In his groundbreaking Aspects of the Theory of Syntax (1965), Noam Chomsky first made explicit what is now arguably the dominant view concerning the aims and objects of linguistic inquiry. Rather than studying the sounds and inscriptions that we produce and comprehend, or the social conventions that govern linguistic usage, Chomsky argued that the primary target of linguistic theorizing must be the “tacit knowledge” that underlies every competent speaker-hearer’s linguistic competence. On a common and natural (though not uncontested) interpretation, this knowledge is encoded in mental representations of grammatical principles.
Chomsky’s “cognitivist” conception of linguistic inquiry generated a great deal of excitement and debate in many areas of cognitive science, and particularly in philosophy. Philosophers immediately questioned the coherence of the notion of tacit knowledge. What sort of knowledge can be neither learned nor taught, neither spoken nor recollected (Quine, 1970; Stich, 1971)? And why should we count as knowledge something that doesn’t play a role in our everyday practical and theoretical inferences? (Most New Yorkers know that there’s great pizza in Brooklyn. Do they really know that only the features on functional heads are subject to parametric variation?) Finally, what prevents us from seeing the linguists’ construction of grammars as simply an effort to delineate the class of grammatical sounds and inscriptions (Soames, 1984; Devitt, 2006)? Why must linguistics traffic in any psychological notions at all?
A number of philosophers have defended Chomsky’s approach by offering clarifications of the concepts of “mental representation” and “subpersonal mechanism” (Davies, 1989; Peacocke, 1989; Rey, 2003). But it’s safe to say that there is, at present, no consensus in the cognitive science community—and certainly none amongst philosophers—concerning the viability and conceptual coherence of Chomsky’s program. Detailed analyses of many languages have, of course, been produced, and much more is now known than ever before about language acquisition and real-time processing. But the philosophical underpinnings of the project, particularly the notions of psychological reality, tacit knowledge, and mental representation continue to defy generally acceptable explication.
In addressing these issues, I take as my point of departure Michael Devitt’s recent contribution to the debate. In Ignorance of Language (OUP, 2006), Devitt challenges nearly every aspect of Chomsky’s theoretical framework. He argues that the objects of linguistic theory are public languages (construed “nominalistically” as sounds and inscriptions) and that the grammars emerging from formal syntax tell us only about the rules that govern those languages, not about the cognitive mechanisms involved in acquiring or using them.
My own project can be seen as a critical response to both Chomsky’s and Devitt’s positions. The aim is to construct a philosophy of linguistics that combines what I take to be the best-supported aspects of each. On broadly philosophical issues—the epistemological status of syntactic theory, its ontological commitments, and its methodology—I side, in large part, with Devitt. Where we part company is over the psychological reality of syntactic rules and principles. Devitt (2006) argues that there is no evidence for the claim that successful grammars are represented, or even embodied, in the human mind/brain. Moreover, he entertains the view that comprehending and producing language is “a fairly brute-causal associationist process, rather than a process involving representations of the syntactic properties of linguistic expressions” (220). Similar views can be found in linguistics, artificial intelligence (NLP, both classical and connectionist). Drawing on a wide range of neurocognitive and behavioral studies, I marshal several lines of evidence agains such views.
Importantly, my arguments for the psychological reality of syntactic rules and principles make no appeal to the controversial innateness hypothesis (Chomsky, 1965). My view is not, therefore, hostage to the outcome of ongoing debates between nativist and empiricist approaches to language acquisition. My reliance on contemporary work in psycholinguistics extends, in large part, to studies of real-time language comprehension. This allows me to draw on established, up-and-running computational models in giving substance to technical notions like ‘tacit knowledge’, ‘procedural rule’, ‘cognitive module’, and ‘subpersonal state’. This is especially evident in chs. 8-9, where I undertake a historical survey of the co-evolving fields of formal syntax and parsing theory.
Before delving into the details of the arguments, let me set out my main conclusions, in as nontechnical a fashion as the subject matter permits.
- Contrary to the dominant Chomskyan position, the formal syntactician is not best seen as engaging in a distinctively psychological inquiry. Even the successful grammars emerging from syntactic theorizing cannot be assumed without argument to be psychologically real. A great deal of support from psycholinguistic research must be marshaled before we can make substantive claims about the psychological role of a successful grammar. This is true even if we equate the success of a grammar with what linguists call “explanatory adequacy”.
2. The explanatory adequacy of a grammar is best construed in the following neutral fashion: a grammar of some particular language is explanatorily adequate if it fits well with the maximally general, simple, and unified theoretical coverage of all human languages. Construed in this way, achieving explanatory adequacy is desirable whether or not the cognitivist conception of linguistics is true. Although explanatory adequacy is often seen as a proprietary notion within the cognitivist framework, I argue that it is in practice assessed independently of psychological data. The pursuit of explanatorily adequate grammars can and should be motivated by general methodological principles that have nothing in particular to do with the cognitivist conception of linguistics.
3. Reflection on the actual practice of both syntacticians and language acquisition theorists reveals their commitment to the reality of public languages. Although this commitment is often disavowed, there is simply no other way to make sense of the persistent and ineliminable references to public languages in the canonical texts of formal syntax and acquisition
4. In practice, the primary target of syntactic theorizing is not an idealized idiolect—what Chomsky calls an “I-language”—but, rather, an idealized speech community and its public language. Public languages have some of the aspects of what Chomsky calls “E- languages”, in that they are collections of sounds, marks, muscle movements, and the like. But the classes of these items that constitute one or another public language are delineated on entirely theory-internal grounds, without reference to political organizations or prescriptive goals. The individuation conditions on E-languages are, at present, not precise, and require idealization away from messy variation within a community. But this is no less true of I-languages, whose individuation conditions are, at this stage of inquiry, only dimly understood. We should not expect to provide precise individuation conditions for either E- or I-languages in advance of sustained inquiry.
5. The public E-languages that syntacticians study are not “abstract objects”, in the philosopher’s That is, they are not causally inert things that exist outside of spacetime, and they are not uncountably infinite. The aims of linguistics are in no way furthered by thinking of language in this “platonist” fashion. Linguists’ claims concerning the infinitude of language do not signal an ontological commitment to abstract entities. Rather, they reflect the modal force of linguistic generalizations, as well as a principled idealization away from mortality, memory constraints, and the like. Nor is it useful to think of the truths of linguistics as being discovered through “nonempirical rational intuition” (Katz, 1985; 2000).
6. Though the target of syntactic theorizing need not be a psychological state or mechanism, there is in fact a very strong case to be made for the claim that some class of grammars is psychologically real. The mind/brain of each competent speaker-hearer employs a specific set of syntactic rules or principles, which play an important role in the cognitive processes underlying language comprehension.
7. Comprehending linguistic input requires constructing and manipulating mental representations of its syntactic Any psychological or computational model that eschews such representations will be unable to account for a vast range of behavioral and neurocognitive data. The models thus far proposed in this genre exhibit a striking pattern of failure.
8. Representations of incoming linguistic structure are not best seen as beliefs, thoughts, or even perceptual judgements. They differ in kind from all of the psychological states that figure in commonsense explanations of behavior. Unlike those states, the mental representations posited by the psycholinguist are invariably nonconscious, inferentially insulated from most beliefs, and inexpressible in speech. Moreover, ascribing such states requires making far less demanding normative assumptions. Whereas erroneous judgments and inferences leave a person open to rational criticism, misrepresentations in language processing are best seen as malfunctions in a dedicated subpersonal mechanism within a person’s mind/brain.
9. Constructing and manipulating representations of syntactic structure requires either representing or embodying a grammar. Any psychological or computational model that neither represents nor embodies a grammar will be incapable of replicating human performance and effectively coping with the massive ambiguity of linguistic inputs. This is true regardless of what additional machinery the model requires, including statistical information about the frequency of various kinds of input.
10. The notion of embodiment is intermediary between the notions of full-blown representation and mere “conformity to a rule”. Embodiment is distinct from both of these, in ways that are open to empirical
11. The hypothesis that grammatical rules or principles are embodied is more parsimonious than the hypothesis that they are represented. Moreover, we have, at present, no principled grounds for asserting that grammars are explicitly represented, rather than embodied, in the human mind/brain.
12. We should tentatively conclude that a common claim in generative linguistics—viz., that grammars are represented in the minds of competent language users—is either a conflation of the notions of embodiment and representation, or simply an attractive but as-yet-ungrounded hypothesis. Although we confidently assert that the syntactic structure of linguistic input is explicitly represented in the course of language comprehension, the claim that the rules or principles of a grammar are likewise mentally represented, rather than embodied, must await significantly more fine-grained measuring techniques and sophisticated experimental paradigms in cognitive neuroscience.
In the posts that follow, I’ll sketch my arguments for several of these conclusions. I’ll begin with philosophical issues—metaphysics, epistemology, methodology—and then switch gears to review the results of psycholinguistic and computational research.
Chomsky, N. (1965). Aspects of the Theory of Syntax, Cambridge, Mass.: MIT Press.
Davies, M. (1989). “Tacit Knowledge and Subdoxastic States,” in Epistemology of Language, A. George (ed.), pp. 131-52.
Devitt, M. (2006). Ignorance of Language. Oxford: Oxford University Press.
Katz, J. J. (ed.) (1985). The Philosophy of Linguistics. Oxford: Oxford University Press.
Katz. J. J. (2000). Realistic Rationalism. Cambridge: MIT Press.
Peacocke, C. (1989) “When is Grammar Psychologically Real?” in Reflections on Chomsky, A. George (ed.), Oxford: Basil Blackwell, pp. 111- 130.
Quine, W. V. (1970). “Methodological Reflections on Current Linguistic Theory” Synthese 21, pp. 386-398. Reprinted in Semantics of Natural Language D. Davidson and G. Harman (eds.).
Rey, G. (2003), “Chomsky, Intentionality and a CRTT,” in Chomsky and His Critics, Antony and Hornstein (eds.), Wiley-Blackwell.
Soames, S. (1984). “Linguistics and Psychology”. Linguistics and Philosophy 7, pp. 155‑79.
Stich, S. (1971). “What Every Speaker Knows,” The Philosophical Review, 80 (4), pp. 476-496.