Nativist Language Acquisition

Nativist Language
Acquisition
Abstract: Nativists about language acquisition employ an inference to the best explanation from
the poverty of the stimulus argument (PSA) to motivate their position. I argue that we can agree
that the PSA is resistant to common objections surrounding the primary linguistic data, without
affirming that nativism about language acquisition is, in fact, the best explanation. I present an
alternative account from usage-based language acquisition to show that we can reject the empiricist
learning method on the grounds of the PSA, but have an explanation of language acquisition that
is prima facie as plausible as nativism – inviting the conclusion that nativists must do more than
merely establish the PSA in order to motivate their position.
Keywords: Chomsky, Language Acquisition, Poverty of the Stimulus Argument, Nativism
Nativist Language Acquisition |1
Introduction
“The study of universal conditions that prescribe the form of any human language is
‘grammaire generale’. Such universal conditions are not learned; rather, they provide the
organizing principles that make language learning possible […] By attributing such
principles to the mind, as an innate property, it becomes possible to account for the quite
obvious fact that the speaker of a language knows a great deal that he has not learned”
(Chomsky, 1966: 59).
The quote above captures Chomsky’s position on language acquisition, referred to as ‘Cartesian
Linguistics’, and some of its most salient features. In this essay, I will be concerned with one of
the main motivations and justifications for Cartesian Linguistics, known as the poverty-of-thestimulus argument. The argument is implicit in the above quote; Chomsky states that the
postulation of an innate property concerned with language acquisition enables us to account for
the simple fact that we are generally able to acquire and manipulate complex linguistic principles,
with seemingly little instruction, evidence or exposure1. My purpose in this paper is to present
and explicate this argument carefully, defend it from initial objections and discuss an objection
that I think has the potential to undermine the argument. In §1 I present and briefly qualify my
formulation of the PSA. In §2 I discuss an objection that I believe can be overcome by the
argument as I formulate it. Finally in, §3, I discuss a further objection that may prove more
difficult to refute.
§1: The Argument
Before presenting the argument there are a few points to note. Firstly, even a sound poverty of
the stimulus argument (PSA) would not be sufficient to establish nativism as a general theory
about the structure of the mind. It is entirely plausible to be a nativist about language acquisition
and nothing more (Simpson: 2005)2. At most, a persuasive PSA will be a specific instance of an
innate module, inviting the view that postulating innate modules is a fruitful and plausible option
when considering the structure of the mind. Secondly, the argument is focused on rejecting an
‘empirical learning method’ of language acquisition, which stands opposed to a nativist model of
language acquisition. Although I borrow the phrase from Laurence & Margolis (2001), what
exactly this method involves is a matter of debate. I will discuss this debate, and its implications,
in §3. Finally, the sum of these points is that the PSA has a weak conclusion, it does not indicate
nativism in its general form, nor does it guarantee a nativist module of language acquisition.
For a more detailed account of Chomsky’s position on the PSA see Chomsky (1967).
Equally, as Simpson notes, a sound argument against a particular aspect of cognition, for example language
acquisition, serves only as to cast doubt on that particular aspect of cognition, not nativism as a theory generally.
Although such an argument will inevitably be damaging to nativism generally, we are perhaps at liberty to be
piecemeal about the structure of the mind.
1
2
Nativist Language Acquisition |2
It does reject one model of language acquisition and, should its premises be true, gives a strong
indication of a move towards a nativist account of language acquisition.
1. Suppose that language was acquired by an empiricist learning method.
2. If language was acquired by an empiricist learning method then one set of grammatical
principles would register as correct against an indefinite number of possible grammatical
principles.
3. If one set of grammatical principles would register as correct against an indefinite
number of possible grammatical principles, then it must be the case that the Primary
Linguistic Data (PLD) that a child is exposed to contains enough positive and negative
evidence to fix a certain set of grammatical principles to a particular language.
4. The PLD does not contain enough positive and negative evidence to fix a certain set of
grammatical principles to a particular language.
5. Thus, it is not the case that one set of grammatical principles would stand out as correct
against an indefinite number of possible grammatical principles. (3, 4)
6. One set of grammatical principles would register as correct against an indefinite number
of possible grammatical principles. (1,2)
7. A contradiction and the supposition must be false. (5,6)
8. Language is not acquired by an empiricist learning method. (1,7)
As noted above, from this conclusion a further step is required to yield a nativist position,
namely an abductive inference to the postulation of a domain-specific, innate language module.
To see this, consider the production of whiskey at a distillery; suppose that we see barley and
empty bottles going in one door of the distillery and crates of whiskey coming out the other. We
know that water is needed to make whiskey, so, given the output of whiskey but no input of
water we infer to the best explanation; that the water is already present inside the distillery (Cook &
Newson, 2007: 53). The same inference is made here, the PLD is the input and the grammatical
principles are the output, but we know that the PLD is lacking the ‘ingredients’ to afford the
output, thus we postulate the existence of an innate, domain-specific, language acquisition
module to do that work.
§2: The Primary Linguistic Data
The premise that stands out for obvious criticism in the PSA is Premise (4). It is certainly a (if
not the) key premise of the argument. In this section I will discuss a major objection that seeks to
deny Premise (4) by rejecting the claim that PLD is insufficient for language acquisition. I will
argue that it does not do much work towards rejecting premise (4) and an argument against PSA,
as I have presented it, would do better to allow this seemingly controversial premise.
Nativist Language Acquisition |3
The first comes from Sampson (2007: 41) who argues that the PLD is much richer than
Chomsky allows, perhaps to the point that it is sufficient to fix a certain set of grammatical
principles to a particular language citing research into the sheer volume and variation of
sentences that a child is likely to hear in the early stages of their life3. The main point here is that
the PSA seriously underestimates the PLD.
A reply to this objection should be apparent due to the flexibility of premise (4); recall, it notes
that the positive and negative evidence is insufficient for language acquisition from the PLD. A
novel aspect of Chomsky’s account is to stress that in order to learn the principles of grammar
there would need to be evidence available in the PLD of incorrect grammar usage as well as
correct usage.4 With this in mind, it would be possible to compile PLD with thousands of
instances of positive evidence, yet still hold that the PLD is insufficient to fix a certain set of
grammatical principles to a particular language due to a lack of negative evidence5.
Fiona Cowie (1999) has a counter-reply to this in questioning whether the negative evidence
contained in the PLD really is as insufficient as the PSA purports it to be. It is suggested that
when considering linguistic rules a child is ‘faced’ with two hypotheses:
H1 – A structure-independent rule: ‘find the first occurrence of ‘is’ in the sentence and move it
to the front of the sentence.
H2 – A structure-dependent rule: ‘find the first occurrence of ‘is’ following the subject nounphrase of the sentence and move to the front of the sentence. (Cowie, 1999: 178)6.
The point at issue here, then, is whether the PLD contains sufficient falsifying instances of H1 in
order to infer the correct rule H2. Drawing on research compiled by Sampson (1989) and Pullum
(1996) Cowie argues that, contrary to the PSA, we should expect the PLD to contain falsifying
instances of principles like H1.7
See also Pullum & Scholz (2002) for a comprehensive discussion of the volume and variation of sentences that a
child is likely to hear in the early stages of their life.
4 This point is often made by way of analogy to learning the rules of a game just by watching, for example Cook &
Newson (2007: 189) make the point using a snooker analogy, “[the observer] would not be able to tell if a new
sequence, say two red balls in succession, was illegal, or one he had by chance not encountered before […] an
adequate knowledge of snooker involves knowing what not to do as well as what to do.”
5 You might also think that this line of objection is not going to be particularly fruitful from the outset, the case
from lack of positive evidence is a fairly routine instance of theory under-determination by evidence, a much wider
issue than the PSA needs to deal with.
6 These are possible rules for learning how to form polar interrogatives (yes-no questions) from declarative
sentences.
7 Cowie argues for a general position that she dubs ‘enlightened empiricism’ which, she argues, maintains some weak
nativist principles (such as Domain Specificity and Biological Boundedness) but rejects universal grammar and
(possibly) innately encoded constraints, see Cowie (1999, 176ff).
3
Nativist Language Acquisition |4
It is tempting to see this dialectic as both positions ‘digging their heels in’ with regards to the
content of PLD. One point of convergence in the literature with regards to PLD is that linguists
simply do not know how poor the PLD is. John Collins (2003) uses this point of convergence as a
starting point to argue that the PSA requires the least of the PLD; much to its advantage. Collins
(2003: 170) notes that PLD, regardless of content, is certainly not uniform, it is wide and varied
across individual children’s situations. So, with this in mind, the PSA claims that, “the pld is poor
relative to what it must be like if children were learning the language from degraded and variable
data.” (loc.cit.) It does not matter whether falsifying instances of H1 are available to lots of
children, the fact is, they are not available to all (or most) children and yet all (or most) children
do reject H1 in favour of H2. Empiricism treats the PLD as a fixed body of data with sufficient
content to make (or not make) inferences to grammar rules but this picture is false – and what’s
more, most agree that it is false. So, the argument concludes, nativism requires the least from the
PLD, it can account for the fact that the PLD wide and varied but yet in general children do
acquire the abstract principles of their language’s grammar. I think Collins’s reply and general
line of argument can successfully defend the PSA and in light of this, an opponent of nativism
would be better to allow premise (4) and switch the focus to another objection. In the section
below I will explore this option.
§3: Inference to the Best Explanation?
With two objections to premise (4) contested I will consider an objection that I think does
threaten to undermine the PSA and nativism about language acquisition more generally. It begins
by claiming that the PSA has chosen the wrong target in attacking the ‘empiricist learning
method’ (ELM) and, as a result, achieves little in establishing nativism about language
acquisition.
To start with we might wonder exactly what the ELM is. From the discussion in §2 it might
already be clear that PSAs characterise the ELM as a Popparian approach to language acquisition
– that is, entertaining possible grammatical principles (or particular sentences) and looking for
falsifying instances of possible principles. Once explicated in this way it should seem pretty
unremarkable that a Popparian approach to language acquisition fails to be compelling, naïve
falsificationism of this sort has been left behind in the philosophy of science and replaced by a
more sophisticated falsificationism or rejected altogether8.
Noting the weakness of the PSA’s characterisation of ELM is a mere reiteration of the claim I
made in §1; the PSA has a weak conclusion. The PSA still stands, but what is certainly up for
debate, given that the PSA merely argues against a weak ELM (possibly even a straw man), is
whether the postulation of an innate language acquisition module really is the inference to the
8
For a discussion of these issues see Lakatos & Musgrave (1970)
Nativist Language Acquisition |5
best explanation. Michael Tomasello (2003) has developed a usage-based theory of language
acquisition which purports to replace innate modules as the best explanation of the poverty of
the stimulus. Note how, unlike the objections in §2, this approach can accept the PSA but argue
that there is a better way to make sense of the PLD than a move towards nativism about
language acquisition. Usage-based theories of language acquisition are often summarised in
slogan form as ‘meaning is use’.
Taking inspiration from Wittgenstein, they argue that the semantics of natural languages are
contingent on the usage of linguistic communities and, “the essence of language is its symbolic
dimension, with grammar being derivative (Tomasello ibid: 5). Tomasello (ibid: 67) postulates two
sets of cognitive skills that are involved in language acquisition; ‘intention-reading’ and ‘pattern
finding’, both of which are (crucially) domain-general rather than domain-specific9.
The merits and difficulties involved in adopting a usage-based theory of language acquisition are
beyond the scope of this paper. The relevant point to take from this discussion is that the PSA is
required to do more than reject an ELM and subsequently claim nativism to be the inference to
the best explanation. We can agree that an ELM, as presented by the PSA, is unpersuasive but
still vastly disagree on how best to make sense of the PLD, or lack thereof. Unlike the objection
in §2, this objection is not contesting the quality of the PLD, but what it is that children actually
do with the PLD available to them – whether it be rich or poor. An interesting point to note,
and perhaps a starting point for reply, is that the PSA is also going to support the ‘theory-theory’
of mind reading and it seems as though usage-based theories will fit more with ‘simulation’
theories of mind reading – the relative pros and cons of these positions may well impact upon
how inclined you might be towards pursuing nativism or rejecting it in favour of usage-based
accounts. The purpose of this section has been to indicate an approach which may well serve to
undermine the PSA, should it be more appealing either theoretically, such as a commitment to
‘simulation theories’, or empirically, or both. Accordingly, I leave the question open as to how
effective this line of objection will ultimately be.
Conclusion
My purpose in this essay has been to present a PSA, defend it from initial objections and discuss
an objection that I think has the potential to undermine the argument. In §1 I presented and
briefly qualified my formulation of the PSA. In §2 I discussed some objections that I believe can
be overcome by the argument as I have presented it. Finally in, §3, I discussed a further
objection that may prove more difficult to refute.
i.e. they are general cognitive functions that play a role in language acquisition rather than nativist, domain-specific,
language acquisition modules.
9
Nativist Language Acquisition |6
Bibliography
Chomsky, N. (1966). ‘Cartesian Linguistics: Acquisition and Use of Language’, in Stephen, P.
Stich (ed.) (1975). Innate Ideas, Berkley: University of California Press.
___________(1967). ‘Recent Contributions to the Theory of Innate Ideas’, in Stephen, P. Stich
(ed.) (1975). Innate Ideas, Berkley: University of California Press.
Collins, J. (2003). ‘Cowie on the poverty of stimulus’, Synthese 136 (2):159-190.
Cook, V. J. & Newson, M. (2007). Chomsky’s Universal Grammar: An Introduction, 3rd Edition,
Maldon, MA: Blackwell Publishing.
Cowie, F. (1999). What’s Within: Nativism Reconsidered, Oxford: Oxford University Press.
Laurence, S. & Margolis, E. (2001). ‘The Poverty of the Stimulus Argument’, in The British Journal
for the Philosophy of Science 52: 217-276.
Lakatos, I. & Musgrave, A. (eds.) (1970). Criticism and the Growth of Knowledge, Cambridge:
Cambridge University Press
Pullum, G. (1996). ‘Learnability, Hyperlearning, and the Poverty of Stimulus’, in J. Johnson, M.
Junge, & J. Moxley (eds.), Proceedings of the 22nd Annual Meeting: General Session and Parasession on the
Role of Learnability in Grammatical Theory, Berkeley: Berkeley Linguistics Society.
Pullum, G. & Scholz, B. (2002). ‘Empirical Assessment of Stimulus Poverty Arguments’, The
Linguistic Review 19: 9-50.
Putnam, H. (1967). ‘The Innateness Hypothesis and Explanatory Models in Linguistics’, in
Stephen, P. Stich (ed.) (1975). Innate Ideas, Berkley: University of California Press.
Sampson, G. (1989). ‘Language Acquisition: Growth or Learning?’, Philosophical Papers 18: 203240.
__________ (2007). ‘There is No Language Instinct’, Ilha do Desterro 52: 35-63.
Simpson, T. (2005). ‘Towards a Reasonable Nativism’, in P. Carruthers & S. Laurence & S. Stich
(eds.) The Innate Mind: Structure and Contents, Oxford: Oxford University Press.
Tomasello (2003). Constructing a Language: A Usage-Based Theory of Language Acquisition, Cambridge
MA, Harvard University Press.