What is it to understand a language, hence others?
By Dr. Paul Tomassi
Former Professor of Philosophy
University of Aberdeen
Current understanding of the nature of language owes much to two authors: Noam Chomsky and the later Wittgenstein. What is interesting is that the conceptions of language proposed by each appear to conflict. The key question is: what is it to understand a language? In these terms, the internalist/individualist view of linguistic understanding which Chomsky has consistently advocated throughout his career appears to flatly contradict the later Wittgenstein’s externalist account of linguistic understanding (i.e. the use-based approach to meaning and understanding advocated throughout Wittgenstein’s later work). In short, the relation between these two conceptions is not well-understood. The aim of this paper is to establish some rapprochement along the following lines: philosophy of language may be the richer for what it can learn from empirical linguistics but that area of philosophy remains the context within which empirical linguistics derives its significance.
Solving the Creativity Problem
Considered purely qua empirical linguistics, the most impressive feature of Chomsky’s work is its sheer success. Chomsky first made effective ‘poverty of stimulus’ objections against Skinner in 1959 (Chomsky, N. 1959) and continues to make such objections with effect in 1995 (Chomsky, N. 1995a). The rationalist character of Chomsky’s thinking was made plain in his Cartesian Linguistics (Chomsky, N. 1966). The lack of success of rival empiricist (behaviourist) accounts of linguistic understanding during the whole period confirms the impression that Chomsky’s is the only linguistic game in town. The contemporary empiricist challenge to the Chomskian paradigm (in empirical terms) derives not from any form of behaviourism but rather from connectionist AI, i.e. from recent and ongoing research into modelling cognitive processes by using parallel distributed processing to produce (computer-simulated) neural nets. While modern connectionist approaches have their roots in work from the 1960’s (McCulloch, W.S. and Pitts, W.H. 1965) a more direct challenge to Chomsky emerged in the mid-1980’s in the work of David Rumelhart and James McClelland who produced a neural net capable of ‘learning’ the past tenses of (regular and irregular) English verbs (Rumelhart, D and McClelland, J. 1986). The net achieved this in virtue of a self-corrective learning algorithm, ‘back propagation of error’, i.e. in the absence of any prior knowledge of Chomskian linguistic rules.
It is not clear that Chomskian and connectionist approaches are mutually exclusive. Given that past-tense generation is frequency-dependent only in irregular cases, Steven Pinker, for example, argues that while irregular verbs may be handled case by case, the general regular case is correctly described by Chomskian rules (Pinker, S. and Prince, A. 1994). Further, even if phonologically adequate, connectionist models may not be semantically adequate. On the other hand, connectionists have made (some) progress in representing linguistic behaviours as temporal sequences inexorably bound up with memory (Elman, J.L. 1990). While connectionism remains a thriving neuroscientific research programme, it has yet to articulate an adequate rival to Chomskian linguistics. For Chomsky, the problem of correctly characterising the neuroscience underlying language is exactly that, a neuroscientific problem:
The current situation is that we have good and improving theories of some aspects of language and mind, but only rudimentary ideas about the relation of any of this to the brain. (Chomsky, N. 1995a, 11)
The primary philosophical motivation for accepting a Chomskian account of what it is to understand a language arises from his putative solution to The Creativity Problem (TCP):
(TCP): How are we to explain the fact that from exposure to a finite stock of more or less grammatically well-formed utterances the language-learner can produce and understand a potentially infinite number of unfamiliar utterances?
As Chomsky has put the point:
The most striking aspect of linguistic competence is what we may call the creativity of language, that is, the speaker’s ability to produce new sentences, sentences that are immediately understood by other speakers although they bear no physical resemblance to sentences which are familiar … the fundamental importance of this creative aspect of language has been recognised since the seventeenth century … (Chomsky, N. 1974, 74)
The importance of the problem is, I believe, obvious and, therefore, it is unsurprising that concern with it cuts across a number of academic boundaries, e.g. accounting for this phenomenon has been a sine qua non not only of empirical linguistics but also of the philosophy of language. Chomsky’s approach to the problem centres on an empirical hypothesis to the effect that prerequisites to language use must include the existence of a language faculty or language processor, i.e. a separable cognitive faculty which enables linguistic understanding by generating knowledge of a particular language. The cognitive faculty is universal both in being part of the genetic endowment of human beings and in the sense that it is structure-independent: exposure to linguistic input in a specific language results in the generation of (an instance of) a grammar for that language in the individual. Thereafter TCP is (ultimately) solved by exploiting recursion. Cognitive linguistic theories attempt to describe the operations of the relevant mental mechanisms. If correct, that description will make clear how in fact the phenomenon of the creativity of language is possible:
… a grammar is an account of competence. It … attempts to account for the ability of a speaker to understand an arbitrary sentence of his language … if it is a linguistic grammar it aims to discover and exhibit the mechanisms that make this achievement possible. (Chomsky, N. 1974, 73)
Chomsky has consistently expressed commitment to the existence of a language processor throughout his career. For example, he comments as follows on Riechling’s (Riechling, 1961) interpretation of his work:
[Riechling] asserts that obviously I could not ‘be said to sympathise with such a “mentalistic monster” as the “innere Sprachform”‘. But in fact the work that he is discussing is quite explicitly and self-consciously mentalistic (… that is it is an attempt to construct a theory of mental processes). (Chomsky, N. 1974, 72).
Further, (Chomsky, N. 1980, 5):
… it makes sense to say … that you and I know English … that this knowledge is in part shared among us and represented somehow in our minds, ultimately in our brains, in structures that we can hope to characterise abstractly, and in principle quite correctly, in terms of physical mechanisms.
Finally, the same view is again endorsed (Chomsky, N. 1995a, 13):
Jones’s language faculty has an ‘initial state’, fixed by genetic endowment … performance systems are fully determined by the initial state- … any state changes are internally directed … Let us … call a state of the cognitive system of Jones’s language faculty a ‘language’-or … an ‘I-language’, ‘I’ to suggest ‘internal’, ‘individual’.
The success of Chomsky’s proposed solution to TCP can be measured (in part) by the fact that the form which the solution to the problem must take is frequently a presupposition of the debate rather than a part of its agenda. As Dummett once put the point:
The fact that anyone who has a mastery of any given language is able to understand an infinity of sentences of that language, an infinity which is of course principally composed of sentences which he has never heard before … can hardly be explained otherwise than by supposing that each speaker has an implicit grasp of a number of general principles governing the use in sentences of words of the language. (Dummett, M.A.E. 1978, 451).
The same attitude is also prevalent in the views of some significant contemporary philosophers of mind, e.g. Chomsky’s solution has been integral to Paul Churchland’s account of Matter and Consciousness for more than a decade. Thus, the correctness of Chomsky’s account is regularly presupposed not only in thinking about language but in thinking about thinking.
Two ‘Wittgensteinean’ Objections
Despite its enduring influence, both the general rationalist character of Chomsky’s theory and his approach to TCP in particular have provoked highly critical reactions from philosophers sympathetic to the later Wittgenstein. Before turning to the content of Wittgenstein’s own later writings I refute two examples of the most direct kind of attempt to undermine Chomsky on ‘Wittgensteinean’ grounds (I will not assume that these grounds are genuinely Wittgenstein’s own). In essence, both strategies attempt to show that TCP is a pseudo-problem. First, Norman Malcolm submits that there is simply no case to answer, i.e. that the proper response to the question ‘how did you learn your native tongue’ is simply to assert: “That is not a possible question.” (Malcolm, N. 1993, 56).
Despite this claim, Malcolm goes on to allow that a person’s having grown up in an English-speaking community counts as an explanation of their understanding English and so, it would seem, an answer to the impossible question can be had. In Malcolm’s view, however:
If someone explains his proficiency in English by informing us that he grew up in an English-speaking community, this is not a ‘superficial’ explanation. There is no ‘deeper’ explanation. This is an example of where explanation has come to an end. (Malcolm, N. 1993, 57).
The problem here is that Chomsky may (and clearly does) agree that growing-up in an English-speaking community enters into explanations of why a person understands English (Chomsky, N. 1995a, 13). To generate understanding of a language the language-processor requires exposure to linguistic input in a specific structure, i.e. it is itself ‘modified’ by experience. But why should explanations of understanding English stop at this point? Without convincing criteria for identifying where, in general, explanations end, Malcolm’s suggested restriction appears dogmatic and ad hoc.
Malcolm is not alone in his suspicion of TCP. Baker and Hacker, for example, have challenged the idea that a child understands an infinity, potential or otherwise, of novel utterances (Baker, G.P. and Hacker, P.M.S. 1984, 345-356). TCP misdescribes linguistic understanding. A child will come to understand many words and sentences, but there will remain many which it does not understand, words which have to be learned, sentences which are not fully perspicuous. Does a speaker ever stop being in this situation, i.e. does any speaker ever possess complete linguistic understanding? On this view, speakers gain understanding via ‘traditional learning theory’, i.e. the view that language-acquisition is properly explained in terms of learning rather than ‘growth’ or ‘development’ (Chomsky, N. 1995a, 13-15). It is therefore wrong to think that speakers ever do understand an infinity of novel utterances. Again, TCP is a pseudo-problem.
Two points need to be made here. First, the claim that speakers cannot understand a potential infinity of meaningful sentences looks false. Consider the simple predicate ‘… is a number’. We can assign a potential infinity of terms to this function resulting in sentences which the actual speaker may never have heard but would understand. Second, the point about completeness of understanding rests upon a confusion. TCP is solved by positing a cognitive faculty which, given finite structure-specific input, generates a complete instance of a grammar for that language in an individual. Linguistic understanding consists in possession of a complete instance of that grammar, i.e. of a formal means of generating syntactic structures. It does not follow that the speaker thereby possesses actual understanding of an infinite or even an indefinitely large number of sentences. Compare the situation with understanding the language of propositional logic. To credit someone with an understanding of the grammar of the language is to credit them with understanding seven rules of grammar, i.e. a recursive structure. It is not to credit them with actual understanding of the entire deductive closure of that set. The confusion here is one between a grammar and a language. For the moment then (pending further arguments to the contrary) I assume that TCP is a genuine problem.
Wittgenstein 1: Rule-Following
If some of the familiar points made against Chomsky by supporters of Wittgenstein have been shown to be ineffective, it remains an open question whether there is conflict between Chomsky’s programme and Wittgenstein’s own later writings. In ‘Wittgenstein’s Rule-Following Considerations and the Central Project of Theoretical Linguistics’, Crispin Wright describes one such area of conflict (Wright, C. 1989, 223-264). Here Wright is concerned with “… saving ordinary first-person epistemology …”, i.e. “… the phenomenon of non-inferential, first-person knowledge of past and present meanings, rules and intentions.” (Wright, C. 1989, 236-237). The point is that “An account of the truth-conditions of such claims which, like Chomsky’s … makes a puzzle of this aspect of their first-personal epistemology should be rejected.” (Wright, C. 1989, 236).
Wright identifies a divergence between Chomsky and “the official Wittgenstein” (Wright’s phrase) as regards the autonomy of rules: the latter repudiates the independence of rules (e.g. as ‘rails’) while some such independence seems essential to Chomsky’s linguistic machine in the ghost. Rehearsing Wittgenstein’s discussion of rule-following in Philosophical Investigations (185-219) and Remarks on the Foundations of Mathematics, Section IV, paragraphs 23-47, Wright notes that Wittgenstein’s own focus is epistemic rather than ontic. Here there are four key theses: (i) understanding cannot exceed explanation in content, (ii) following a rule is neither a matter of intuitive cognition of any independent requirement nor (iii) is it a matter of interpretation. Finally, (iv) language is grounded in primitive dispositions of agreement in judgement and action. It follows that what is required by a rule in a particular case is constituted neither by agreement about that case (which could be consequent upon error) nor by the autonomy of the rule itself. Hence, while the factive character of talk about meaning is rescued (e.g. from ‘Kripkenstein’), conflict with Chomsky remains. Agreement in judgement grounds language-use. Talk about meaning is factive. But the relevant kind of agreement here issues neither from intuition nor interpretation. In short, ‘there is no inner epistemology of rule-following’. What remains to be explained (agreement in judgement) is not the product of any cognitive process. Hence, if TCP is a well-formed question, cognitive linguistics proposes exactly the wrong kind of answer.
As Wright notes, Wittgenstein supplies a critique of cognitive models without providing a positive constitutive account of the requirement of a rule on any particular occasion. However, Wittgenstein’s conclusion differs from Malcolm’s: TCP is well-formed. The upshot, then, is a question to which Wittgenstein offers no answer. In contrast, Wright goes on to tease out a test (order-of-determination) on the basis of which an answer might be given. The thought is that there are two contrasting kinds of judgement to be considered. The tracking epistemology of the cognitive linguist makes judgement answerable to independent states of affairs as regards distributions of truth-value. In contrast, certain classes of judgement (analogous to those of secondary qualities and avowals of intention) remain objective in the absence of a tracking epistemology. In the latter case, there is an a priori co-variance of (best) judgement and truth as regards fixing the extension of the rule as concept (the point is not that meaning is a species of intention but rather that judgements about content might be of a similar kind, i.e. objective in the absence of a tracking epistemology). But there is a double underdetermination here: the order-of-determination test itself needs refinement to establish its applicability to judgements of content. Second, even if the results are as hinted, the notion of best opinion awaits analysis as regards its compatibility with theoretical linguistics.
Commenting on Wright (1989) in Language and Nature, Chomsky compares the case of the human language-master, Peter, to that of the Martian language-master, M. Theory T explains the linguistic behaviour of both. Unlike Peter, M enjoys conscious access to T. In a sense, then, M’s inner epistemology clarifies effortless Martian first-personal authority. By the same token, human first-personal authority remains mysterious. For Chomsky, Peter follows the rules of T because he is made that way. Moreover, T provides an account of the behaviour of both which is at least scientifically adequate. Arguably, an important concession follows:
The kind of account that can be offered today, including T, does not ‘make a mystery’ of first-person authority, though it does leave a mystery, about both M and Peter. For both, we have an account that meets the conditions of the sciences … but we lack any insight into the nature of consciousness (Chomsky, N. 1995, 36).
We know why Peter follows the rules he does even if we do not know how. If I read Chomsky correctly and if Wright correctly develops Wittgenstein’s rule-following considerations then it may be that there is no conflict here. Neither author offers a positive account of first-personal epistemology. Therefore, there is no conflict between rival positive theories. But that may be a conclusion too far. After all, Wittgenstein does outline and critically consider the cognitive linguist’s epistemic dilemma (between intuition and interpretation) and offers arguments designed to establish the unpalatability of either horn. Hence, there is conflict with the idea that Chomskian linguistics is properly understood as cognitive.
Supporters of Chomsky have recognised and risen to the challenge here. In ‘Rule-Following and Cognitive Linguistics’, James Page (1998) attempts to provide an account of first-personal epistemology which is precisely both Chomskian and immune to Wittgenstein’s critique. For Page, an adequate account of linguistic rules must establish minimal autonomy: “[the] condition that there must be distinction between one’s believing that one is following a rule and one’s actually doing so …” (Page, J. 1998, 195). Page concedes that the regress-of-interpretation argument precludes an account of tracking on an interpretative basis and defends an intuition-based account. In short, cognitive linguistics provides a theoretical framework of rules and structures on the basis of which an inference to the best explanation (IBE) of language-use can be sustained – these rules are followed because they are parts of abstract systems instantiated in individuals. In effect, Page canvasses a naturalised Platonism which meets the minimal autonomy condition by positing rules as abstracta without bringing in its wake the familiar epistemological problems of its mathematical cousin. To the claim that we lack a model for such intuition Page exploits an example used by James Higginbotham (1989) as regards pronominal anaphora:
(PA) No pronoun can have an antecedent within its own scope.
If correct, it explains why, in a normal context, the sentence
(I) He is one of the reasons I passed Ray.
Does not mean the same as
(2) Ray is one of the reasons I passed Ray. (Page, J. 1998, 199)
Intuiting the rule is instantiating the rule and IBE of our understanding of pronominal anaphora ensures that the rule is (minimally) autonomous. On this view, individual (mis)use can be corrected with reference to the rule itself while agreement in judgement (agreement in intuition) is explained by appeal to the (contingent) fact of the universality of endowment of the language faculty. In other words, linguistic logical priorities are Chomskian.
As Wright notes, one of Wittgenstein’s points against the appeal to intuition is that such an appeal may amount to labelling rather than explanation. The identification of intuition with instantiation may, however, be vulnerable to precisely that point. The challenge was to fill out a sense of cognition appropriate to the language-user’s grasp of linguistic rules, i.e. to characterise the relation between the two in a way which generated an adequate first-personal epistemology. On Page’s view, rather than providing evidence of cognition, correct use appears to constitute cognition. To assert that cognition is identical with instantiation threatens to undercut the possibility of their being anything beyond use which could constitute cognition here, i.e. despite Chomskian intentions, Page may in fact have offered a behaviourist account of relevant mental processes. Whether or not this point goes through, if cognition is identical with intuition and intuition is in turn identical with instantiation there seems little prospect of any substantive first-personal epistemology. On the other hand, it seems unlikely that Wittgenstein’s negative remarks alone establish the impossibility of cognitive linguistics. To know that would be to know not only that those remarks destroyed their targets but that all the targets had been correctly identified. Perhaps intuition-based accounts could be refined. Perhaps consciousness will yield to scientific investigation. Perhaps not. The fact remains that none of the positions considered in this section generate an adequate first-personal epistemology. Thus, a mystery is left. However, empirical linguistics offers rules and structures which appear empirically adequate to the linguistic phenomena including, crucially, TCP. Therefore, I submit, IBE demands recognition that we do follow rules of that kind even if we presently lack a full-blooded account of how it is that such rules are followed.
Cognitive Linguistics and Scientific Realism
Page’s attempted defence raises a further point which merits consideration: Chomsky’s construal of linguistic theory is that of the scientific realist. As noted, the solution to TCP involves hypothesising the existence of a language processor as cognitive system whose operations generate that understanding in virtue of which speakers actually master their language. Chomsky has also consistently claimed both that the formal theory of grammar which is ultimately going to provide the answer to TCP in any particular case is an empirical hypothesis (Chomsky, N. 1974, 121-122, 125) and that the whole project of cognitive linguistics should be understood in naturalistic terms, i.e. as itself an empirical theory evaluable in empirical terms (Chomsky, N. 1995a, Part I). Thus Chomsky’s position is indeed properly construed as that of the scientific (in this respect, entity) realist. Again, Norman Malcolm exemplifies ‘Wittgensteinean’ suspicions:
Chomsky’s proposed solution [to TCP] has a vaguely scientific look. But it isn’t ‘science’ in the sense in which science involves observation, experimentation, testing. Chomsky’s position may be better understood if we take his initial perplexity to be philosophical and his solution metaphysical. (Malcolm, N. 1993, 56)
For any particular hypothesised grammar it can plausibly be argued that empirical evaluation criteria exist by means of which testing and, indeed, refutation may be effected, i.e. any such grammar must be empirically adequate to the linguistic phenomena it attempts to model. A grammar either fits the linguistic facts or not and, even allowing that ‘fit’ may be a matter of degree, it is not difficult to imagine circumstances under which the degree of fit is so low as to constitute a refutation of a particular grammar as an adequate model of English. But, of course, this is not Malcolm’s point:
Could Chomsky set up tests on two groups of infants, one group consisting of those infants who have a grasp of the theory of ‘universal grammar’ … the other … composed of infants who have no conception of universal grammar … and could Chomsky discover that only those in the first group actually learn the language of their community?(Malcolm, N. 1993, 52)
The criterion of empirical adequacy is of application only to specific hypothesised grammars – not to the existence or functioning of those grammars as realisations of the language processor itself. The hypothesis that there exists such a cognitive system is not refuted when a particular grammar is refuted. Every time a child learns a language, the hypothesis that there exists such a cognitive faculty is confirmed while the idea that a child could learn a language without possessing that faculty is ruled out as impossible; despite the lack of any clear notion of what would count as direct evidence against Chomsky’s fundamental hypothesis. Thus, there is no test, crucial, conclusive or otherwise for the concrete existence of the language-processor.
Malcolm’s conclusion that Chomskian cognitive linguistics is unscientific follows only on the narrowest Popperean account of the nature and status of scientific theories. For Popper, the raison d’être of the methodology of conjecture and refutation consists in the possibility of falsification. Testability provides systematic constraints, (allegedly) guarantees objectivity and so constitutes rationality. In virtue of the role of falsifiability as demarcation-criterion, what is not testable is not scientific. Chomsky does offer specific falsifiable theories; most recently, the minimalist theory (Chomsky, N. 1995b). But, again, Malcolm’s point concerns more fundamental elements, i.e. the language-processor hypothesis. Arguably, on a narrow Popperean view, Malcolm’s conclusion follows. But does falsificationism provide the correct account of the nature and status of scientific theories? And does that view supply an account of the notion of falsification adequate to those purposes? A negative response to the latter question undermines the possibility of a positive response to the former. As is well known, there are good grounds for just such a negative response.
If scientific theories are more articulated than Popper allows then the language-processor hypothesis is properly construed as part of the frame of linguistic explanation, that which makes such explanation possible. That hypothesis functions as the Lakatosian hard core of the Chomskian research programme or, in Kuhn’s (1970) terms, defines the Chomskian paradigm. It would be unsurprising that this hypothesis was held fast, i.e. that it was not itself up for grabs in terms of truth-value, if it determined what was so appraisable. No doubt the accumulation of anomalies might bring about change at this level in the structure. Equally, the emergence of adequate, ontically-leaner, alternatives would constitute evidence against the hypothesis: redundant in explanatory terms, it would be a prime candidate for Ockham’s Razor. As yet, there is no such Quinean alternative and while an adequate connectionist model might be epistemically leaner (devoid of rationalistic presuppositions) it would not be ontically leaner (it will posit neural nets whose operations enable language-acquisition).
The fact that the language-processor is internally realist to cognitive linguistics does not show that Chomskian linguistics is unscientific. Rather, it shows that Chomsky is a scientific realist. As such, his position is open to anti-realist objections. But anti-realism may come at a high price here. If the existence of the language-processor is deemed verification-transcendent then there is no fact of the matter about any such faculty. It follows that no appeal to any such faculty can be made. In the absence of an alternative theory, it is then unclear how TCP is to be solved. Moreover, rather than opening up anti-realism, connectionist alternatives replace realism about one kind of entity with realism about another. Presently, then, the hypothesis is not a wheel which turns while touching nothing else. Hence, IBE of the linguistic facts currently requires that cognitive systems responsible for speech production include a language-processor.
Wittgenstein 2: Understanding
Chomsky’s approach to cognitive linguistics is that of the scientific realist, i.e. the existence of the language-processor is (at least) internal to Chomskian cognitive linguistics. Moreover, Chomsky’s position must be realist if an explanation of language-acquisition and, thus, an actual solution to TCP is to be given. On exposure to an initial finite stock of linguistic input, the language-processor produces an instance of a grammar for an individual in which that individual’s linguistic understanding consists. On hearing a novel utterance, that utterance is then processed by the rules and structures composing the grammar so formed. Understanding of the utterance is precisely the effect of such processing. Iteration of the process makes actual the possibility of understanding an infinite number of novel utterances. Chomsky’s solution to TCP is therefore causal.
Can causal explanation properly account for correct use? Crispin Wright (1981) has expressed doubt about the ultimate coherence of any account of tacit knowledge sufficiently robust to supply an actual solution to TCP, i.e. a notion of semantic knowledge adequate to representing compositionality. Commenting on Wright’s discussion, Gareth Evans is similarly sceptical (Evans, G. 1981, 120). Evans attempts to elucidate a possible account of a tacitly known semantic theory capable of representing the structure-reflecting characteristic in terms of dispositions to judge utterances as tokens of sentence-types with associated truth-conditions. Of the notion of disposition employed by the account Evans remarks:
… it is essential that the notion of a disposition used in these formulations be understood in a full-blooded sense. These statements of tacit knowledge must not be regarded as simple statements of regularity, for if they were anyone who correctly judged the meanings of complete sentences would have a tacit knowledge … When we ascribe to something the disposition to V in circumstances C, we are claiming that there is a state S which, when taken together with circumstances C, provides a causal explanation of all the episodes of the subject’s V-ing (in C) … The decisive way to decide which model is correct is by providing a causal, presumably neurophysiologically-based, explanation of comprehension. (Evans, G. 1981, 125-127)
Questions about knowledge of a language unavoidably import wider epistemic issues into an already complex debate. The focus of the present paper is not knowledge per se but understanding. Arguably, the two are intimately connected. Then again, while knowledge may plausibly be said to entail understanding it seems unlikely that understanding entails knowledge. If that is the case, the latter is logically stronger than the former and the two are independent. So far as possible, I focus on understanding in what follows. In these terms, the key point is that a Chomskian solution to TCP presupposes an explanation of linguistic understanding which is ultimately causal in character. That this fairly represents Chomsky’s position is clear:
The language faculty has an input receptive system and an output production system … more than that … performance systems access a common body of information which links them and provides them with instructions of some kind. The performance systems can be selectively impaired … while the cognitive system remains intact. (Chomsky, N. 1995a, 12)
Further, while the initial state of the language faculty is “fixed by genetic endowment” and “the performance systems are fully determined by the initial state”, “The cognitive system of Jones’s language faculty is modified in response to linguistic experience …” (Chomsky, N. 1995a, 13)
Let us tentatively call a state of the cognitive system of Jones’s language faculty a “language” … If the cognitive system of Jones’s language faculty is in state L, we will say that Jones has the I-language L. (Chomsky, N. 1995a, 13)
Hence, Chomsky’s account of linguistic understanding is indeed causal. As such, it could be argued that linguistic understanding is a property of brains rather than persons and, therefore, that no account of personal understanding has been given; a charge to which Chomsky responds, “People in certain situations understand a language; my brain no more understands a language than my feet take a walk.” (Chomsky, N. 1995a, 8). Even so, such understanding is a property of matter rather than mind, of the physical rather than the mental. Chomsky urges a naturalistic investigation of persons as human beings which wholly eschews ‘methodological dualism’ and bites the bullet as regards what has and has not (yet) been achieved (Chomsky, N. 1995a, 2). This stance is underpinned by a strong thesis of incommensurability: of ‘the physical’, ‘the material’ and the notion ‘mechanical cause’ Chomsky argues:
These terms had some sense within the mechanical philosophy, but what do they mean in a world based on … mysterious notions of fields of force, curved space, infinite one-dimensional strings in ten-dimensional space, or whatever science concocts tomorrow? Lacking a concept of “matter” or “body” or “the physical”, we have no coherent way to formulate … the “mind-body problem”. (Chomsky, N. 1995a, 4-5)
Undoubtedly, science has moved on from the mechanical paradigm but the claim that the mind-body problem is now incoherent conflicts (at least) with Kuhn’s (1970) conception of incommensurability according to which there is no possibility of any neutral, logical comparison between paradigms. But the appeal to incoherence suggests exactly that possibility. Kuhn’s point is not that successive paradigms are incomprehensible to one another – given the temporal proximity between two successive paradigms the majority of vocabulary will be shared. Therefore, (partial) communication and thus persuasion will be possible; albeit logical demonstration will not (Kuhn, T. 1970, 198-204). Can we make no sense in contemporary terms of scientific insights in more distant paradigms? Despite the worlds between general relativistic physics and its Cartesian a priori ancestor in particular, Einstein comments that:
Descartes was not so far from the truth when he believed he must exclude the existence of an empty space … It requires the idea of the field as representative of reality, in combination with the general principle of relativity, to show the true kernel of Descartes’ idea; there exists no space “empty of field”. (Einstein, A. 1988, 155-156)
To show that concerns about a causal analysis of personal understanding can be meaningfully formulated is to show nothing about the well-foundedness of such concerns. However, the causal character of the explanation does open up a second line of attack from the writings of the later Wittgenstein. In Philosophical Investigations (e.g. 150-155, 199) Wittgenstein draws attention to an affinity between understanding and ability (the ability analogy). The relation between an ability and its exercise is not causal. Hence, in so far as understanding is (akin to) an ability, understanding is not properly explained in causal terms. If understanding is (akin to) an ability, it has been argued, consideration of other abilities in Chomskian terms exposes his approach as absurd (Baker, G.P. and Hacker, P.M.S. 1984, 339-345). If one is able to ride a bicycle one is thereby able to recognise and ride an infinite number of bicycles but no one will argue that this phenomenon presupposes possession of a recursive theory or mental mechanism in which such understanding inheres. Unlike language-use, bicycle-riding is non-representational but the same point may hold for painting in a genuinely representational sense. The ability analogy suggests a very different approach to TCP: the potential infinitude of linguistic output simply reflects the open-endedness, the plasticity, of language-mastery as an ability (Baker, G.P. and Hacker, P.M.S. 1984, 362-368).
It is clear from Chomsky’s response to Scott Soames (1989) that he does not consider representation of the world to be the defining feature of language (Chomsky, N. 1995a, 26-27). The sense of representation with which Chomsky is concerned is essentially mentalistic and, as such, anathema to many whose philosophy of language is shaped by the thought of the later Wittgenstein (Chomsky, N. 1995a, 19, 53). The key point here, however, is simply that both representational painting abilities and bicycling abilities lack syntactic structure. Languages are syntactic structures. Hence, understanding a language involves grasping syntactic structure, i.e. rules of composition within that structure. Were this not the case, it would be difficult to see how well-formed complex linguistic expressions could be functions of their constituent (logical and non-logical) parts, i.e. to see how any theory of such could have the property of “structure-reflection”. The ability analogy provides no basis for an alternative account of the projective qualities which the appeal to recursion can establish. But if there were no finite basis for composition in a language it would be difficult to see how that language could be learned.
Even if understanding is akin to an ability, TCP could presumably be re-formulated in terms which reflect that fact, i.e. the suggested re-description fails to preclude the possibility that Chomsky has identified the nature of those structures which underpin and explain linguistic understanding. Arguably, this reply overlooks a key element in the analogy: understanding is not a psychologically continuous process, i.e. it is not a mental state. In contrast, states of the language-processor as cognitive system will be precisely that, i.e. mental states. Again, this raises questions of first-personal epistemology: linguistic rules come with no associated inner phenomenology. The crux of the issue here is whether or not we credit speakers with an understanding of rules which they cannot articulate and which do not explicitly feature in their intentions as regards correct use. For example, speakers ordinarily construct well-formed interrogatives without any rule-statement to the effect: ‘Move the first verb in the main clause to the front’ ever figuring in their intentions. Most speakers will not be able to state this rule and, when confronted with it, some will fail to realise that the rule describes this aspect of their linguistic behaviour. Presumably, such a rule could figure in a speaker’s intentions but while speech is in some sense a conscious activity, understanding of the rules underpinning the production of grammatically well-formed sentences in English is not an item in the stream of consciousness of the average English speaker. It follows, so the argument goes, that the average English speaker either has no understanding of those rules or understands such rules tacitly or implicitly. If it is allowed that speakers understand rules which they do not, perhaps could not, articulate or recognise, what could speakers be said not to understand?
The first point to make here is that this criticism applies equally to connectionist strategies. While an adequate connectionist model will restore elements of traditional learning theory (e.g., repetition and imitation) it will not result in a set of linguistic processing operations of which speakers are consciously aware. In this respect, then, connectionism offers no advantage over the Chomskian approach. The claim that linguistic rules generally do not feature in the intentions of speakers as regards speech-construction is no doubt incontrovertible. But it does not follow that the rule for forming interrogatives is in error. That rule accurately describes correct composition, i.e. the rule saves the (linguistic) phenomena. Moreover, the argument here is enthymematical; the hidden premise being that all understanding must be explicit, at least, in principle. Speakers can certainly be taught empirical linguistics, however, and so could, in principle, consciously reflect upon principles of correct composition. Hence, a stronger thesis is required, i.e. understanding must be able to be made explicit without residue.
If understanding must be explicit then English speakers, in general, do not understand the rules of correct composition in English. But what reason have we to accept the premise here? Knowing how to play the piano or drive a car involves adhering to rules and practices and exploiting techniques which need not, and generally do not, feature in the consciousness of the piano player or the car driver. But we would not be tempted to say, as a result, that such persons did not know how to play the piano or drive a car. Again, questions about the nature of knowledge itself lie beyond the scope of the present paper. However, the obvious move may seem to be to distinguish, with Ryle, between theoretical knowledge and practical knowledge, identifying knowledge of a language as a disposition belonging to the latter category (Ryle, G. 1949, 25-60). But how plausible is that move? While Dummett (1993) continues to defend the idea that language is known, the following observation is pertinent here:
The dichotomy theoretical/practical knowledge is too crude to be applied to knowledge of a language. If I cannot ride a bicycle I may still know what it is to ride one; but if I do not know Tibetan, I do not know what it is to speak it. Our need is therefore not to dispute over whether knowledge of a language is genuinely knowledge, still less over whether it is theoretical or practical knowledge, but to refine our excessively coarse conceptions of theoretical and practical abilities. (Dummett, M.A.E. 1987, 283)
The ability analogy succeeds in pointing up some categorial differences in our ways of talking about cognitive systems and understanding. Other such differences are also noteworthy. While it might make sense to talk of degrees of understanding, talk of a cognitive system being partly in a certain state makes little sense, i.e. the cognitive system either is or is not in a given state. More generally, while understanding admits of qualitative description (e.g. as deep or shallow) mental states do not. It would be unwise to infer that these considerations preclude a Chomskian account of linguistic understanding. As noted, the thesis that understanding must be explicit breaks down as regards numerous abilities. If linguistic understanding involves analogous abilities then such understanding should not be expected to be explicit, i.e. there is conflict between the idea that understanding must be explicit and the idea that understanding is akin to an ability. Similar conflicts may also arise as regards knowledge itself. If it is allowed that understanding can derive from theoretical knowledge in the sense captured by traditional analysis of ‘S knows that P’ then, given that ‘S knows that P’ is intended to define theoretical rather than practical knowledge, it would be odd to describe such understanding in terms of ability. Equally, given that S either does or does not know that P it is difficult to see how such states could admit of qualitative rather than quantitative description. These considerations hint at something of the ‘coarseness’ of the theoretical/practical distinction alluded to by Dummett. More importantly, they also hint at the implausibility of the attempt to identify understanding with practical abilities. In so far as manifestation arguments presuppose just that premise, so far has the linguistic realist grounds for rejecting anti-realist conclusions. For present purposes, however, the key point is that in the absence of a clear general theory of understanding there seems to be little reason to reject a Chomskian conception of linguistic understanding tout court, rather than to reflect further on grounds for commitment to what are, at best, analogies.
The conclusion of the previous section was just that, pending adequate alternatives, explanation adequate to the linguistic facts currently presupposes acceptance of a Chomskian account of linguistic understanding. Here I have argued that the notion of understanding is itself too little understood to preclude Chomskian ideas, i.e. without a much more developed theory to appeal to, Chomsky’s account of linguistic understanding cannot be ruled out in virtue of the nature of understanding qua understanding.
If the arguments of the foregoing sections are accepted, a complete account of an individual’s linguistic understanding should, at least in principle, include recognition of the existence of the language-processor as cognitive system. If there does exist such a cognitive system then, like other cognitive systems, it will, in principle, be describable in physical terms, i.e. correct descriptions of that cognitive system will appeal to relevant natural laws. Hence, arguments from analogy can be constructed, e.g. just as complete description and full explanation of visual cognition should include relevant neurophysiological facts and laws, so description and explanation of linguistic understanding should include neurophysiological facts about, and laws governing, the language-processor. Given this conclusion, a further question naturally arises: could an individual’s linguistic understanding be completely described in empirical terms?
The focus of Chomsky’s investigations concerns the nature and function of the language-processor as cognitive system. What is of particular importance here is the generation by that cognitive system of a structure-specific grammar in an individual. Correct descriptions of any such structure-specific grammar in an individual will be correct descriptions of that in virtue of which a speaker masters a language, i.e. any such description must ultimately describe a syntactic structure adequate to a solution to TCP. Up to this point I have been concerned to establish that possession of the right kind of syntactic structure does constitute a necessary condition of linguistic understanding; the claim to sufficiency, however, seems much harder to establish. While the internalist/individualist (and naturalistic) focus of Chomskian investigations is valuable, it may set unduly stringent limits both to individual linguistic understanding and to language itself. As regards the former, for example, to identify linguistic understanding with the state of a cognitive system in an individual may well deliver knowledge of syntactic structure adequate to a solution to TCP but at the cost that any actual use of such understanding is secondary to, derivative from, and, perhaps, even incidental to knowledge of a language. If linguistic understanding is identical with the state of the relevant cognitive system in the individual then any use of such understanding, for example, to make reference to the world is exactly that, i.e. a derivative use of language rather than a constitutive element of an individual’s understanding of that language. By parity of reasoning, exploiting one’s understanding of English to describe or to think about (in the intentional sense of thinking about Uncle George, for example) are, again, applications of linguistic understanding rather than constitutive elements thereof. Chomsky (1995a) contains strong evidence that, in his view, this construes such phenomena correctly:
As for semantics … the argument for a reference-based semantics … seems to me weak. It is possible that natural language has only syntax and pragmatics … there will be no provision for what Scott Soames calls “the central semantic fact about language, … that it is used to represent the world”, because it is not assumed that language is used to represent the world. (Chomsky, N. 1995a, 26-27)
Further, “… general issues of intentionality, including those of language use, cannot reasonably be assumed to fall within naturalistic inquiry” (Chomsky, N. 1995a, 27).
That the wide variety of phenomena involved across the entire spectrum of language use are presently too ill-defined to admit of purely (natural-scientific) empirical investigation is uncontroversial. That such phenomena can be subtracted from the linguistic abilities of an individual while leaving that individuals’ understanding of language intact is controversial. It is difficult to imagine an individual possessing a language who did not, or could not, make reference to the world, describe or think intentionally in that language. The point here is not simply one about the limits of imaginability. Would it be correct to ascribe language-mastery to an individual who did not know how to carry out any or all of these activities within that language? In the fullness of time some of these phenomena might be pinned-down in ways exact enough to enable empirical investigation. Even if this is true, however, an important stumbling block remains. Among such phenomena are activities which could not be completely defined by a state of the cognitive system because they are defined, at least in part, by how we conceive of what is being done linguistically, e.g. intentional thinking about the state of Uncle George, making a promise, taking vows at a marriage ceremony or signing a cheque. More generally, such socially-defined concepts and practices unavoidably import normative elements into descriptions and explanations of linguistic practices. Again, many will balk at the idea that a causal explanation could properly characterise any such practice. As Friedrich Waissman once put it:
We shall say that the meaning of a sign ‘a’ is the effect which it should have, not the effect that it will have. This, of course, is nothing but a contribution to the grammar of the word ‘meaning’ to prevent it being confounded with that of the word ‘effect’. (Waissman, F. 1965, 116)
… if a command to do p were only a command to do p in virtue of p being done, it would no longer be the command to do p if it ceased to effect this movement. According to this view, … there would be no such thing as a disobedient child. (Waissman, F. 1965, 116)
The crux of the matter is just whether neurophysiological phenomena can possibly be identical with any normative aspect of a concept or practice with such a concept? Sufficiently robust eliminativists will respond positively while their opponents respond negatively. Settling this matter is beyond the scope of the present paper. However, failure to settle this issue need not prevent further pursuit of the issue at hand. As hinted above, difficulties of this kind follow not only from the Chomskian definition of individual linguistic understanding but also from the definition of a language (of language itself) as identical with an ‘I-language’, i.e. a state of the cognitive system of an individual’s language faculty. Given the latter identification, it is unclear how socially-defined concepts and normative features of linguistic practices can be represented in purely naturalistic, individualistic terms. Pending convincing counter-arguments, it remains equally unclear that a complete account of an individual’s linguistic understanding can be given without appeal to some such concepts.
Wittgenstein 3: Privacy
Current understanding of the nature of understanding does not provide a basis sufficient to warrant rejection of Chomskian accounts of linguistic understanding in terms of mastery of syntactic structure. Explanation adequate to the facts of language-acquisition requires acceptance of Chomsky’s solution to TCP. Hence, a Chomskian account of syntactic structures is a necessary part of any complete account of what it is to understand a language. However, in the previous section I noted how Chomsky’s internalist/individualist definition of language forces key facts about language-use such as reference, intentionality and normativity onto the sidelines as derivative applications of linguistic understanding. To many, reference will seem a particularly significant phenomenon without which linguistic understanding and language itself are incompletely specified. Thus, for example, the exclusion of reference has fuelled critics such as Hilary Putnam (1989) and Scott Soames (1989) – as the latter puts it, reference is ‘the central semantic fact’. For the later Wittgenstein, though we may be misled by surface appearance, reference is but one kind of use to which language may be put. The diversity of language-games and the idea of language as a tapestry woven from such threads entail that there is no single central semantic fact.
The notion of a language-game naturally embeds linguistic practices within social contexts. Thus, socially-defined concepts and normative aspects of their use are at the heart of Wittgenstein’s later thinking about language. On psychological predicates, for example, he is explicit in his rejection of the idea that meaning is any kind of mental process. In Philosophical Investigations, a number of arguments are devoted to this negative conclusion. While these arguments are usually lumped together under the banner of ‘The Private Language Argument’ a number of different but importantly related groups can be separated out. Preliminary considerations pointing up categorial differences between meaning and mental processes are given on page 59 (see also the Brown Book, pages 158-167). Arguments against both the necessity and sufficiency of any mental event as regards meaning are given in 139, 141, 151-2 and 151-2, 329-30, page 18e, respectively. A further group of remarks are devoted to establishing the impossibility of definition of psychological predicates in terms of mental states or processes, e.g. 258, 293. In contrast with the case of rule-following, many remarks point to the general shape of a more positive thesis in terms of criteria, i.e. public and irreducibly externalist elements central to language-games with psychological predicates. (See, for example, the Blue Book, pages 24-25, or Philosophical Investigations 269, 288, 290, 404 and especially 580). There are parallels between criterially-based judgements and that class of judgements teased out by Crispin Wright in his discussion of rule-following (Section III above). Criteria constitute standards of correctness and provide a basis for best judgement as regards third-person ascriptions of psychological predicates. Unlike first-person avowal, a priori co-variance of best judgement and distribution of truth-value is harder to establish here. In Wright’s terms, the third-personal counterpart to the non-trivial condition excluding akratic, self-deceptive possibilities in first-person cases will seek to rule out mimicry, deception and the like. Given his comments in Volume 1 of Remarks on the Philosophy of Psychology (especially at 137) it seems unlikely that Wittgenstein himself would argue for the satisfiability of such a condition. For present purposes, a weaker notion may suffice. For example, there are criteria for someone’s being in pain recognition of which is (weakly) positive-presumptive in favour of third-personal pain-ascription. Even if there is no guarantee of correctness in any given case, I cannot be wrong in general that such behaviour warrants pain-ascription, i.e. descriptions of the satisfaction of pain-criteria (in part) fix the content of the concept. If this is correct, psychological concepts are ultimately socially-defined and sentences utilising such concepts come not merely with syntactic structure but with inferential structure fixed by normative rules of inference. It follows that to understand this aspect of the use of English is to grasp irreducibly social and normative aspects of use.
Prima facie, the competing conceptions of linguistic understanding offered by Chomsky and Wittgenstein appear mutually exclusive here. If that is the correct diagnosis, the work of both authors may be incomplete, i.e. neither is alone adequate to account for language-mastery completely. First, consider Chomsky’s position. The identification of linguistic understanding with I-language precludes from such understanding any use to which I-language is put. But some such uses (reference and intentionality, for example) and some aspects of the use of socially-defined concepts (normativity, for example) are sufficiently central to what it is to understand a language that anyone lacking such abilities could not be correctly described as having mastered the language. Such features of language-use appear to belong not to syntax or combinatorial semantics but rather to semantics proper – an analysis which chimes with early definitions of the subject given by the generative grammarians themselves. Chomsky (1995a) recalls the earliest such definition as, “the study of how this instrument, whose formal structure and potentialities of expression are the subject of syntactic investigation, is actually put to use in a speech community” (Chomsky, N. 1995a, 26-27). Currently, supposing language to have a semantics, Chomsky (1995a) offers a sketch of an individualist account within which an internalist notion of mental representation plays the pivotal role (Chomsky, 1995a, 26-27). This approach opens up further conflict with Wittgenstein, i.e. with the arguments against privacy cited in the previous section. But, in this instance, a sketch is all that is on offer, Chomsky’s case is not proven and the nature of semantics remains an open question. Whatever the bigger picture turns out to be, it at least seems likely that some (if not all) of the features of use considered in this and earlier sections properly belong to semantics. If that is so, key aspects of the use of the linguistic instrument are ineliminable from a full account of linguistic understanding.
To concede that some uses of language are sufficiently semantically central to count as constitutive elements of what it is to understand a language is not to dismiss the importance of Chomskian linguistics. The value of Chomsky’s work is exemplified in his account of syntactic structures and combinatorial semantics, i.e. in his provision of a solution to TCP. If the arguments of earlier sections are correct that solution must be accepted as a necessary part of a complete account of linguistic understanding. Therefore, if there really is no place in Wittgenstein’s later account of linguistic understanding for grasp of syntactic structure then Wittgenstein’s account must also be incomplete. If such a reading is correct exegetically, then, each author does, indeed, give a distinct account of linguistic understanding which might be separated out, e.g. ‘Understanding 1’ and ‘Understanding 2’. But if these are mutually exclusive conceptions then, just in itself, the work of both authors is incomplete. Neither Understanding 1 nor Understanding 2 is alone adequate to completely capture both the syntactical and semantic elements of such understanding. Therefore, neither is a complete account of linguistic understanding.
Cognitive Linguistics and the Philosophy of Language
I noted in the introduction to this paper that a correct account of linguistic understanding has implications for the nature, scope and limits of the philosophy of language. In the previous section I argued that, while necessary, grasp of syntactic structure alone is not sufficient to fully capture the notion of understanding involved in understanding a language. Rather, language-mastery presupposes abilities to think intentionally, make reference to the world, recognise socially-defined concepts, normative aspects of use etc. Chomsky’s central reason for excluding such key features of language-use from the programme of cognitive linguistics was that such phenomena are not amenable to naturalistic investigation. Chomsky also urges that (methodological) naturalism is the appropriate methodology for the philosophy of language itself. Pace Dummett’s claim (in Seas of Language) that naturalistic investigation might deliver (correct) psychological hypotheses but would thereby ‘contribute nothing to philosophy’, Chomsky comments, “Had Hume achieved [his] goals, he would have established “psychological hypotheses” … but would not yet have contributed anything to philosophy.” (Chomsky, 1995a, 34)
Hume scholarship aside, the key question here is just whether a complete account of the philosophy of language can be given in Chomskian terms? This raises two related issues. Is naturalism the appropriate methodology as regards philosophical linguistic investigations? Further, are individualist/internalist concerns exhaustive as regards the philosopher’s interest in the nature of language? No doubt, methodological naturalism is the correct methodology for empirical linguistics but the claim that the correct account of method is identical in both science and philosophy is a harder one to establish, presupposing, as it does, that empirical knowledge is the only kind of knowledge to be had. Empirical theories may contribute significantly to philosophical puzzles, e.g. Chomskian linguistics and TCP. Natural science generates models and theories which are themselves proper objects for philosophical reflection. But such models and theories do not automatically provide solutions to any or all relevant philosophical problems. For example, despite that fact that general relativistic physics is well-understood, the debate between substantivalists and relationalists over the ontological status of spacetime persists (even if that theory did deliver a decisive answer here it is not clear that the philosophy of space and time would, in effect, be at an end).
On individualism, Chomsky admits that phenomena such as intentionality, reference, representation and, indeed, actual language-use are not suitable cases for naturalistic treatment (Chomsky, 1995a, 27). Naturalistic investigation pursues its ends independently of and neutral to questions of ‘human significance’ (Chomsky, 1995a, 10). Again, phenomenological approaches to issues in space and time no doubt have no place in contemporary spacetime physics but it is much less clear that they have no place in the philosophy of space and time. Coming closer to home, the claim that issues arising from actual language-use in social contexts are not suitable cases for philosophical treatment seems false. It is, at least, difficult to see why philosophers of language should not concern themselves with the broad range of irreducibly social phenomena involved, at least some of which are relevant to language-mastery and linguistic understanding. Providing an exact definition of what is distinctively philosophical about philosophical enquiry is notoriously difficult but an example may provide a hint. Consider the problem of other minds (or, for that matter, the mind-body problem). The argument from incommensurabilty considered earlier fails to show that these problems cannot be meaningfully stated. It also remains unclear how a complete physical description of a human being could, of itself, solve those problems. In contrast, the kind of conceptual analysis of ‘human being’ given by Wittgenstein in Philosophical Investigations (280-90) may show precisely that scepticism about other minds cannot ultimately be coherently stated. Methodologically, Wittgenstein’s approach to analysis is non-naturalistic and the object of that analysis is precisely a socially-defined concept. While it may not be possible to define philosophical explanation, the point is not that philosophy owes empirical linguistics such an explanation but rather that empirical linguistics needs to provide more justification for the wholesale revisionism proposed as regards philosophy of language.
A complete account of linguistic understanding must include sufficient empirical-linguistic theory to allow a solution to TCP. At the same time, no account of such understanding is complete unless it accommodates central uses of language. Empirical linguistics confines itself to notions sufficiently well-understood to admit of empirical investigation. The fact that this precludes consideration of reference etc. provides good reason to include such uses of language within the boundaries of the philosophy of language proper. Thus, while philosophy of language may be enriched by what it can learn from empirical linguistics, the philosophy of language remains the context within which empirical linguistics ultimately derives its significance.
- I am grateful to James Page for extensive comments on an earlier draft of this paper.
- I do not believe that the relation between the two is properly understood as that between a Chomskian and a convention-based semantics, i.e. in terms of Stephen Laurence’s ‘A Chomskian Alternative to Convention-Based Semantics’, Mind, 1996. The theories of Grice (and Fodor) seem as opposed to the later Wittgenstein’s account of what it is to understand a language as Chomsky’s theory. Again, Wittgenstein’s account of understanding (section V of the present paper) may conflict equally with both convention-based semantics and Chomskian theoretical linguistics. Hence, the later Wittgenstein’s account of understanding a language is not a ‘convention-based semantics’ in the sense considered by Laurence.
- Churchland’s assumption of the correctness of Chomskian linguistics is as integral to Churchland (1994) as it is to Churchland (1984). That the correctness of Chomsky’s position is assumed by Churchland is particularly clear in Churchland’s (1984) account of Matter and Consciousness. There Churchland literally presupposes the correctness of a Chomskian solution in his account of ‘conscious intelligence’. Consider, for example, a few of Churchland’s introductory remarks on Artificial Intelligence: “These rules [of arithmetic] you already know … So you already possess a self conscious command of one formal system. And given that you can think at all, you also have at least some tacit command of the general logic of propositions as well, which is another formal system. What is more interesting is that any formal system can be automated.” (Churchland, P. 1984, 100). Why does Churchland say that unconscious knowledge of a formal, mathematical system is a precondition of conscious thought? No argument is given to substantiate this claim, though it is far from self evidently true. It is quite clear why Churchland holds such a view given that author’s attitude to Chomsky: “… these artificial languages [ BASIC, PASCAL etc.] are much simpler in structure and content than human natural language, but the differences may be differences only of degree, … the theoretical work of Noam Chomsky and the generative grammar approach to linguistics have done a great deal to explain the human capacity for language-use in terms that invite simulation by computer. (Churchland, P. 1984, 16).… these artificial languages [ BASIC, PASCAL etc.] are much simpler in structure and content than human natural language, but the differences may be differences only of degree, … the theoretical work of Noam Chomsky and the generative grammar approach to linguistics have done a great deal to explain the human capacity for language-use in terms that invite simulation by computer.” (Churchland, P. 1984, 16). Churchland’s first argument on AI is an enthymeme: thinking is linguistic. Hence, his conclusion is presupposed. But once that is conceded Churchland’s reasoning obviously runs as follows: Given that to think is always to think in a language and that language-acquisition is only possible in virtue of something like Chomsky’s rules which we know innately then unless we had such innate knowledge language and, consequently, thought, would be impossible.
- A summary of some central objections can be found in Tomassi, P. 1995: ‘Logic and Scientific Method’ in Phillips, C.I. (ed.) 1995: Logic in Medicine, (Second Edition), British Medical Journal Supplement. Pages 48-53.
- This expression is borrowed from Wright, C. 1981: ‘Rule-Following, Objectivity And The Theory of Meaning’ in Wittgenstein: To Follow a Rule. Holtzman, S and Leich, C (eds.). Routledge and Kegan Paul. London. Page 118.
- There is a familiar response to the kind of argument used here (and to that used by Dummett in Section I above) which is well-put by Putnam, H. 1974, i.e. in his ‘The Innateness Hypothesis’ and Explanatory Models in Linguistics’ in ‘Symposium on Innate Ideas (b) ‘in The Philosophy Of Language. J.R. Searle (ed.). Oxford Readings in Philosophy. Oxford University Press. Oxford, page 133: “There is the ‘argument’ that runs ‘what else could account for language learning?’ The task is so incredibly complex that it would be miraculous if one tenth of the human race accomplished it without innate assistance. (This is like Marx’s proof of the Labour Theory of Value … which runs, what else could account for the fact that commodities have different value except … that the labour-content is different).” Putnam is rightly unhappy about this form of argument, not least because it is no form of argument at all, i.e. the question is rhetorical. My inability to conceive of alternative forms of explanation other than my own has no implications as regards alternative explanatory possibilities. But this point is merely motivational., i.e. so far, nothing has been said which challenges the correctness of Chomsky’s solution to TCP and, indeed, (on this occasion) Putnam concedes that neither he nor traditional learning theory has any alternative solution.
- I am grateful to my colleague Jonathan Friday for this point.
- Baker, G.P. and Hacker, P.M.S. 1984: Language, Sense and Nonsense, A Critical Investigation into Modern Theories of Language, Basil Blackwell. Oxford.
- Chomsky, N. 1959: ‘A Review of B.F. Skinner’s Verbal Behaviour’, Language, 35.
- Chomsky, N. 1966: Cartesian Linguistics: A Chapter in the History of Rationalist Thought. Harper and Row. New York & London.
- Chomsky, N. 1974: ‘Topics in the Theory of Generative Grammar’ in The Philosophy Of Language. J.R. Searle (ed.). Oxford Readings in Philosophy. Oxford University Press. Oxford.
- Chomsky, N. 1980: Rules and Representations. Oxford: Basil Blackwell.
- Chomsky, N. 1995a: ‘Language and Nature’. Mind, Vol. 104. No. 416.
- Chomsky, N. 1995b: The Minimalist Program. MIT Press. Cambridge, Mass.
- Churchland, P. 1984: Matter and Consciousness. MIT Press, Cambridge, Mass.
- Churchland, P. 1994: ‘Review of Searle (1992)’ in: London Review of Books, 12th May, 1994.
- Dummett, M.A.E. 1978: Truth and Other Enigmas. Duckworth. London.
- Dummett, M.A.E. 1987: ‘Reply to Prawitz’ in Michael Dummett, Contributions to Philosophy. B. Taylor (ed.) Martinus Nijhoff.
- Dummett, M.A.E. 1993: Seas of Language. Clarendon. Oxford.
- Einstein, A. 1988: Relativity, the Special and General Theory. Methuen & Co. London.
- Elman, J.L. 1990: ‘Finding Structure in Time’, Cognitive Science, 14, 1990.
- Evans, G. 1981: ‘Semantic Theory and Tacit Knowledge’ in Wittgenstein: To Follow a Rule. Holtzman, S. and Leich, C. (eds.). Routledge and Kegan Paul. London.
- Higginbotham, J. 1989: ‘Knowledge of Reference’ in Reflections on Chomsky, A. George (ed.). Blackwell. New York.
- Kuhn, T. 1970: The Structure of Scientific Revolutions. University of Chicago Press Ltd. London.
- Malcolm, N. 1993: Wittgenstein, A Religious Point of View? Routledge. London.
- McCulloch, W.S. and Pitts, W.H. 1965: ‘A Logical Calculus of the
- Ideas Immanent in Nervous Activity’ in McCulloch, W.S. 1965: Embodiments of Mind, MIT Press. Cambridge, Mass.
- Page, J. 1998: ‘Rule-Following and Cognitive Linguistics’ in Southwest Philosophy Review, The Journal of The Southwestern Philosophical Society. Vol.14. No. 1.
- Pinker, S. and Prince, A. 1994: ‘Regular and Irregular Morphology and the Psychological Status of Rules of Grammar’ in S.D. Lima, R.L. Corrigan and G.K. Iverson (eds.) The Reality of Linguistic Rules. John Benjamins. Philadelphia.
- Putnam, H. 1989: Representation and Reality, MIT Press. Cambridge. Mass.
- Reichling, A. 1961: ‘Principles and Methods of Syntax: Cryptanalytical Formalsim’, Lingua 10. Pages 1-7.
- Rumelhart, D and McClelland, J. 1986: ‘On Learning the Past Tenses of English Verbs’ in Rumelhart, D., McClelland, J. and PDP Research Groups (eds.). 1986: Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Vol. 2. MIT/Bradford. Cambridge, Mass.
- Ryle, G. 1949: The Concept of Mind. Hutchinson & Co. London.
- Soames, S. 1989: ‘Semantics and semantic competence’ in J. Tomberlin (ed.) Philosophical Perspectives 3. Ridgeview. Atascadero, California.
- Waissman, F. 1965: The Principles of Linguistic Philosophy. Macmillan. London
- Wittgenstein, L. 1967: Philosophical Investigations. Oxford: Basil Blackwell. Translated by G. E. M. Anscombe.
- Wright, C. 1981: ‘Rule-Following, Objectivity and the Theory of Meaning’ in Wittgenstein: To Follow a Rule. Holtzman, S and Leich, C (eds.). Routledge and Kegan Paul. London.
- Wright, C. 1989: ‘Wittgenstein’s Rule-Following Considerations and the Central Project of Theoretical Linguistics’ in Reflections on Chomsky, A. George (ed.). Blackwell. New York.
Originally published by Minerva: An Online Open Access Journal of Philosophy (Volume 4, November 2000), an open access resource, republished for educational, non-commercial purposes.