Faculty Disputes
Abstract
Jerry
Fodor, among others, has maintained that Chomsky’s language
faculty hypothesis is an
epistemological proposal, i.e., the faculty comprises propositional
structures known (cognized) by the speaker/hearer. Fodor contrasts this notion
of a faculty with an architectural (directly causally efficacious) notion of a module. The paper offers an independent
characterisation of the language faculty as an abstractly specified
non-propositional structure of the mind/brain that mediates between sound and
meaning - a function in intension that maps to a pair of structures that determine
sound-meaning convergence. This conception will be elaborated and defended
against a number of likely complaints deriving from Fodor’s faculty/module
distinction and other positions which seek to credit knowledge of language with an empirical or theoretical
significance. A recent explicit argument from Fodor that Chomsky must share his
conception will be diagnosed and the common appeal to implicit knowledge as a foundation for linguistic competence will
be rejected.
“Some
questionable terminological decisions also contributed to misunderstanding.”
— Chomsky,
1986, pp.28-9.
1: Introduction
Jerry
Fodor, 1975, 1981a, 1983, 2000, has long maintained a particular understanding
of Chomsky’s hypothesis that linguistic competence is principally subserved by
a language faculty. This faculty
hypothesis, according to Fodor, is an epistemological
proposal about what speaker/hearers know
- that propositional knowledge which essentially enters into the explanation of
language acquisition and the maintenance of mature performance. What Chomsky’s
faculty hypothesis is not, claims Fodor, is a proposal about the architecture of the mind/brain, where
such a thesis offers a causal explanation of how speaker/hearers acquire and maintain a knowledge of language
and put it to use. In the works cited, this reading has been offered as more or
less non-tendentious exegesis, with scant support from Chomsky’s texts; more
recently, Fodor, 2001, presents an argument on Chomsky’s behalf that putatively
shows that Chomsky must cleave to the
epistemological view.
The sequel will agree with Fodor in his
negative claim - that the faculty is not a causal mechanism. Here, there is
some range of agreement within the literature, although significant differences
persist (see, e.g., Higginbotham, 1987; George, 1989; Peacocke, 1989; Davies,
1989; Matthews, 1991; Segal, 1996; Smith, 1999, Knowles, 2000). Where I depart
from Fodor, as well as most of the above, is in my further proposal that
Chomsky’s understanding - one we should share - is not epistemological either;
an alternative will be offered and defended.
My focus will be on Fodor’s fairly strict understanding according to
which a speaker/hearer’s knowledge of language is constituted by internal
propositional states, a species of ‘knowledge-that’. The negative conclusions
reached, however, will be seen to impact equally on more relaxed construals of
the epistemological thesis that don’t carry assumptions as to the character of
the relevant internal states.
2: Fodor’s Distinction
The term faculty, and it close associate, module, have acquired an almost Humpty
Dumpty status, whereby they are used to designate, willy-nilly, whatever the
particular author is interested in defending or attacking (see, e.g., Segal,
1996; Garfield, et al., 2001). Fodor,
2000, p.57, takes his 1983 use of Chomsky’s ‘modular’ jargon to be responsible
for much of the confusion he is presently seeking to rectify. For what it’s
worth, I’ve always found Fodor clear on the matter; further, few thinkers in
the field worthy of attention have been guilty of reading Chomsky’s use of
‘module’ (used interchangeably by Chomsky with ‘faculty’) as synonymous with
Fodor’s use. What has been confusing is the willingness of so many to many to
read Chomsky through Fodor’s positive epistemological reading, and then to find
Chomsky obscure. To ward off yet more needless confusion over nomenclature, let
Fodor’s distinction be the following:
Faculty, in the epistemic sense (= faculty, smpliciter): That (usually innate) dedicated information or
knowledge which a system possesses such that it acquires and maintains a
competence in some area of cognition (language, mathematics, social relations,
etc.)
Faculty, in the architectural sense (= module): A computational component of
the mind/brain that is domain specific,
in the sense that it outputs ‘answers’ specific to some domain, and is informationally encapsulated, in the
sense that the information the computations are defined over such that they are
domain specific is restricted to a fixed database (plus input) represented by
the module. The module’s computations over the database causally account for a
system’s acquisition and maintenance of competence in the given domain
(language parsing, etc.)
As
intimated here, it would not be a gross mistake to think of Fodor as proposing
that what Chomsky calls faculties just
are the proprietary databases of modules (Fodor, 2000, p.57). Officially,
Fodor does not insist on this understanding, although he does think that there
is a language module (in effect, a parser) with a proprietary grammatical
database. Fodor, 1983, p.9, well recognises that the epistemological thesis - a
claim about what is known - is
logically independent of the architectural thesis - a claim about the physical
realisation of what is known. That is, the claim that a speaker/hearer stands
in an epistemic relation to her language faculty does not entail any particular
account of how the faculty is realised in the mind/brain architecture. Still,
Fodor does think that the database of the language module just is what Chomsky
calls the language faculty (see n.2). For our purposes, what is of primary
interest is the distinction itself, not Fodor’s further (and independent)
proposal about how faculties stand towards modules. Following Fodor, 1983,
p.6-7, 2000, p.11, we may cast light on the distinction by considering the
issue of poverty of stimulus.
That a cognitive capacity is acquired in
the face of a poverty of stimulus from the domain of the capacity is understood
to militate decisively for the innate basis of the capacity. Indeed, such is
the felt association between poverty of stimulus and innateness that Prinz,
2002, pp.193-4, suggests that ‘acquired under poverty of stimulus’ may serve as
an operational definition of ‘innateness’. Similarly, Cowie, 1999, pp.46-7,
argues against a developmental model of innateness on the basis that innateness
just is what sound arguments from the
poverty of stimulus entail, and no such arguments entail any particular
developmental account (cf. Samuels, 2002). I don’t think that claims of this
strength possess the least warrant (Collins, forthcoming). Even so, familiarly,
arguments from poverty of stimulus have been central to Chomsky’s nativist
claims since the 1960s: a poverty of linguistic stimulus (inter alia) shows that a learning theory model of language
acquisition is empirically inadequate; a model that does explain (uniquely?)
language acquisition in the face of a poverty of linguistic stimulus is one
which attributes to the child a biologically endowed range of language specific concepts. The details
of such reasoning and the orbiting controversy need not detain us. The most
simple way of making the point is via what Jackendoff, 1993, p.26, has called
the paradox of language acquisition.
The professional communities of linguists and psychologists have all the data
they desire on, say, English, certainly much more than any child has ever
enjoyed. They also pool their mature intellects, have equal data on many other
languages, and design subtle experiments and conduct longitudinal studies over
many years. Yet they still can’t work out the principles, rules and concepts
that constitute competence with English which the normal child masters by the
time she is five! So, it is not so much that the child’s data is poor according
to some absolute standard (as if such a notion makes sense); rather, if the
child were relying on data, then her acquisition of language would be
miraculous. The acquisition ceases to be at all miraculous if we understand the
child to possess innate resources which already determine most of what is to be
acquired. On this view, the child’s task is so ‘easy’ precisely because she
isn’t dependent on rich data, while the data inundated scientist’s task is so
hard precisely because he is trying to work out what the child already knows
independent of data or general principles defined over it.
Now, and this is Fodor’s point, however
such a style of argument may be evaluated, by itself, it only seems to militate
for the hypothesis that the child possess some innate knowledge (rules, concepts, etc.) specific to language. The
argument leaves entirely open the question of whether there is a dedicated
computational device that has access to that knowledge alone, is encapsulated with respect to it. The
argument might convince us that, say, the principles of binding theory are innate, but it shouldn’t convince us that there
is a module which represents such principles in its database. As far as the
argument goes, it might be that cognition is served by a general computational
device that has free access to different stores of innate knowledge -
faculties. In short, poverty of stimulus considerations tell us that some
knowledge is innately represented; they don’t tell us how the knowledge is
represented or processed.[1]
So, arguments for faculties (knowledge) are not ipso facto arguments for modules. It doesn’t follow, of course,
that Fodor reads Chomsky aright merely by dint of the latter’s commitment to
arguments from the poverty of stimulus. All that does follow is that such
arguments cannot be used in direct defence of a modular thesis (see §5).
The sequel will not be directly
concerned with ‘modularity’ theorising in general, nor, in particular, with
whether either of Fodor’s notions, or the alternative to be presented, is apt
to make sense of the claim that the mind is ‘massively modular’, i.e., all cognition is sorted into domain
specific modules. Fodor’s thesis, concomitant with his distinction, is that
modules are input/output devices associated with specific channels of
transduction such as vision, olfaction, hearing, etc. Thus, for Fodor, whether
there are ‘central’ modules is at best moot; the thesis that it’s all modules he considers to be
virtually a priori false (Fodor,
2000, chp.4). Much of Fodor’s motivation for his distinction is to bring
clarity to these issues. The sequel’s concern is with the language faculty
alone. Still, it bears emphasis now that if Fodor does think that all modules
are peripheral, in the sense of processing transduced information, and the
language faculty is a database for a
module, then the language faculty will essentially be in the service of a
parser.[2]
We shall come back to this (see §5).
3: The Language Faculty, in Chomsky’s Sense
Notwithstanding
any diagnostic virtues Fodor’s distinction enjoys, it will be argued that
Chomsky’s notion of the language faculty eludes Fodor’s epistemic/architectural
distinction. Let us characterise this conception as follows:
The language faculty, in Chomsky’s sense: The
language faculty is a function in intension whose specification describes an
aspect of the human brain. The function is from selections from a lexicon to
infinite pairs of structures - <PF, LF> - that determine the respective
forms of merged lexical items as they interface with external systems governing
sound articulation and intention/conceptuality. The convergence of the pairs
internal to the faculty accounts for the robust sound/meaning association upon
which our linguistic performance is based.[3]
The
following is a typical statement of Chomsky’s: “[T]here is a special component
of the human brain (call it ‘the language faculty’) that is specifically
dedicated to language. That subsystem of the brain (or mind, from the abstract
perspective) has an initial state which is genetically determined, like all
other components of the body” (Chomsky,1996, p.13).[4]
Further: the language faculty is “dedicated to the use and interpretation of
language… [it] assum[es] states that vary in limited ways with experience.
Interacting with other systems (cognitive, sensorimotor), these states contribute to determining the sound and
meaning of expressions” (Chomsky, 2000a, p.168). And this integration is what
makes the language faculty linguistic: “Each linguistic expression [PF, LF
pair] generated by the I-language [a steady state of the language faculty]
includes instructions for performance systems in which the I-language is
embedded. It is only by virtue of its integration into such performance systems
that this brain state qualifies as a language” (Chomsky, 2000a, p.27).
Similarly, Chomsky (e.g., 1986, p.3; 2000b, p.54) explicitly states that universal grammar (UG) - the initial
state of the language faculty - can be innocently understood as the language
acquisition device. The reasoning
here is that, on the assumption that the other systems with which the faculty
interfaces are autonomous and in working order, the development of the faculty
amounts to the acquisition of language, the proper integration of the
performance systems.
Some words of clarification are in order.
Firstly, the very idea of the
language faculty is a methodological idealisation. Properly speaking, the
target of our investigation is whatever horrendously complex ensemble of
systems accounts for the observed linguistic phenomena. If one wishes to
preserve the ‘faculty’ label for this manifold, there can be no a priori complaint. The sequel will
follow the tradition, however, and understand the faculty to be that system
that is dedicated to ‘syntax’ and relates to external performance systems which
serve our use/articulation of language. Here, ‘syntax’ is not a simple notion of ‘form’ or ‘rules’ construed to be
independent of words with their phonological and semantic properties. Rather,
‘syntax’ may be seen as a systematic relation between lexical items and complex
structures built from lexical items. More precisely, ‘syntax’ (= the faculty)
covers a lexicon and operations defined over selections from it that unfurl, as
it were, phonological and semantic features inherent in the lexical items in
such a way that structures are generated that may input to systems that put
such structures to use in our production and consumption of language. In this
sense, the faculty at a steady state is the result of the development of UG,
whose parameters of variance are perhaps restricted to the morphological
idiosyncrasies of the acquired lexicon.[5]
This system stands in contrast to the so-called external systems with which it
is associated inasmuch as while the faculty’s output necessarily answers to the conditions the systems impose upon it
(this is the core idea of the minimalist program - Chomsky, 1995), what use the
systems make of its outputs is not itself encoded in UG or, perforce, the
faculty’s steady states. Whether this perspective carves the ensemble of
systems at its salient joints is an empirical question, not a conceptual one;
that is, whether ‘syntax’ itself marks a theoretically salient distinction is
to be determined. We maintain the
assumption so long as doing so continues to be theoretically fecund. If
recalcitrant data arises, or a simpler model consistent with the facts is formulated,
etc., then we are rationally free to alter our assumption (see §5).
Secondly, even though we have said
that the faculty is a brain state, the faculty explanation of how sound/meaning
pairings are attained and maintained is not a causal cum neurological
explanation of how one is able to order a coffee or write a paper on the
language faculty, still less is it predictive of such behaviour. The faculty
may be viewed as a set of sui generis
conditions met by the operation of a ‘normal’ brain. The conditions are sui generis in the sense that they
specify an aspect of a complex system that is not visible minus such
conditions. The aspect at issue is simply the recursive/structured integration
of sound and meaning in human cognition.[6]
The idea here correlates with the thesis, going back to chapter 1 of Aspects (Chomsky, 1965), that the
faculty accounts for competence not performance. It should be said that the
correlation is not precise, and the distinction itself has never been
absolutely clear. Still, the distinction does have a real point and can, I
think, be usefully recast in my favoured terms. As an initial characterisation,
we may say that the faculty counts as competence just because it explains how
there is a systematicity between independent systems of sound and meaning; to
have access to such systematic pairings is just what it is to be linguistically
competent. Thus, rather than explaining performance, acts of speech or thought,
the faculty hypothesis explains how performance is so much as possible. This,
note, does not entail that competence is in any sense restricted to, or
exhausted by, the provision of performance. The generation of structures
designed to pair sound and meaning might well exceed the bounds of any
performance parameters. Indeed, we may say that the very nature of the
components conjoined in language use require a third system whose principles
are not dependent on the use we make of them in performance: we need a third
(competence) system to integrate otherwise separate systems (see §5).
Thirdly, the faculty is further
abstracted from the causal structure of the brain in that it is a function in intension. For example, we may describe
a given Turing machine as computing some number or as working out the GDP of
Luxembourg. As far as the computability of the function (in extension), qua ordered pair, is concerned, it does not make any difference
what description we use: the same numbers in will give us the same number out.
It makes all the difference in the world, however, if, rather than being given
the function, we have an extant system and our job is to figure out what
function it is computing given its outputs.
In essence, here we are interested in what distinctions and properties
the system is sensitive to. Assuming Church’s
Thesis, we know that the function (in extension) will be computable in any
number of ways in any number of formats, but this does no explaining for us. We
want to know the function under that
description (in intension) which
will explain how the system can produce an independently describable output and
keep to any constraints which issue from the nature of the system and the
character of its development. So, turning to language, we want a function that
produces the observed systematic and highly specific structures that are
realised by meaning/sound pairings. In other words, what we don’t want is a
function which merely generates symbol strings. The point here goes back to the
heart of the argument of early generative grammar (Chomsky, 1955/75, 1957).
Relations hold between linguistic structures that are not tractable if the
structures are construed as immediate
constituent symbol strings (e.g.,
active vs. passive, scope ambiguities, declarative vs. interrogative, etc.).
Thus, the computation must be sensitive to items under whatever descriptive
terms make sense of these relations which constitute our competence with the
sentences. The function must also be such that the system may fixate on it via
partial and degraded data; it would seem, for example, that it can’t be one
which merely works on statistical regularities. So, not any old function that
determines sound/meaning pairs will do; we want one which will explain an
independently observed systematicity between and within sentences.
Fourthly, this conception of the
language faculty is not intended to be one Chomsky or others have always held,
as if there were such a notion. My construal is intended to be one which is
flush with many aspects of Chomsky’s developing position (if not every sentence
he has ever written), especially in its course to ‘internalism’. Further, it is
independently coherent, as well as being distinct from the epistemological
reading. Still, a slight exegetical digression will serve us later.
The notion of a grammar presented in Syntactic Structures (Chomsky, 1957) -
in fact, undergraduate lecture notes - and the unpublished 1955-56 manuscript
of Chomsky, 1955/75, which in essence prevailed up to the advent of the principles and parameters approach (e.g., Chomsky, 1981, 1986), appears to fit pretty much with Fodor’s
epistemological reading. Perhaps the most revolutionary feature of Chomsky’s
early work was its “methodological approach”, the idea that a grammar for a
language should be evaluated much like a scientific theory. The job of linguistic (meta-) theory is to specify a set of
theories (grammars) that recursively assign a set of descriptions (at various levels) to the sentences of a language,
where the descriptions cohere with the speaker/hearer’s intuitions over the
language. The assignment is recursive in that it is infinitely generative
(“predictive”) of descriptions of the sentences of the language. The
psychological analogue of this approach - not broached in the earliest work -
is that the general theory details what grammars a speaker/hearer can
acquire/know which in turn explain the intuitions to which the explicit
grammars answer, i.e. the general theory becomes an account of the ‘language
acquisition device’ that maps from data to a highest valued grammar
“predictive” of the data. Thus, rather than grammars being operationally
definable from the corpora to which human children are typically exposed, they
are theories of the data, internalised by children, that posit unobservables
and abstract rules (as made available by the LAD) to capture hidden
generalisations in the respective target languages, which in turn account for
the structure of, inter alia,
linguistic ‘creativity’ (unbounded and continuously novel understanding and
production, uncaused by features of
the prevailing situation but appropriate
to it and coherent in the context).[7]
Fodor, with many others, appears to think
that Chomsky’s position, although developed in all sorts of other ways, still
retains this kind of methodological/psychological split. This is not so.
It seems to me that in Chomsky’s early
thought the methodological and psychological approaches were at least equal;
the former prevailed in the published work, the latter was implicit, but soon
to emerge (Chomsky has clearly stated that psychological issues of innateness,
explanatory adequacy, UG, etc. were important to him in the mid 1950s, in the
“immediate background”, “taken for granted”, but their articulation would have
been “too audacious” (Chomsky, 1955/75, p.13/35/37). It is worth noting in this
regard that the famous opening chapter of Aspects
(1965) was essentially written in the late 1950s (Chomsky, 1982, p.62).) What
was crucial to Chomsky, and trivially shared by the two approaches, was a commitment
to explicitness, which was perhaps
best expressed methodologically. (An explicit
grammar is one which is infinitely generative of structural descriptions and
does not rely on speaker/hearers’ intuitions.) Such is why, I think,
psychological factors were certainly not so central for Chomsky that they
formed clear constraints on the adequacy of grammars, even if this was just
because he couldn’t see how the two would properly mesh (Chomsky, 1955/75,
p.116-7). So, while Chomsky did think, contra
the behaviourists/empiricists, that grammars were psychological kinds - the
kind of thing speaker/hearers know -, one doesn’t find a notion of knowledge or
explanatory adequacy that imposes any particularly tight constraint on the
nature of the human mind such that it gives rise to the grammars (only in the
famous Skinner review does Chomsky, 1959, first raise such issues and there
merely sketches a cognitive model as a potential alternative at the end of the
article after the Skinner model had
been refuted.[8]) This
reading is not at all curious when it is appreciated that explicit cognitive
considerations were unnecessary for the refutation of behaviourism/empiricism
in linguistics. The refutation proceeded simply on the basis of grammars being explicit, for their very abstractness
already sounded the death knell for behaviourism and its operational discovery
procedures; the structure of the grammars was not recoverable from the
‘visible’ properties of corpora. Further, as intimated above, the introduction
of an explicit psychological construal of the theory (Chomsky, 1962, 1964,
1965) was, in essence, a simple recasting of already extant methodological
notions. The “external” condition of descriptive adequacy and the “internal”
condition of explanatory adequacy of the 1960s work were simply interpretations
of the earlier notions of descriptive conditions, which pertain to particular
grammars, and generality conditions, which situate such grammars within general
linguistic (meta-) theory (e.g., compare Chomsky, 1955/75, p.108-9, 1957,
pp.49-50 with Chomsky, 1964, pp.62-3).
Thus, we have a natural concordance
between the methodological and psychological approaches: the
speaker/hearer’s relationship to her
grammar vis-à-vis data could usefully
be cast as being the same as the scientist’s relationship to her theories vis-à-vis data inasmuch as both involve
theory construction whose hypotheses are not inductively determined (see, e.g.,
Chomsky, 1957, pp.49-50; 1959, pp.574-8; 1962, pp.528-9/535; cf. Katz, 1966, p.275).
All in all, the general impression
that was to persist was that grammars, as explicitly specified theories of
languages, could (and should) be understood as objects of knowledge, where, in
Higginbotham’s 1991, p.556, words, it is a “rational achievement” of the
speaker/hearer to acquire her grammar, a theory of her language (Higginbotham
appears to endorse this model; cf. Fodor, 1983, p.5). It appeared to be such an achievement precisely because specifying a
theory of a language - a grammar - was one thing, specifying the psychological
facts which enabled a speaker/hearer to know/acquire such a theory was another
thing (George, 1989, p.91 explicitly endorses this model). When psychological
concerns came to the forefront in the early/mid-1960s, and the old notions were
recast in the ways indicated, the epistemological terminology of the
methodological approach became less clear cut. For instance, the evaluative
notions (e.g., simplicity) became internal to general linguistic theory and
fell under general psychological constraints given that it was the child not
the theorist who was doing the evaluating. Such issues of learnability and
grammar selection eventually led to the triggering model - as opposed to an
inference from evidence model (which was always a tentative hypothesis, not an
a priori claim) - of the principles and parameters approach and its concomitant
internalism (see §§5, 6). With this approach there appears to be nothing left
that is theoretical about the language faculty, at least not in any
epistemological sense. This is reflected in the eschewal of the very notion of
a child learning a language in favour
of the notion of a child’s linguistic growth:
to acquire a language is to undergo a specific internal development, it is not
to acquire knowledge about - a grasp
of - an external object. Similarly, it is a mistake to think of the poverty of
stimulus considerations in favour of UG as an instance of the general
(Humean/Goodmanian) inductive underdetermination of theory by evidence, as if the
child were seeking to find justification for its ‘theory’. One key difference
is that the child’s data appear not to confirm the UG generalisations at all,
but do appear to confirm simpler but false ones; the ‘data’ act as a trigger
for the realisation of some principle, not as evidence for it. There is nothing
rational about acquiring a language.
The point of this little bit of
history is to highlight that after the development of P&P it really became
transparent that grammars and/or languages could no longer be sensibly thought
of as independent ‘objects’ of knowledge. In bald terms, grammars, and so
languages, ceased to be understandable as ‘things’ which speaker/hearers know;
they are simply states of the speaker/hearer. Otherwise put, the only operative
notion of language is the grammar ‘in the head’ - the I-language - which is not
represented as an external object in
any epistemological sense whatsoever, nor, a
fortiori, does the internal state
represent such an object - there is nothing to represent.[9]
The difference between the represented
or known and the representation or state of ‘knowledge’ is, at best, a formal
reflex of the notion of ‘representation’: a representation requires a
represented. But even the formal distinction is misleading. If the faculty is
not represented, then there are no
causally efficacious representations in the appropriate substantive sense, for
there is nothing for them to represent. Of course, there is no harm in
retaining the term ‘representation’ along, indeed, with its reflex, but it has
a “technical sense”, marking levels of interface between an ensemble of
internal systems (Chomsky, 2000a, pp.175-6). Indeed, Chomsky’s use of
‘representation’ has always been technical; it derives from concatenation
algebra, not philosophical theories (see Chomsky, 1955/75, p.105). So, we may
call a LF-structure a representation because (inter alia) it meets conditions internal to the faculty imposed
upon it by other systems. The structure isn’t about such conditions; it can’t go right or wrong about
them. The whole arrangement is ‘syntactic’, internally specifiable (cf.
Jackendoff, 2002). Chomsky, 1986, p.28-30, has suggested that the “systematic
ambiguity” between the methodological and psychology approaches to
grammar/language has been the root of much confusion. The internalism that resolves the ambiguity will be returned to at
length and be shown to be inconsistent with Fodor’s general argument that there
must be a representation/represented distinction.[10]
Let us now consider some ‘Fodorian’
complaints that might be expected. After that, we shall turn to his argument
that his epistemic understanding must be correct.
4: ‘Knowledge of Language’
The first
issue which needs to be broached is the simple fact that Chomsky has for many
decades freely appealed to knowledge of language, which would prima facie suggest that Chomsky has an
epistemic understanding of the language faculty. Does Chomsky not mean what he
says?
The negative aspect of this central
locution bears emphasis. To say that a subject knows a language is to signal
that linguistic capacity is not
constituted by a set of dispositions or a practical ability or a capacity to
communicate. In short, the knowledge locution is, apart from anything else,
intended to make clear that competence
not performance is at issue (see
especially Chomsky, 1965, pp.8-9; 1975, pp.22-3.) So much should be agreed upon
by all. However, under the distinction of know-that/know-how, where the latter
is read in terms of capacities or dispositions, it seems obligatory to read
linguistic competence as a species of knowledge-that, albeit one which
potentially departs in various ways from our intuitive notion. Chomsky, though,
has been consistently leery of the know that/know how distinction, mainly because he doesn’t see
dispositions or other behavioural notions as possessing any explanatory value.[11]
Thus, it matters not at all to linguistic theory whether the speaker/hearer
knows-that or knows-how a language, or ‘has a language’, ‘speaks a language’,
etc. so long as what is being picked out
is a mental structure. To this end Chomsky, 1975, pp. 164-5 (cf. 1980,
pp.69-70), coins the term cognize to
divorce his appeal to knowledge from those properties that have traditionally
differentiated know-that from know-how: especially ‘justification’ and
‘conscious access’. Still, Fodor assumes, explicitly contradicting Chomsky (see
n.11), that “cognize” is a species of know-that in a substantial propositional
sense. As Fodor, 1983, pp.4-5, puts it: “what is innately represented should
constitute a bona fide object of propositional attitudes; what’s innate must be
the sort of thing that can be the value of a propositional variable in such
schemas as ‘x knows (/believes/cognizes) that P” (cf., Fodor, 2000, pp.10-11.)
While Fodor is certainly right that ‘cognize’ may serve as a propositional attitude verb (Why
not? It can have any features we wish to preserve from ‘know’), it is not
intended to be a notion which contributes in any substantial sense to
linguistic theory.
The language faculty is cognized in the sense that it is a state
of the mind/brain that is separate from, although stands in an explanatory (not
causal) relation to, performance, much as, notwithstanding obvious differences,
knowledge of a precise score explains one’s competence with Beethoven’s Hammerklavier, a knowledge which is
retained even if one develops a debilitating arthritis. The precise thesis,
then, does not essentially involve any epistemic terms, or propositional
attitudes. The ‘knowledge’ at issue is simply whatever autonomous core concepts
and principles structure one’s performance and judgements, although, again, the
extent to which it outstrips performance is an empirical matter. In particular,
then, it does not follow that the cognized falls under any broader cognitive
state or process such that we may sensibly speak about it being an object of
knowledge. One difficulty in seeing this is that the information the language
faculty generates is easily conflated with the faculty itself; indeed,
Chomsky’s ‘cognize’ is explicitly intended to cover both notions, but this just
shows how little it has to do with our intuitive propositional attitude notion
of knowledge. This becomes transparent with Chomsky’s more recent understanding
of the faculty.
It seems natural to think of the language
faculty of S as representing S’s language, or at least its
grammatical structure: the faculty generates grammars that are true of languages. This is Fodor’s
model: grammars are “innately specified propositional contents… truths” (1983,
p.7; cf., 2000, pp.95-6). But what could such contents be possibly true of?
There is, as it were, nothing to get
right. Languages are not external objects we can go right or wrong about.
Fodor, 2000, chp.5, suggests that what we might be getting right is the
language of our conspecifics, those from whom we learn our language. But this
cannot be right. The language we end up with will reflect, along certain narrow
parameters, our initial experience, but we do not end up representing (or misrepresenting) that experience. The complex
grammatical principles and features are just not in the data; that’s the very
point of the poverty of stimulus considerations. The only thing there is to
represent is the language faculties of our conspecifics, or at least UG, but
that is not something which we represent on the basis of another speaker, it is
simply shared qua species trait. The
problems here arise from the very idea that language is something which may be represented,
something which exists external to us which we can get right or wrong, have
true beliefs about. In essence, this model is what Chomsky rejects under the
label of ‘E-language’, ‘E’ for an extensionally
individuated set of external objects.
As go E-languages, so go grammars as things which might represent them.
As far as scientific investigation is
concerned, as opposed to conceptual reflection, Chomsky, 1986, 2000a,
understands ‘language’ to be the internal finite
generative procedure that accounts for our infinite competence. Chomsky dubs
languages, so construed, I-languages:
the internal steady states of our individual language faculties (UG is
separately designated as the initial state, but any state of the faculty may be
construed as an I-language). In effect, an I-language just is a particular
function in intension, a variation on
a biologically determined theme. Thus, an I-language is not something one may
be right or wrong about; it is a structure of the mind, not an independent object
the mind represents.
Now it is perfectly sensible to think
of such I-languages as providing us with infinite knowledge-that (in the colloquial sense) about, say, relations of
co-reference, elliptical antecedence, distribution of argument roles, etc.
(Chomsky, 1975, p.165; 1980, p.71/93; 1986, p.269). For example, it is our
language faculties that determine (in part) our knowledge that in Bill expects to leave by himself it must
be Bill who expects to leave (by himself), but when we embed the sentence as a wh-complement - e.g., Harry wonders who Bill expects to leave by
himself - we know that it’s not necessarily Bill who expects to leave, but
someone or other (who might, accidentally, be Bill). We may also even say
(theory permitting) that the competent speaker knows that in the first sentence
the empty subject - PRO - of the infinitive complement is (i) obligatorily controlled by (roughly, co-referential
with) the matrix subject and (ii) is the clause-mate antecedent of the
reflexive, while in the second sentence, the empty subject is the trace (or
copy) of the raised wh-argument, with
the lower copy binding the reflexive.
But the relevant I-language here is not
the set of such judgements nor even the set of LF structures which encode the
chains of dependence hinted at. I-languages themselves are not infinite abstract objects nor ‘statements’ of such knowledge
that can coherently fit into Fodor’s schema (especially see Chomsky, 1994,
p.158; 2000a, p.73.) An I-language is that which, in part, generates such knowledge, not the products generated. Indeed, the
nature of the faculty is a theoretical issue; it is not a descriptive notion
determined by a mere abstraction from our intuitive judgements.
It is wholly unclear, then, in what
sense we might know-that an
I-language (a state of the language faculty). It really does bear emphasis
that, if we bracket Chomsky’s various informal statements about ‘knowledge’,
which, over the years, have mainly been in response to philosophical queries,
and look just at the theories proposed (especially the recent ones), then we
don’t find likely propositional objects of knowledge, whether realised by
internal propositional states a la
Fodor, or not. We find descriptions of lexical items as (non-phrasal) feature
clusters; a computation from lexical selections that merges and moves lexical
items, and constraints/principles the computation inherently meets that define levels of description of the interfaces
at which external systems integrate with the computation. There are no
components here that are constrained to be propositional so as to cleave to our
commonsensical attitude verbs. Chomsky is explicit about this:
…
in English one uses the locutions "know a language," "knowledge
of language," where other (even
similar) linguistic systems use such terms as "have a language,"
"speak a language," etc. That may
be one reason why it is commonly supposed (by English speakers) that
some sort of cognitive relation holds
between Jones and his language, which is somehow "external" to
Jones; or that Jones has a "theory of his language," a theory that he
"knows" or "partially knows."… One should not expect such
concepts to play a role in systematic inquiry into the nature, use, and
acquisition of language, and related matters, any more than one expects such
informal notions as "heat" or "element" or "life"
to survive beyond rudimentary stages of the natural sciences. (Stemmer, 1999,
p.397; cf. Chomsky, 2000a, p.72/119, Chomsky, 2000c, p.23.)
The
position here, note, is not premised upon a rejection of any particular
theoretical substantiation of knowledge,
be it Fodor’s RTM or something more neutral. The much more simple point is that
knowledge of language is just a
reflection of English collocation; it is not a notion that requires any
theoretical substantiation, still less does it play a constraining role in
linguistic theory. There is no relation R
between speakers and theories/grammars, save, of course, those we can stipulate
willy-nilly (see §7).
We may view the retention of the
epistemic locution as a species of Wittgensteinian sophisticated naiveté.
Occasionally, Chomsky (e.g., 1980, p.70, 1986, pp.32-3/267-8) does point out
that his use of ‘knowledge’ is not necessarily at odds with its common-sense
understanding, although any correspondence is of marginal interest. The point
here is that ‘knowledge’ is employed as an informal term, one which may readily
give way to a more precise specification. This approach, however, is not
legislative. To propose that the faculty hypothesis is not best thought of in
epistemic terms is not to be an eliminativist about ‘linguistic knowledge’, as
if English collocation were somehow expressive of an outmoded ontology and
should hence be regimented so as to lose such an ontic commitment. Rather, we
are led to the simple thought that the job of linguistic theory is to
understand the nature of the language faculty (UG, developed I-languages and
their integration with external systems), not
how the language faculty stands to our colloquial locutions. It turns out, as
part a matter of discovery, part methodology, that we do not know languages (better: I-languages) in
a range of senses which the verb carries. For example, notions of conscious
access, justification, authority, and norms are simply not applicable. Most
centrally, the verb also loses its transitive/relational aspect; this was one
of the main motivations for Chomsky’s cognize
neologism and is explicit in the above quotation.[12]
An I-language is not an independent ‘object’ - a set of propositions - that is
represented/known by a speaker/hearer; it just is a state of the
speaker/hearer. From this perspective, we may say that ‘language goes on
holiday’ when the colloquial knowledge relation is recapitulated internally, as
if to possess a faculty requires us to stand in some epistemic relation towards
our internal states so that they may serve as the putative objects for our
transitive verb. Only persons know
languages; mind/brains do not, just as they do not see cats on mats. More collocation. The import of the common verb -
the aspect retained - is that the knowledge is a state (or a set of states)
rather than a relation to an independent thing, and this state accounts for the
way in which speaker/hearers produce and consume language without the state
being exhausted by the specification of such an input/output profile.
5:
Collapsing a Distinction
The
proposal offered is that we should collapse the representation/represented distinction. The language faculty is not
represented by the mind/brain; it is
an aspect of the mind/brain. In essence, it is a function from a lexicon to
pairs of structures that integrate external systems. The function is defined
(in intension) by the descriptive conditions it must meet. The states of the
faculty are solutions to an equation between the features of the lexical items
and the conditions imposed by the external systems as to which features are
legible to them. The states are not amenable to causal generalisation; nor,
perforce, do they contribute to the aetiology of linguistic acts. Two potential
problems arise here. Firstly, in what way does the poverty of stimulus argument
support the hypothesis? Secondly, does the proposed collapse of the
representation/represented distinction vitiate the performance/competence
distinction?
The first worry here is wholly
misplaced. As seen above, poverty of stimulus considerations cannot directly
militate for the claim that there is a language module, a component of the causal architecture of the mind/brain.
Equally, though, they do not militate for a propositional
conception of innate knowledge. The classic arguments of Chomsky are only
intended to show that the mind/brain is structured (biologically or physically)
to develop/grow I-languages given certain broad environmental conditions; the
mind is not an unformed general mechanism whose eventual shape is determined by
experience. On the present conception, a theory of the language faculty is
simply a specification of such a structure. Thus, the poverty of stimulus
considerations tell us, in effect, that the mind/brain must be natively
structured in such a way as to be sensitive to concepts of (theory permitting)
categorical features, case properties, theta-roles, head projections etc., for
these concepts are not recoverable from the data but are necessary (by present
theory) to account for the observed structure of linguistic output. So, we do
not need to think of Chomsky as confusedly using an epistemological argument
for an architectural conclusion.
What of the concern that a departure from
the epistemological model leads to a destabilisation of the
competence/performance distinction? As noted above, the distinction is not
written in stone; it marks a methodological simplification of the complexity
that constitutes linguistic cognition in
toto. In a sense, the distinctness or autonomy of the language faculty just
amounts to the idea that it is a central system rather than an input/output
one. That is, systems of the mind which deal with production and understanding
of language - performance - interface with the faculty, but the faculty is
independently constituted (especially
see Chomsky, 2000a, p.117-8.) This is not a stipulation but a theoretical assumption
that is sufficiently supported to be our default position. The structure of
language cannot be explained by performance in that it is not reducible to -
explicable from - the meaning or sound properties it is used to encode.
Trivially, sound is linear, but grammatical structure is hierarchical.[13]
Likewise, there are perfectly well structured ‘representations’ which, for
most, are unusable; and there are perfectly acceptable meanings that are
expressed by structures, but which are not typically encoded in those
structures.[14] As a matter of fact, then, rather than
stipulation, it looks as if the language faculty is not a performance system.
It is that third component that is required to effect a convergence of sound
and meaning, for properties of neither are predictable or explainable from the
other. Further, this component counts as a system of competence just in the
sense that it plays this third role, ‘instructing’ performance systems while
being independently constituted. In perhaps picturesque terms, the faculty is
an engineering solution to the problem of effecting a convergence between the
two independent systems. The faculty remains autonomous, though, in that its
output pairs answer to independent constrains (sound and meaning respectively).
The faculty cannot see or know what use, if any, its interfacing systems may
make of its outputs; that the pairs are convergent is determined by the
internal structure of the faculty - it does not have one eye, still less two,
on what, if anything, the rest of the mind may make of them (Chomsky, 2000a,
p.27/117-8). Chomsky goes as far as to suggest that the faculty could have been
employed for locomotion in a differently organised mind/brain, if it had been
integrated with different systems (Chomsky, 2000a, p.27). The point here is
just that, while the faculty meets the conditions imposed by the external
systems, its internal design does not encode for just those conditions as
opposed to a potential myriad of others. Obviously, no theory (a set of
propositions) that ‘truly represents’ an external language could effect
locomotion in a differently organised system.[15]
A final worry which might be raised is
to do with our ‘free access’ to language. We can, for example, reflect on
linguistic structure and discern previously unnoticed ambiguities or see that
what appeared to be gibberish is in fact meaningful (see n.14). Indeed, much of
the data for linguistic theory comes from such reflection by the linguist,
rather than data on performance, although, of course, there is no bar on such
data. What this appears to show is that the object of linguistic theory is our
propositional knowledge, that which, in part, is reflectively available. The
impression is wrong. As earlier stressed, we can indeed be said to know or to
cognize the products of the faculty,
but they are not the faculty itself; the faculty is that which, in part,
generates the products, and it is not open to conscious reflection.
Again, to bar conscious access to the
language faculty is not, perforce, to consider it a parser module - an
encapsulated device that mandatorily produces structural descriptions for
acoustic inputs. Chomsky and others have been vociferous about this: as
emphasised, there is a difference between acceptability
(parsing results) and interpretability
(what’s determined by the faculty). Indeed, Chomsky, 1996, pp.14-5, even doubts
whether there is a universal human parser (cf. Chomsky, 1986, p.14, n.10; 1991,
pp.19-20; 2000a, pp.117-8; Stemmer, 1999; Higginbotham, 1987). This is not to
say, of course, that parsing processes do not involve the language faculty; it
is only to say that the faculty is neither a parser itself nor an encapsulated
database for one (cf. Weinberg, 1999, on minimalism and parsing). So, the disavowal of conscious access would
translate into an argument for an epistemic understanding of the faculty only
if the only alternative were a parser module. This is not the situation we are
in. The kind of freedom and conscious access language users enjoy is, for
Chomsky, a mystery, potentially beyond human comprehension. Whether this is so
or not, no light appears to be shed on the problem by saying that the faculty
is propositional rather than an abstract structure (a function in intension).
After all, the problem of free will is not solved by saying that, unlike in the
case of a knee-jerk reflex, one has access to the knowledge that helps one
decide whether to act in this way or that. There would be a difference if
linguistic performance could be explicated in causal-computational terms, but
that is not on the cards.
Still, it might be thought that only an
epistemic understanding of the faculty is so much as consistent with the phenomenon of free access. Yet again, it is
difficult to see such a claim as more than an insistence on Fodor’s exhaustive
disjunction. On the current position, the job of at least making sense of the
free access or the ‘dawning of meaning’ phenomenon, as witnessed in the
examples of n.14, is necessarily divided between the language
faculty and other systems of the mind. We may view the faculty’s role as
determining the range of interpretations (potentially zero) a given structure
may possess, independent of whether we immediately recognise such interpretations.
When we work out, say, that a structure does in fact have some unnoticed
interpretation, we are not consulting the faculty at all. To see this,
consider: Sailors sailors sailors fight
fight fight. One way of explaining the coherence of this sentence to the
uninitiated is to ask them to drop the central clause - sailors fight - and stress the remaining relative, to produce:
Sailors SAILORS FIGHT fight (additional pauses before and after the relative
helps). Then they see that the material they dropped is just another
clause. But we are not here simply
analysing the sentence, working out its phrasal hierarchy or the selection
features of the verbs or some such; rather, we are appealing to performance
factors of stress (or intonation, or context) in other examples. (Similar remarks hold for ‘garden paths’, such as
the illuminating contrast between The
horse raced past the barn fell and The
onions fried in the pan burnt.)
We may say that the faculty has done its job and remains in aloof silence on
the matter, while we delve around in matters of meaning, sound, context, and
comparison. We are, for sure, learning something about the faculty - e.g., that its principles generate
interpretable n-degree centrally
embedded relatives - but we are not being instructed
by the faculty (see §7).[16]
Let us now look at Fodor’s explicit
argument that a non-epistemic understanding of the language faculty cannot be
what Chomsky has in mind.
6: Fodor’s Argument
Fodor,
2001, offers an argument whose conclusion is that Chomsky must be making a clear distinction between representation and
represented. As previously noted, many others agree with Fodor in this regard,
but neglect to provide something like an argument; instead, they rely on the
simple inference, ‘If the faculty isn’t doing the representing, then it must be
the represented’ (e.g., George, 1989; Higginbotham, 1987, 2001; Knowles, 2000).
Fodor’s argument is directed at Cowie, 1999, who appears to be insensitive to
the distinction between the language faculty and performance systems. If Cowie
indeed recognises no such distinction, then she is simply mistaken, independent
of Fodor’s argument. So, let us put aside a
priori inferences and just think about Fodor’s argument on its own terms.
Fodor argues as follows. Distinct
grammars G and G* for a language L of speaker S can be co-extensive in
the sense that they (strongly)
generate the same set of L-sentences
with the same structural descriptions. That is, the grammars are in agreement
about the sentences S understands and how he would, say, parse them. Let this
mean that G and G* are descriptively
adequate grammars of S’s language. It does not follow, though, that G and
G* are equally true of S. G*, say,
might include principles of ordering or substitution, or might even be
infinite, which are all, ex hypothesi, incompatible with UG (if,
say, S would have tried to acquire his language by fixating on G*, he would
have failed due to poverty of data.) Thus, while G* extensionally picks out S’s
language up to phrase structure - what S knows
- it cannot be the grammar S internally represents, for it is inconsistent with
UG, the hypothesised initial state of S’s language faculty. Now Fodor’s key
point is that if grammars are internally represented intentional objects, this
difference between descriptive adequacy and truth is easily explained. The
difference is that while G and G* express the same content - G and G* are descriptively adequate - S can
represent G, but not G*. Thus, there is a notion of psychological reality - the underlying representations - that is
not exhausted by descriptive adequacy, i.e., what is represented. If there were no difference between
representation and represented, then there would be no difference between G and
G*, both would be true of S qua descriptively adequate.
The basis of Fodor’s reasoning here, I
think, has its roots in Quine’s, 1969, 1972, transposition of indeterminacy
considerations from translation/meaning to grammar.[17]
Thus, one finds the essence of the above argument succinctly intimated twenty
years early:
If…
the notion of internal representation is not
coherent, the only thing left for linguistic theory to be true of is the
linguist’s observations [i.e., that covered by the constraint of descriptive
adequacy]… Take the notion of internal representation away from linguistic
metatheory and you get positivism by subtraction (Fodor, 1981a, p.201).
I’m not
here concerned with whether Fodor’s argument is a good response to Quine-style
scepticism. In fact, in Chomsky’s, 1975, 1980, extended discussions of Quine,
his response has simply been that there is no coherent indeterminacy thesis
beyond the trivial claim of theoretical underdetermination (cf., Chomsky and
Katz, 1974, in response to Stich, 1972). Of course, a concomitant of this claim
is that we shouldn’t artificially impose operational criteria on the data which
may be relevant to linguistic hypotheses. But it just doesn’t follow that the
only way to forego an indeterminacy inducing operationalism is to endorse a
representation/represented distinction. Whatever the proper response to Quine
and those influenced by him might be, the argument proposed by Fodor is not
Chomsky’s; if, then, Fodor wishes to endorse its generalisation as part of his
intentional realism, as he does, the ‘Chomskyan’ approach to language is not a
basis from which to proceed. First, something will be said about Fodor’s
endorsement, then the argument will be diagnosed.
The question of what verb we should use in
relating a subject to his linguistic competence is somewhat trivial, a matter
of collocation. The issue, however, is not trivial for Fodor precisely because,
for him, propositional attitudes are the level of theorisation which may
track, and so capture in
generalisations, the causal springs of human behaviour. In crude terms, then,
the problems for Fodor we have been witnessing may be due to his reading
Chomsky as if he were Fodorian. That is, Chomsky is being read as if he
endorses some general theory of cognition, such as Fodor’s representational theory of mind (RTM), with the language faculty
being a particular set of belief contents (something which may be the object of
‘cognize’) represented in some more generally specifiable format (a language of
thought) whose tokens form the causally efficacious states of our beliefs and
knowledge (see Fodor, 1983, §1.1 for the explicit reasoning). Again, this
perfectly fits with Fodor’s position that the language faculty is essentially
at the disposal of a parser module.
Fodor has explicitly talked this way from at least the early 1970s (see
n.2). In effect, the argument Fodor attributes to Chomsky is just an argument
for a particular instance of what Fodor sees as a general distinction between
mental content and underlying representational structures. As we have seen,
though, Chomsky endorses no general theory of cognition in terms of which
language might be understood as an instance, still less a language of thought
in Fodor’s sense (Chomsky, 2000a, pp.176-8). The computation which underlies
linguistic competence is sui generis,
essentially linguistic in that it is defined in terms of operations from a
lexicon to a set of <PF, LF> pairs. Moreover, as also noted and
supported, Chomsky has no commitment whatsoever to the integrity of our
intuitive concepts of belief or knowledge. A simple point should make this
palpable. Chomsky has long contested that human action, and linguistic
behaviour or performance in particular, is not caused (especially see Chomsky,
1975, Chp.1; 2000a, pp.72/95.) But Fodor’s general view is that RTM is the only
way to fit content with causal states. If, therefore, Fodor’s endorsement is an endorsement of the
idea that the language faculty is content that may feature in intentional
generalisations causally implemented in computational states, then it is not an
endorsement of anything for which Chomsky argues. Let us now turn to the
argument itself.
The first thing to note is that Fodor’s
argument, if from anywhere, is from the standard
theory.[18]
But even on this understanding, the argument is misleading. The classic
distinction of the standard theory is one between descriptive adequacy and
explanatory adequacy - not
‘truth’.[19]
That is, the thought was that where general linguistic theory might allow for,
in principle, two or more grammars that would explicitly generate the same
structural descriptions for some language L,
those grammars would only be equal in terms of explanatorily adequacy just if
(i) they could be acquirable given data + UG principles and (ii) were equally
‘simple’ by a UG metric. If two or more grammars were so equal, then, effectively, they would be notational variants. Now, if
‘truth’ is intended to cover grammars sanctioned by ‘explanatory adequacy’,
then his argument just does not go through. A descriptively adequate grammar is
not ipso facto a grammar a speaker might know in any sense
that plays a role in linguistic theory, i.e., descriptive adequacy just doesn’t
equate with possible content. The very point of explanatory adequacy, as a
condition on general linguistic theory, is to constrain the grammars speakers can know, to “distinguish natural
languages from arbitrary symbolic systems” (Chomsky, 1965, p.36): it is
nomologically impossible for a speaker’s language faculty to generate a
descriptively but non-explanatorily adequate grammar, i.e., one that is not
appropriately selected by the constraints on UG. Thus, of Fodor’s two grammars,
G* qua not generatable by UG just
cannot be represented, even if it is co-extensive up to sentential phrase
structure with the acquirable G. Chomsky’s actual point was much simpler than
Fodor’s attribution. It was that descriptive adequacy is only necessary, not sufficient, for a grammar to be ‘known’
(ibid, p.34); the adequacy condition
provides an “external” justification by relating the grammar to observed facts
(ibid, p.27). Explanatory adequacy provides
an “internal” justification of the same
grammar by relating it to UG in showing that the grammar is in fact acquirable,
an actual human language, rather than an “arbitrary symbolic system”. So, the
two adequacy constraints jointly determine
a single representational system: an adequate grammar is one which ‘fits the
facts’ and is explicable as an instance of general rules and conditions that
may be understood to be universal, innate (cf. Chomsky, 1964, p.63). As
remarked in §3, this difference between ‘internal’ and ‘external’ was already
present in Chomsky’s earlier work of the 1950s, where it had a methodological
gloss as opposed to a psychological one. On either gloss, it is flat wrong to
think that one condition is to do with external content - what is known - while the other is to do with
its internal representation - how it is known or represented. The distinction
is to do with particularity, determinable via ‘external’ observation, and
generality, determinable via ‘internal’ investigation of the resources
available to any human language. Unsurprisingly enough, if Fodor’s argument
does not even track the reasoning of Aspects,
we should not expect it to track the reasoning of the more recent principles and parameters (P&P)
model.
Under the P&P model, UG is theorised
as the schematic initial state of the language faculty; exposure to data fixes
a lexicon and triggers a given vector of values to a finite set of parameters,
which results in a steady state of the faculty. Roughly, this state is a
speaker’s I-language (what we have just been calling a grammar) (see Chomsky,
1986, chp.2, and 2000a, chp.1). Clearly, Fodor’s argument makes no sense on
this model.[20]
There is no conception here of a grammar
or language to which notions of extensional equivalence or strong - let alone weak -
generative capacity apply. Chomsky (2000d, p.141, n.21) is explicit on this
matter:
“[We should] put aside [the view] that
restricts “linguistic evidence” to identification of “well-formed”
(“grammatical”) expressions [i.e., “external” evidence which bears on
descriptive adequacy], so that the linguist then faces the alleged problem of
selecting among grammars that are extensionally equivalent over these objects.
Such demands [are] incoheren[t].” (Cf. Chomsky, 1996, p.48, and 2000a, p.132.)[21]
Note:
Chomsky is not answering the problem, he is dismissing it as one which does not
so much arise because the operative extensional notions are senseless. There is
no threat of “positivism by subtraction”. An I-language is simply a state of
the mind/brain: a procedure from a lexicon to a <PF, LF> pair. There is
nothing here to be descriptively adequate but not true, and so there is no
distinction which needs explaining via a representation/represented distinction.
In this light, there is no real distinction at all, still less one of the form
Fodor assumes. There is the one system, which can realise distinct states, and
we want an explanation for all of them (cf. Chomsky, 2002, p.131). Still, we can perhaps say that a theory of
the faculty is descriptively adequate if we can see how the developed states of
the system support mature competence, and it is explanatorily adequate if the
account of the initial state (inter alia)
of the system is such as to furnish an explanation of the maturation of the
descriptively adequate states. This is a distinction of perspective, not fact
or principle. A uniform computational procedure (the intensional function) is
taken to constitute the initial state (UG), and the determination of an
I-language will reduce, under a minimalist
understanding, to a specification of lexical idiosyncrasies and how these
affect the morphological marking of functional categories. Trivially, on this
understanding, the idea of a pair of strongly generated I-languages, only one
of which may be represented, is meaningless: (i) I-languages are not generated
at all; they are determined by parameter settings. (ii) By definition, an
I-language must be ‘represented’
(realised in some brain), for it is nothing other than a state of a brain,
period. (iii) All I-languages are the same up to lexical differences and
morphology; the only ‘grammar’ there is is UG, up to the idiosyncrasies
indicated. (iv) Consequently, there is no concept of a grammar that applies to
a language, and so no possibility of two grammars equally applying to the same
language. We can think, in theory at least, of distinct I-languages generating
the same E-language, where an E-language is a set of symbol strings
type-classified according to some rough phonological or grammatical criteria:
all the sentences of ‘English’, say. But this distinction is no part of
linguistic theory proper, and an E-language is certainly not what the
linguistic mind represents.
7: (Tacit) Knowledge of Language
The
argument so far has been that Chomsky’s understanding - and the understanding we should also have - of the language
faculty falls outside of Fodor’s faculty/module distinction. In particular, we
have seen that there is no basis for the view that the faculty comprises a
potential propositional attitude ‘object’: the faculty is not a set of
propositions represented by the mind/brain over which relations of knowledge or
belief might be realised. There is a philosophical tradition, however, in which
tacit knowledge is understood to be a
substantive notion in the (potential) absence of explicit propositional states,
i.e., a theoretical attribution of propositional knowledge to a speaker/hearer
does not entail the speaker/hearer explicitly (internally) tokening ‘sentential’
states with such propositional content (Evans, 1981; Davies, 1986, 1989;
Peacocke, 1989). Does this difference make a difference to the preceding
arguments?
Beginning with Evans, 1981, the
motivation for this sort of account arises from a concern over our putative
knowledge of the axioms of a semantic theory for our respective languages. The
axioms express propositions, but the speaker/hearer cannot be said to token
such propositions under an attitude; for Evans, genuine propositional attitudes
must be inferentially promiscuous, at the service of the speaker/hearer for
general reasoning implicated in a myriad of projects, rather than harnessed to
particular capacities (for discussion, see, e.g., Stich, 1978; Davies, 1989;
Knowles, 2000). The details of Evans’s resolution of this quandary need not detain us. The upshot is
that the ascription of propositional knowledge is justified to the extent that
the inferential role of the axioms in delivering truth conditional statements
the speaker/hearer is disposed to accept map onto the causal roles of
sub-doxastic states that enter into the causal explanation of the
speaker/hearer’s disposition to assent to the truth conditional statements
mandated by the theory. A more precise formulation, which does without
dispositions, is articulated in Davies’s, 1986, mirror constraint, where the deductive structure of the theory
known - a set of propositions - is mirrored in the causal structure of
non-propositional states that explains the speaker’s judgements that the theory
mandates. In essence, this kind of approach is transposed to linguistic theory
by Peacocke, 1989, and supported by Davies, 1989. There are differences between
the two accounts offered, but for our purposes, they may be subsumed under a mapping condition:
(MC) A
speaker/hearer tacitly knows the axioms, A1,…An, and
rules, R1,…Rn, of a grammar G (G is ‘psychologically
real’) if, for any p, such that G d p, the explanation of p
holding of the speaker/hearer’s language essentially appeals to a causal
process that may be factored into a set of states which respectively map onto
the axioms/rules of G from which p
follows.
In essence, MC says that the technology
of a grammar can be said to be known by a speaker/hearer because the
propositional units of the grammar that are common factors in derivations map onto states that
are the common factors in the causal explanation of the speaker/hearer’s
judgements that are mandated by the grammar (cf. Davies, 1989, p.132). My
concern here is not with the precise formulation of some such condition, nor
even with the general notion that propositional attribution may be legitimate
in the absence of propositionally structured internal states. I do think, for example, that Peacocke, 1989, pp.122-4, is right to
suggest that a non-sentential ‘connectionist’ architecture is perfectly
consistent with the ‘psychological reality’ (as understood via MC) of
generative grammars (cf. Davies, 1989, p.151; George, 1989, p.102). My only
concern is whether the advertised model provides a substantiation of the
locution knowledge of language that
has any salience for linguistic theory. I think it does not.
In broadest terms, the moral of my
position is that knowledge of language
is innocent, independent of any connotation of consciousness or justification,
so long as we are immune to any serious methodological/psychological split,
under which we are obliged to instantiate, in some sense, the formula R[S, T(L)]. Here, ‘S’ is the speaker/hearer, ‘T(L)’ is a theory of a language L
as determined by general linguistic theory, and ‘R’ is a (non-stipulated/substantial/external) relation between S
and T about which we require some psychological theory. Fodor approaches the formula in
straightforward realist terms: R is knowledge which involves the
representation of T such that S realises a state that is true of L.
Our problem with Fodor has not so much been his particular brand of
intentional/representational realism, but the very structure of the formula
which is instantiated. According to internalism, there is no relevant relation R because there is no L which we require a theory of, and so
no substantive relation which may hold between it and the speaker/hearer. The
formula which is our target might be expressed simply and trivially as S(I),
where I is an internal state of speaker/hearer S. We say that the states are
I-languages and to know a language is to be in such state. The speaker/hearer
stands in no ‘internalised’ relation to our
theories, for our theories are just of
the internal states, not an independently specifiable ‘object’ the states
represent as the content of linguistic knowledge (cf., Chomsky, 1975, chp.1;
2000c, pp.19-20).
Well, the problem with accounts of tacit knowledge, as the notion has been
construed, is that they retain as their target the formula R[S, T(L)] (e.g., see
Peacocke, 1989, p.114; Davies, 1989, p.133). Under such theories, R is not directly propositional; rather,
R is constituted by a complex of
causal roles whose nodes or states map onto (‘realise’) the contents of
primitive ‘axioms’ and ‘rules’ of T(L).
We have already seen at length that the states of the faculty do not enter into
the causal nexus leading to linguistic acts. The fundamental problem with the
tacit approach, though, is the very idea of accounting for the chimerical R between speaker/hearers and
propositional contents, not particularly with its causal assumptions.
The continuing echo of the split approach
of Chomsky’s early work, which persisted in the nomenclature up to the
development of P&P, makes the point here difficult to appreciate. Once we
properly acknowledge the internalist/intensionalist framework, however, the fog
clears. First off, generative grammars (I-languages) are not deductive
theories, such as have been proposed for semantics, nor, as Peacocke, 1989,
p.114, suggests, do they “correspond” to them; there are no axioms or
particular rules (we may exclude general operations such as Move/Affect
a or Merge) to be mapped onto internal
states. Of course, as Peacocke, ibid, p.117, also remarks, MC can potentially
be reformulated to cater for distinct conceptions of a grammar. The crucial
point, however, is that there simply are no propositional informational units
which linguistic theory attributes to a speaker/hearer, and so there are no
particular states to be discriminated (propositional or not) which express such
information qua a correspondence
between a causal and deductive factorisation.
Consider the principles of binding
theory as classically laid out in Chomsky, 1981.
(A) A
reflexive is bound in D.
(B) A
pronoun is free in D.
(C) A
R-expression is always free.
(‘D’ is a
dummy label for a phrasal domain in which (at least) the respective nominal
occurs; the nature of the domain is disputed). It seems that a theory which trades in A-C attributes knowledge of A-C
to the competent speaker/hearer as part of its explanation of her competence
with nominals. Just so, following MC, it seems that we should say that a
speaker/hearer tacitly knows-that A-C just if there is a causal process that
(i) essentially involves states that (non-propositionally) encode A-C and (ii)
eventuates in a speaker/hearer’s judgements which involve the interpretation of
structures featuring binding relations that adhere to A-C. Again, precisely how
this might be made clear need not detain us, for the very idea, independent of
its causal presumptions, is mistaken, a misunderstanding of the principles A-C.
That A-C are stated as atomised
informational units - propositions, things knowable - is an artefact of our theorisation; it is not
part of the meta-theory that the speaker/hearer has internalised such units.
The theory is our account of the
faculty, but the faculty is not the theory put inside the head (as propositions
or not), as it were. A-C are
internal conditions which the computation of the faculty meets, but there is no
presumption that there are any states at all which are specifiable by the
principles, still less that some such states enter into causal roles
eventuating in explicit judgements. So, for example, the common idea during the
1980s was that A-C hold of S-structure and LF, where this just meant that
legitimate ‘objects’ at these levels meet the conditions specified by A-C; if
they didn’t, they wouldn’t be at that level. (The principles act as filters on otherwise acceptable
structures (at D-structure).) The essential idea here is much more transparent
within the minimalist program, which may be properly seen as a logical
progression of, rather than an alternative to, the earlier framework. Levels (essentially,
just LF, since PF is just features with no syntactic structure) must meet
certain conditions if the computation is not to crash (is not to produce illegible objects at the interface), but
the conditions can be inherently realised in the nature of the computation
without there being any independently specifiable state corresponding to any
condition which is met (see, e.g., Hornstein, 2001, Kayne, 2002, and Zwart,
2002). In effect, A-C are descriptive of an intensional computation; it is
redundant to think of them as independent items of knowledge reflected in
states that uniformly enter into certain causal processes.
Such considerations may be deployed
generally over the technology of recent linguistic theory: X-bar theory, ECP,
Subjacency, relativised minimality, minimal link condition, head movement
constraint, and so on and on. Further, this reasoning holds equally of both UG
and particular I-languages. Peacocke, 1989, pp.126-8, suggests that there is a
difference: with the acquired state (I-language), knowledge of a condition is
legitimately attributable merely if the given condition obtains (i.e., it need
not correspond to some specifiable sub-state), but with the initial state (UG),
the conditions are “second order informational states” ‘about’ the conditions
which hold for acquired languages. Fodor, 1981b, p.258, appears to express the
same thought with his claim that speaker/hearers “believe” that
“transformations must apply in cycles” (clearly, no particular grammar could state/encode such a claim). This
kind of construal depicts UG as a set of meta-instructions on how to form a
grammar, which may give way once a grammar is formed that cleaves to the
instructions. This is a mistake. UG has long been theorised as a schema, a
polymorphic state, whose end-states are parameterised along just a few
dimensions, perhaps just the morphological realisation of functional categories
as proposed under minimalism. UG is
not about particular languages. To
see the point here, consider the projection
principle (PP), one of Peacocke’s own examples to illustrate his
‘meta-information’ hypothesis. As initially proposed (Chomsky, 1981, pp.29-32),
PP says that syntactic structure at all levels is projected from lexical items
as a reflection of their selection features. According to Peacocke, then, UG
contains meta-information about the
form of any particular language in the guise of PP. This is nonsense. PP is our theoretical statement about the
relationship between lexical items and syntactic levels as realised in the
endstates of the faculty; PP is nowhere encoded. This is trivial to see. In
effect, PP says that if lexical items inherently contain their selection
features (as they must, due to idiosyncrasy), then it is redundant for that
information to be duplicated in phrasal rules, for any structure that does not
respect an item’s selection frame will be illegitimate at any level (i.e.,
contravene principles of binding, case and q-role assignment); thus, the system doesn’t require a
principle to encode this (e.g., one doesn’t need a rule to tell one that Bill is probable to leave is ill-formed,
if probable independently carries the
information that it only takes finite complements.) Ironically, to construe PP
as meta-information is to court the very redundancy PP sweeps away: UG keeps to
PP as an effect of the relationship between the lexicon and licensing
conditions at the syntactic levels.
Similar reasoning applies to Fodor’s
example. The classic idea of Aspects
(1965) is that recursive base rules
defined over ‘S’ (sentence/clause) allow for non-transformational clausal
embedding. This permits an economising of the transformational component that
may now just consist of independently required cyclical singular
transformations, rather than such transformations and (unordered) generalized
transformations that target two or more base generated mono-clausal structures.
Thus, on the Aspects model, it would
be wholly redundant for UG to encode the meta-information that all
transformations apply cyclically; the ‘information’ follows directly from the
applicability of singular transformations to any multi-clause base generated
D-structure (see Chomsky, 1965, pp.132-6).
In sum, the failure of the tacit knowledge
model is its very presumption that the target of explanation is how we may
properly ascribe propositional knowledge to a speaker/hearer, whether or not
this involves ascribing explicit internal propositional states. This endeavour
is no part of linguistic meta-theory: a linguistic theory is an account of the
(defining) conditions an internally realised computation meets. We state the theory propositionally, but
we don’t attribute the propositions to the speaker/hearer. There is no
methodological/psychological split, with ‘what is known’ as a specifiable set
of propositions (a theory/grammar) and an account of a relation R which holds between speaker/hearer and
theory.
8: Conclusion
The
language faculty is an abstractly specified computational (= intensional
function) system of the mind/brain. It is abstract because, while realised in
the brain and having no independent existence, it is individuated in terms of sui generis concepts particular to the
domain of the computation’s outputs. This characterisation renders the faculty
neither as a set of propositions to be known, nor as a mechanism, a part of a
causal nexus.[22]
Notes
[1] Ultimately, the point here is simply that modularity is not the same thing as innateness (cf. Khalidi, 2001, Garfield, et al., 2001).
[2] Fodor has consistently held the view that there is a parser and that it has an encapsulated representation of a grammar. A “representation” of a language is “part of a sentence encoding-decoding system” (Fodor, et al., 1974, p.370); “the production/perception of speech [is] causally mediated by the grammar that the speaker/hearer learns” (Fodor, 1981a, p.201); language is a “psychological mechanism that can be plausibly thought of as functioning to provide information about the distal environment in a format appropriate for central processing” (Fodor, 1983, p.44; cf., p135, n.28); and “the domains of perceptual modules (like the language processor) can be detected psychophysically… Modules (especially Chomskian modules) are inter alia innate databases” (Fodor, 2000, p.78/96). Fodor, at least early on, acknowledged that this approach was not Chomsky’s: “I shall… propose a somewhat eccentric way of reading the linguistics and psycholinguistics that developed out of Syntactic Structures… this work is best viewed as contributing to a theory of verbal communication” (Fodor, 1975, p.103). Fodor’s (p.c.) general argument for this position is that parsing requires a rich grammatical description of the signal. Thus, unless the mind/brain represents the same grammar twice, it appears that the language faculty is a database for a parser module. Chomsky is certainly at pains to distance himself from such a view See §5.
[3] The levels of PF and LF are not essential. In recent work, Chomsky, 2001a, 2001b, among others, raises doubts whether it is appropriate to think that there are any levels of representation. For present purposes, nothing at all hangs on this recent development (but see Collins, 2003). I thus shall keep with ‘PF/LF’ for convenience.
[4] Chomsky has been consistent in using non-epistemic terms, such as ‘device’, ‘mechanism’, ‘brain area’, etc., to pick out the language faculty. See, e.g., Chomsky, 1965, pp.53-6; 1980, p.28; 1986, pp.12-3; 1994, pp.153-4; 2000a, pp.4/117-8; 2001b, p.1. For the explicit rejection that the language faculty is a Fodorian module, see Chomsky, 2000c, p.20; 2000d, p.140n2.
[5] For new thinking on this familiar model of UG, see Yang, 2002.
[6] In particular, the conditions do not have to be ‘linguistic’ in the sense of only being specifiable in terms of noun or subject or some such. Indeed, such conditions would amount to a departure from principled explanation (Chomsky, 2001b). In line with this thought, Huaser, Chomsky and Fitch, 2002, speculate that the faculty proper might be unadorned recursion.
[7] This is not to say that a theory-like competence explains creativity; it is only to say that the fact of creativity can be seen to be partly realised by such a competence - an infinitely generative system - whereas the fact appears to be flatly inconsistent with the notion that language is a form of learnt behaviour, a habit.
[8] Some eight year later, Chomsky, 1967, expressed regret that his positive cognitive proposal was “apologetic and hesitant”, not, of course, that the ’59 review was a failure for not properly supporting the innateness of language - as if.
[9] Consider: “I do not know why I never realized [it] before, but it seems obvious, when you think about it, that the notion of language is a much more abstract notion than the notion of grammar. The reason is that grammars have to have a real existence, that is, there is something in your brain that corresponds to the grammar… But there is nothing in the real world corresponding to language” (Chomsky, 1982, p.107).
[10] In effect, the proposal here is precisely that we should “become confused about linguistics” in the sense of George, 1989, 1990. Somewhat like Fodor, 1983, 2000, and Knowles, 2000, George’s argument turns on the spurious exclusive disjunction ‘mechanism or propositional knowledge’, or, perhaps better, ‘representation or represented’: since the faculty is not a representational mechanism, it must be that which is represented.
[11] E.g., Chomsky, 1968/72, p.191, writes: “In general, it does not seem to me true that the concepts “knowing how” and “knowing that” constitute exhaustive categories for the analysis of knowledge”; and one would “find it difficult to understand” linguistic nativism if one keeps to the distinction.
[12] Knowles, 2000, pp.325-6, takes ‘cognize’ to differ from ‘knowledge’ just in respect of conscious access. His thought is wholly based on Chomsky’s, 1980, p.70, remark that if we were to be conscious of the faculty, then we would not hesitate to say that the conscious state constitutes what we know. But it patently doesn’t follow that what would be so known would be propositional (consciousness is not typically propositional), still less that what is in fact ‘known’ is propositional. Chomsky’s simple point is that conscious access is an orthogonal issue.
[13] By Kayne’s, 1994, influential linear correspondence axiom (LCA), the precedence relation of PF is determined by a (hierarchical, antisymmetrical) c-command relation within the syntax, which remains non-spelt-out to meet semantic conditions at LF. Some such arrangement is widely assumed, although we need not be committed to LCA as an axiom - its content can be deduced.
[14] The first feature reflects the fact that structures generatable by the language faculty are not necessarily useable, parsable; e.g., The boat the sailor the dog bit built sank, appears to be gibberish, yet the language faculty determines its interpretability just as it does that of The boat that the sailor that the dog bit, built, sank. The second feature is exemplified by ‘illusions’ such as No head injury is too trivial to ignore, which is taken to express the perfectly legitimate meaning, No matter how trivial a head injury is, it should be ignored. The language faculty, however, determines the sentence to be interpretable as No matter how trivial a head injury is, it should not be ignored.
[15] Higginbotham, 1987, like many others after him, insisted that the language faculty is to be understood as represented by the mind. The central part of what Higginbotham means here, however, is simply that the language faculty is not a performance system such as a parser (cf. George, 1989; Knowles, 2000). Higginbotham (e.g., 2001) has since entertained a sort of Platonism about syntactic and semantic structures, where it is the job of psychology to discover how a given speaker/hearer may represent (or realise) such structures. Such Platonism appears to be the kind of reflex mentioned in §3: representations require a represented. Yet, as Higginbotham, 1991, p.556, himself, in an earlier incarnation, put it, the represented is a mere “platonic shadow”, which it is “pedantic” to distinguish. Higginbotham has been concerned with the extent to which the shadow may be more substantial than a conceptual reflex. Internalism, we may say, rules against all species of the representation of the external, even the notional external (cf. Chomsky, 2000a, p.73).
[16] The issues raised in the preceding paragraphs relate to the question of the possibility of error: Can a speaker/hearer go wrong in her linguistic judgements, if the basis of how it seems to her is just how she represents her grammar to be (as ‘internalized’)? See George, 1990; Smith, 1998, 2001; Higginbotham, 1991, 1998; and Barber, 2001. Let it suffice to say that the substantive issue here appears to be independent of the question of whether the faculty is propositional or not. Let us assume, as the problem presupposes, that our propositional judgements partially track the structure of the faculty (under some construal or other), but it wouldn’t follow that the tracked is propositional, only that the tracked contributes, perhaps quite opaquely, to the determination of the kind of judgements available to us. Otherwise put, there is an issue about the relation between our judgements and the faculty, but there is no ‘problem of error’, for the relation is not epistemic.
[17] Suffice it to say, by questioning Fodor’s argument, I don’t intend to endorse the Quine-style argument, which has been variously entertained by Stich, 1972, Soames, 1984, Devitt and Sterelney, 1987, 1989. The Quine argument, and its progeny, are beyond the pale, so much so that no positive thesis is lent any weight by their trivial refutation. See §7.
[18] Fodor’s argument appears to be a reading of §6 of chapter 1 of Chomsky, 1965; §II of Chomsky, 1964, might also be relevant. Fodor’s curious use of ‘true of’, however, might hold the key. See n. 19.
[19] Fodor’s use of ‘true of’ is very curious. It derives from Stich, 1972, as far as I can see, who uses it to suggest that a grammar carries no implication about internal structure. As Chomsky and Katz, 1974, make clear in their response to Stich, the use of the ‘true of’ idiom, so construed, amounts to a mere stipulation that the linguist should restrict his interest so as to exclude developmental and psychological issues; the notion of a grammar G being true of a speaker/hearer has no positive role to play in linguistic theory; a fortiori, it is no part of an argument in defence of a representation/represented distinction.
[20] Fodor (p.c.) reads ‘I-language’ as idiolectal. Notwithstanding any obscurity which may beset the notion, an I-language cannot be an idiolect, for although individual, an idiolect is a not an internal state.
[21] Contrary to
the impression of much pre-standard theory, Chomsky, 1964, p.53, n.4;
1955/1975, p.5/53, n.75, has made it clear that he has never seen weak generative capacity as intrinsically interesting.
Whatever the case might be, once the distinction between grammars and languages
dissolves, it is difficult to make sense of either weak or strong generative
capacity.
[22] My thanks
go to Noam Chomsky, Jerry Fodor, Jonathan Knowles, Guy Longworth, and an
anonymous referee for many helpful suggestions and clarifications.
References
Barber, A.
2001: Idiolectal error. Mind and Language,
16: 263-283.
Chomsky, N.
1955/75: The Logical Structure of
Linguistic Theory. Chicago: Chicago
University Press.
Chomsky, N.
1957: Syntactic Structures. The
Hague: Mouton.
Chomsky, N.
1959: Review of B. F. Skinner’s Verbal
Behaviour. Language, 35: 26-
58. References to the reprint in J.
Fodor and J. Katz (eds.), The Structure
of
Language:
Readings in the Philosophy of Language (pp.547-578). Englewood
Cliffs: Prentice-Hall
Chomsky, N.
1962: Explanatory models in linguistics. In E. Nagel, P. Suppes, and A.
Tarski (eds.), Logic, Methodology and Philosophy of Science (pp.528-550).
Stanford: Stanford University Press.
Chomsky, N.
1964: Current issues in linguistic theory. In J. Fodor and J. Katz (eds.),
The
Structure of Language: Readings in the Philosophy of Language (pp.50-118).
Englewood Cliffs: Prentice-Hall.
Chomsky, N.
1965: Aspects of the Theory of Syntax.
Cambridge, MA: MIT Press.
Chomsky, N.
1967: Preface to "A Review of B. F. Skinner's Verbal Behavior". In L.
A. Jakobovits and M. S. Miron (eds.), Readings in the Psychology of Language
(pp.142-143). Englewood Cliffs:
Prentice-Hall.
Chomsky, N.
1968/72: Language and Mind (enlarged edition). New York:
Harcourt
Brace Jovanovich.
Chomsky, N.
1975: Reflections on Language.
London: Fontana
Chomsky, N.
1980: Rules and Representations. New
York: Columbia University
Press.
Chomsky, N.
1981: Lectures on Government and Binding.
Dordrecht: Foris.
Chomsky, N.
1982: The Generative Enterprise: A
Discussion with Riny Huybregts
and
Henk van Riemsdijk. Dordrecht: Foris.
Chomsky, N.
1986: Knowledge of Language: Its Nature,
Origin, and Use. Westport:
Praeger.
Chomsky, N.
1991: Linguistics and adjacent fields: a personal view. In A. Kasher
(ed.), The Chomskyan Turn (pp.3-25). Oxford: Blackwell.
Chomsky, N.
1994: Chomsky, Noam. In S. Guttenplan (ed.), A Companion to the
Philosophy of Mind (pp.153-167). Oxford: Blackwell.
Chomsky, N.
1995: The Minimalist Program.
Cambridge, MA: MIT Press.
Chomsky, N.
1996: Powers and Prospects: Reflections
on Human Nature and the
Social
Order. London: Pluto Press.
Chomsky, N.
2000a: New Horizons in the Study of
Language and Mind. Cambridge:
Cambridge University Press.
Chomsky, N.
2000b: The Architecture of Language.
New Delhi: Oxford University
Press.
Chomsky, N.
2000c: Linguistics and brain science. In A. Marantz, Y. Miyashita, and
W. O’Neil (eds.), Image, Language, Brain (pp.13-28). Cambridge, MA: MIT
Press..
Chomsky, N.
2000d: Minimalist inquiries: the framework. In R. Martin, D. Michaels,
and J. Uriagereka (eds.), Step by Step: Essays on Minimalist Syntax
in Honor of
Howard
Lasnik (pp.89-155). Cambridge, MA: MIT Press.
Chomsky, N.
2001a: Derivation by phase. In M. Kenstowicz (ed.), Ken Hale: A Life
in
Language (pp.1-52). Cambridge, MA: MIT Press.
Chomsky, N.
2001b: Beyond explanatory adequacy. MIT
Occasional Papers in
Linguistics,
No.20, pp.1-28.
Chomsky, N.
2002: An interview on minimalism. In On
Nature and Language (pp.92-
161). Cambridge: Cambridge University
Press
Chomsky, N.
and Katz, J. 1974: What the linguist is talking about. Journal of
Philosophy
71: 347-367.
Collins, J.
2003: Expressions, sentences, propositions. Erkenntnis
59: 233-262
Collins, J.
forthcoming: Nativism: substantial vs. deflationary approaches.
Philosophical
Psychology.
Cowie, F.
1999: What’s Within? Nativism
Reconsidered. Oxford: Oxford University
Press.
Davies, M.
1986: Tacit knowledge and the structure of thought and language. In C.
Travis (ed.), Meaning and Interpretation (pp.127-158). Oxford: Blackwell.
Davies, M.
1989: Tacit knowledge and subdoxastic states. In A. George (ed.),
Reflections
on Chomsky (pp. 131-152). Oxford: Blackwell.
Devitt, M.
and Sterelny, K. 1987: Language and Reality. Cambridge, MA: MIT Press.
Devitt, M.
and Sterelny, K. 1989: Linguistics: what’s wrong with ‘The Right View’.
In J. Tomberlin (ed.), Philosophical Perspectives 3: Philosophy of
Mind and
Action
Theory (pp. 495-531). Atascadero: Ridgeway.
Evans, G.
1981: Semantic theory and tacit knowledge. In S. Holtzman and C. Leich
(eds.), Wittgenstein: To Follow a Rule (pp. 118-137). London: Routledge.
Fodor, J.
1975: The Language of Thought.
Cambridge, MA: MIT Press.
Fodor, J.
1981a: Some notes on what linguistics is about. In N. Block (ed.), Readings
in the
Philosophy of Psychology, Vol. II (pp.197-207). Cambridge, MA:
Harvard
University Press.
Fodor, J.
1981b: On the present status of the innateness controversy. In
Representations:
Philosophical Essays on the Foundations of Cognitive Science
(pp.257-316). Cambridge, MA: MIT
Press.
Fodor, J.
1983: The Modularity of Mind.
Cambridge, MA: MIT Press.
Fodor, J.
2000: The Mind Doesn’t Work That Way: The
Scope and Limits of
Computational Psychology. Cambridge, MA: MIT Press.
Fodor, J.
2001: Doing without What’s Within:
Fiona Cowie’s critique of nativism.
Mind,
110: 99-148.
Fodor, J.,
Bever, T., and Garrett, M. 1974: The
Psychology of Language: an
Introduction to Psycholinguistics and Generative Grammar. New York:
McGraw-Hill.
Garfield,
J., Peterson, C., and Perry, T. 2001: Social cognition, language
acquisition and the development of the
theory of mind. Mind and Language,
16: 494-541.
George, A.
1989: How not to become confused about linguistics. In A. George (ed.),
Reflections
on Chomsky (pp.90-110). Oxford: Blackwell.
George, A.
1990: Whose language is it anyway? Some notes on idiolects.
Philosophical Quarterly, 40: 275-298.
Hauser, M.,
Chomsky, N., and Fitch, W. T. 2002: The faculty of language: what is it,
who has it, and how did it evolve? Science 298: 1569-1579.
Higginbotham,
J. 1987: The autonomy of syntax and semantics. In J. Garfield
(ed.), Modularity in Knowledge Representation and Natural Language
Understanding (pp.119-131). Cambridge, MA: MIT Press.
Higginbotham,
J. 1991: Remarks on the metaphysics of linguistics. Linguistics and
Philosophy,
14: 555-566.
Higginbotham,
J. 1998: On knowing one’s own language. In C. Wright, B. C. Smith,
and C. McDonald (eds.), Knowing Our Own Minds (pp.429-441).
Oxford:
Oxford University Press.
Higginbotham,
J. 2001: On referential semantics and cognitive science. In J.
Branquinho (ed.), The Foundations of Cognitive Science (pp.145-156). Oxford:
Oxford University Press.
Hornstein,
N. 2001: Move! A Minimalist Theory of
Construal. Oxford: Blackwell.
Jackendoff,
R. 1993: Patterns in Mind: Language and
Human Nature. Harvester
Wheatsheaf: London
Jackendoff,
R. 2002: Foundations of Language: Brain,
Meaning, Grammar,
Evolution. Oxford:
Oxford University Press.
Katz, J.
1966: The Philosophy of Language. New
York: Harper and Row.
Kayne, R.
1994: The Antisymmetry of Syntax.
Cambridge, MA: MIT Press.
Kayne, R.
2002: Pronouns and their antecedents. In S. Epstein and T. D. Seely (eds.),
Derivation
and Explanation in the Minimalist Program (pp.133-166). Oxford:
Blackwell.
Khalidi, M.
A. 2001: Innateness and domain specificity. Philosophical
Studies, 105:
191-210.
Knowles, J.
2000: Knowledge of grammar as a propositional attitude. Philosophical
Psychology,
13: 325-353.
Matthews,
R. 1991: Psychological reality of grammars. In A. Kasher (ed.), The
Chomskyan
Turn (pp. 182-199). Oxford: Blackwell.
Prinz, J.
2002: Furnishing the Mind: Concepts and
Their Perceptual Basis.
Cambridge, MA: MIT Press.
Quine, W.
V. O. 1969: Linguistics and philosophy. In S. Hook (ed.), Language and
Philosophy
(pp.95-98). New York: NYU Press.
Quine, W.
V. O. 1972: Methodological reflections on current linguistic theory. In D.
Davidson and G. Harman (eds.), Semantics of Natural Language (pp.
442-454).
Dordrecht: Reidel
Samuels, R.
2002: Nativism in cognitive science. Mind
and Language, 17: 233-265.
Segal, G.
1996: The modularity of theory of mind. In P. Carruthers and P. Smith
(eds.), Theories of Theories of Mind (pp. 141-157). Cambridge: Cambridge
University Press.
Smith, B.
C. 1998: On knowing one’s own language. In C. Wright, B. C. Smith, and
C. McDonald (eds.), Knowing Our Own Minds (pp.391-428). Oxford: Oxford
University Press.
Smith, B.
C. 2001: Idiolects and understanding: comments on Barber. Mind and
Language,
16: 284-289.
Smith, N.
1999: Chomsky: Ideas and Ideals.
Cambridge: Cambridge University Press.
Soames, S.
1984: Linguistics and psychology. Linguistics
and Philosophy 7: 155-179.
Stemmer, B.
1999: An on-line interview with Noam Chomsky: on the nature of
pragmatics and related issues. Brain and Language, 68: 393-401.
Stich, S.
1972: Grammars, psychology, and indeterminacy. Journal of Philosophy 69:
799-818.
Stich, S.
1978: Beliefs and subdoxastic states. Philosophy
of Science, 45: 499-518.
Weinberg,
A. 1999: A minimalist theory of human sentence processing. In S. Epstein
and N. Hornstein (eds.), Working Minimalism (pp.283-315). Cambridge,
MA:
MIT Press.
Yang, C.
2002: Natural Language and Language
Learning. Oxford: Oxford
University Press.
Zwart, J-W.
2002: Issues relating to a derivational theory of binding. In S. Epstein and
T. D. Seely (eds.), Derivation and Explanation in the Minimalist Program
(pp.269-304). Oxford: Blackwell.