: the point is even if in waking talk, I am in the realm of freedom yet not all of the strata of my talking are in that realm.