1

1010The Poverty of The Poverty of the Stimulus the Stimulus

ArgumentArgument

2

General dataGeneral data

Except for congenital defect or trauma we Except for congenital defect or trauma we all end up using (at least) a all end up using (at least) a particularparticular language, although we language, although we mightmight have ended up have ended up using using any otherany other language. language.

Our brains, unlike those of other species, Our brains, unlike those of other species, are such as to enable us to acquire language are such as to enable us to acquire language as suchas such, although they are not primed to , although they are not primed to acquire any particular language.acquire any particular language.

3

The acquisition of language is species The acquisition of language is species specific.specific.

Whatever distinguishes us from other Whatever distinguishes us from other animals must be specific enough (animals must be specific enough (notnot necessarily specific to language) for us to necessarily specific to language) for us to arrive at English or Italian or Navajo, etc.arrive at English or Italian or Navajo, etc.

It must also be general enough to target It must also be general enough to target any language with equal ease.any language with equal ease.

4

InnatenessInnateness

Humans possess innate equipment, Humans possess innate equipment, whether specific to language or not, that whether specific to language or not, that enables them to acquire any language. enables them to acquire any language.

So far, then, we don’t have any argument So far, then, we don’t have any argument for the claim that the human child begins for the claim that the human child begins with something specifically linguistic. Some with something specifically linguistic. Some other, species specific, capacity could do other, species specific, capacity could do the job.the job.

5

The languages we speak are very different. The languages we speak are very different.

There might be universal features shared by all There might be universal features shared by all languages, but they are not apparent in the languages, but they are not apparent in the seemingly infinite variety of data to which seemingly infinite variety of data to which children are exposed. children are exposed.

What else does the child have other than the What else does the child have other than the data? data?

It It seems seems that we are infinitely far from the that we are infinitely far from the explanatoryexplanatory ideal situation, i.e., the more ideal situation, i.e., the more languages there are, the more inclusive must languages there are, the more inclusive must be our initial capacity to represent language.be our initial capacity to represent language.

6

TrivialityTriviality..

That the child begins with innate equipment That the child begins with innate equipment is true enough, but we seem to require is true enough, but we seem to require something decidedly less trivial.something decidedly less trivial.

What the child’s innate equipment is required What the child’s innate equipment is required

to actively constrains its ‘choices’ as to what to actively constrains its ‘choices’ as to what is part of the language to be attained.is part of the language to be attained.

But no child is wired to target any particular But no child is wired to target any particular language: the child can make the right language: the child can make the right ‘choices’ about any language with equal ease. ‘choices’ about any language with equal ease.

7

Initial stageInitial stage

The children must begin with ‘knowledge’ The children must begin with ‘knowledge’ specific to language, i.e., the data to which the specific to language, i.e., the data to which the child is exposed is ‘understood’ in terms of child is exposed is ‘understood’ in terms of prior linguistic concepts as opposed to general prior linguistic concepts as opposed to general concepts of pattern or frequency, say.concepts of pattern or frequency, say.

E.g.: children distinguish phonemes from E.g.: children distinguish phonemes from rumours.rumours.

8

Poverty of the stimulusPoverty of the stimulus

A child may acquire a language even though A child may acquire a language even though the data itself is too poor to determine the the data itself is too poor to determine the language: the child needs no evidence for language: the child needs no evidence for much of the knowledge she brings to the much of the knowledge she brings to the learning situation. learning situation.

Children acquire language from Children acquire language from pidginpidgin..

Roughly, children always make the right Roughly, children always make the right ‘hypotheses’ as a function of their genetic ‘hypotheses’ as a function of their genetic endowment.endowment.

9

Since the child can fixate on Since the child can fixate on anyany language language in the face of a poverty of stimulus about in the face of a poverty of stimulus about eacheach language and since all languages are language and since all languages are equally acquirable, children equally acquirable, children all begin all begin with the same universal linguistic with the same universal linguistic knowledgeknowledge. .

This is the essence of This is the essence of the poverty of the poverty of stimulus argumentstimulus argument. .

10

The poverty of the stimulus argument The poverty of the stimulus argument does does notnot telltell us: us:

1.1. What information is innate.What information is innate.

2.2. How the innate information is represented How the innate information is represented in the in the mind/brain.mind/brain.

3.3. Whether the information is available to a Whether the information is available to a general general learning mechanism or specific to a learning mechanism or specific to a dedicated one dedicated one (i.e. general intelligence or (i.e. general intelligence or language module). language module).

These issues are to be decided by the These issues are to be decided by the normal normal scientific routescientific route of the testing and comparison of of the testing and comparison of hypotheses. hypotheses.

11

Positive dataPositive data tells the child that some tells the child that some construction is construction is acceptableacceptable..

Negative dataNegative data tells the child that some tells the child that some construction is construction is unacceptableunacceptable. .

There is much discussion of this difference, There is much discussion of this difference, for it has been claimed that negative for it has been claimed that negative evidence is typically unavailable and not evidence is typically unavailable and not used by the child even where it is available.used by the child even where it is available.

12

Children are innately constrained to Children are innately constrained to initiallyinitially ‘chose’ the smallest possible ‘chose’ the smallest possible language compatible with their language compatible with their positive data.positive data.

13

Much of the debate around the Poverty of Much of the debate around the Poverty of the Stimulus Argument focuses on the Stimulus Argument focuses on negative evidence.negative evidence.

If there is lot of negative evidence there If there is lot of negative evidence there are more chances that the child’s learning are more chances that the child’s learning is based on trial and error.is based on trial and error.

Even if there is plenty of negative data Even if there is plenty of negative data (which is questionable), the Poverty of the (which is questionable), the Poverty of the Stimulus Argument is not refuted.Stimulus Argument is not refuted.

14

The relative neutrality of the Poverty of the The relative neutrality of the Poverty of the Stimulus Argument suggests something Stimulus Argument suggests something surprising: the fact that the child can acquire surprising: the fact that the child can acquire any language without seemingly enough data to any language without seemingly enough data to do so, indicates, counter-intuitively, that do so, indicates, counter-intuitively, that languages are not so different. languages are not so different.

The innate ‘hypotheses’ the children employ The innate ‘hypotheses’ the children employ must be universal, rather than language must be universal, rather than language particular. particular.

Imagine that each language is radically distinct, Imagine that each language is radically distinct, an effect of a myriad of contingent historical an effect of a myriad of contingent historical and social factors. This seems to be what the and social factors. This seems to be what the pursuit of descriptive adequacy tells us. Now, if pursuit of descriptive adequacy tells us. Now, if this were the case, then the child’s data would this were the case, then the child’s data would still be poor.still be poor.

15

But how would innate knowledge help here? But how would innate knowledge help here?

Since, Since, ex hypothesiex hypothesi, each language is as , each language is as distinct as can be, there is no generality distinct as can be, there is no generality which might be encoded in the child’s brain.which might be encoded in the child’s brain.

That is, the child would effectively have to That is, the child would effectively have to have separate innate specific knowledge have separate innate specific knowledge about each of the indefinite number of about each of the indefinite number of languages it might acquire. languages it might acquire.

16

This is just to fall foul of the Poverty of the This is just to fall foul of the Poverty of the Stimulus Argument: how does the child know Stimulus Argument: how does the child know that the language it is exposed to is a sample that the language it is exposed to is a sample of grammar X as opposed to any of the other of grammar X as opposed to any of the other grammars?grammars?

The best explanationThe best explanation..

The specific conjecture is that The specific conjecture is that we all begin we all begin with with universal grammaruniversal grammar (UG), the one (UG), the one language, as it were. language, as it were.

UG is innate and is informed in the sense that UG is innate and is informed in the sense that it encodes certain options or it encodes certain options or parametersparameters which are set by exposure to certain data.which are set by exposure to certain data.

17

To acquire a language is simply for the To acquire a language is simply for the values of UG’s parameters to be set in one values of UG’s parameters to be set in one of a finite number of permutations (given of a finite number of permutations (given the acquisition of a lexicon.)the acquisition of a lexicon.)

Chomsky understands Chomsky understands UG to be the UG to be the initial state of the initial state of the language facultylanguage faculty (an abstractly specified system of the (an abstractly specified system of the brain.) brain.)

18

To acquire a language is to To acquire a language is to acquire a particular acquire a particular systematic mapping between sound and systematic mapping between sound and meaningmeaning. .

How do we fixate on such a pairing? How do we fixate on such a pairing?

Think of the language faculty as a genetically Think of the language faculty as a genetically determined determined initial stateinitial state prior to experience. prior to experience.

Experience Experience triggerstriggers the setting of values along the setting of values along certain parameters that determine the output certain parameters that determine the output conditions. conditions.

Experience also provides the assignment of Experience also provides the assignment of features in the lexicon, although not the features features in the lexicon, although not the features themselves. themselves.

19

From the Initial State to I-LanguageFrom the Initial State to I-Language

Different experiences set the parameters to Different experiences set the parameters to different values (cf. switch analogy).different values (cf. switch analogy).

This finite variation ramifies to produce This finite variation ramifies to produce languages of seemingly infinite variety. languages of seemingly infinite variety.

Once all parameters are set, the faculty Once all parameters are set, the faculty attains a attains a steady statesteady state we call an we call an I-languageI-language..

I-language is a generative system which I-language is a generative system which explains an individual’s competence with her explains an individual’s competence with her idiolect.idiolect.

20

UG is not implied by the above general UG is not implied by the above general reasoning about acquisition in the face of the reasoning about acquisition in the face of the poverty of the stimulus. poverty of the stimulus.

It is, rather, a somewhat speculative It is, rather, a somewhat speculative hypothesis based upon a myriad of hypothesis based upon a myriad of considerations, both empirical and considerations, both empirical and theoretical. theoretical.

The form of the Poverty of the Stimulus The form of the Poverty of the Stimulus Argument is quite general and based on Argument is quite general and based on what Chomsky has called what Chomsky has called Plato’s problemPlato’s problem. .

The problem occurs The problem occurs whereverwherever a a competence is exhibited which we have competence is exhibited which we have apparently too little data to acquireapparently too little data to acquire..

21

The Poverty of the Stimulus Argument is The Poverty of the Stimulus Argument is notnot employed in employed in directdirect defence of UG defence of UG (under some proprietary specification).(under some proprietary specification).

On the contrary, On the contrary, UG is supported to the UG is supported to the extent that it is the best theory of the extent that it is the best theory of the knowledge which the Poverty of the knowledge which the Poverty of the Stimulus Argument tells us existsStimulus Argument tells us exists..

UG is a scientific hypothesis.UG is a scientific hypothesis.

22

What must a child know such that it can What must a child know such that it can correctly go from this kind of data to the correctly go from this kind of data to the correct interrogative form in general? correct interrogative form in general?

(1)(1) a. a. That man is happyThat man is happyb.b. Is that man happy?Is that man happy?

(2)(2) a. a. That man can singThat man can singb.b. Can that man sing?Can that man sing?

Chomsky asked this question as a challenge Chomsky asked this question as a challenge to Putnam, who had contended that the to Putnam, who had contended that the child need only have at her disposal child need only have at her disposal general principles (not domain specific general principles (not domain specific linguistic ones). linguistic ones).

23

The empiricist challenge.The empiricist challenge.

SI:SI: Go along a declarative until you Go along a declarative until you come to the come to the first ‘is’ (or, ‘can’, first ‘is’ (or, ‘can’, etc.) and move it to the etc.) and move it to the front of the front of the sentence.sentence.

SI is SI is structure independentstructure independent in that it in that it appeals merely to the morphology and appeals merely to the morphology and linear order of the declarative.linear order of the declarative.

24

The important point here is that an The important point here is that an empiricist may happily appeal to SI as the empiricist may happily appeal to SI as the rule upon which the child fixates, for it rule upon which the child fixates, for it involves no linguistic concepts and so is involves no linguistic concepts and so is one at which a child may arrive without one at which a child may arrive without the benefit of specific linguistic the benefit of specific linguistic knowledge. knowledge.

25

Now the child would proceed correctly Now the child would proceed correctly with SI so long as she continued to meet with SI so long as she continued to meet such monoclausal constructions as such monoclausal constructions as (1)+(2). (1)+(2).

(1)(1) a. That man is happya. That man is happyb. Is that man happy?b. Is that man happy?

(2)(2) a. That man can singa. That man can singb. Can that man sing?b. Can that man sing?

26

But the rule But the rule does not generalisedoes not generalise..

(3)(3) a. That man who is blonde is happya. That man who is blonde is happy

Application of SI would produce the Application of SI would produce the nonsensicalnonsensical

(3)(3) b. * Is that man who blonde is happy?b. * Is that man who blonde is happy?

(3)(3) c. [NP That man [CP who is blonde]] is c. [NP That man [CP who is blonde]] is happy happy

27

It is unreasonable to assume that, for a It is unreasonable to assume that, for a child to fixate on a rule R, it needs child to fixate on a rule R, it needs exposure to all the distinct types of exposure to all the distinct types of construction to which R applies, i.e., all construction to which R applies, i.e., all those construction types which would those construction types which would refute potential prior hypotheses of ‘false’ refute potential prior hypotheses of ‘false’ rules.rules.

28

This is the gift of Plato’s point in the This is the gift of Plato’s point in the Meno.Meno.

There is nothing in particular being There is nothing in particular being withheld from the slave boy, but he arrives withheld from the slave boy, but he arrives at an understanding of Pythagoras’ theorem at an understanding of Pythagoras’ theorem on the basis of data that would not be on the basis of data that would not be sufficient were he relying on just that data. sufficient were he relying on just that data.

Hence, we conclude (Hence, we conclude (nonnon-demonstratively) -demonstratively) that he has prior knowledge about the that he has prior knowledge about the domain. domain.

29

The Rarity of Negative Evidence.The Rarity of Negative Evidence.

The kind of negative evidence putatively The kind of negative evidence putatively exploited by children is very weak, only exploited by children is very weak, only appears in mothers with young children. appears in mothers with young children.

Crucially, the relatively rich mother-child Crucially, the relatively rich mother-child interaction observed is typical of the Western interaction observed is typical of the Western middle-class, but it is far from universal. middle-class, but it is far from universal.

The The factfact that children acquire normal that children acquire normal competence without negative evidence competence without negative evidence showsshows that the children who do have it do not need that the children who do have it do not need it.it.

30

There is no need of negative evidence.There is no need of negative evidence.

This is corroborated by the fact that there is This is corroborated by the fact that there is no correlation between negative evidence no correlation between negative evidence supplied by an attentive mother and the supplied by an attentive mother and the rapid acquisition of mature competence.rapid acquisition of mature competence.

So, (i) children don’t require negative So, (i) children don’t require negative evidence and (ii) even when they have it, evidence and (ii) even when they have it, they don’t use it. they don’t use it.

This observation is also supported by a This observation is also supported by a wealth of anecdotal data on the sheer wealth of anecdotal data on the sheer recalcitrance of children’s errors.recalcitrance of children’s errors.

31

Children’s errorsChildren’s errors..

All the data we have indicate that children’s All the data we have indicate that children’s errors (morphological, semantic, syntactic) errors (morphological, semantic, syntactic) are quite rare, certainly rarer than they are quite rare, certainly rarer than they would be were the child seeking to falsify or would be were the child seeking to falsify or test initial hypotheses. test initial hypotheses.

Moreover, the errors made are neither Moreover, the errors made are neither random nor occur equally for all random nor occur equally for all constructions. For example, usually, children constructions. For example, usually, children (as well as adults, of course) make (as well as adults, of course) make regularisation errors with the past tense regularisation errors with the past tense affix -ed.affix -ed.

32

It is very difficult to talk sensibly about It is very difficult to talk sensibly about children’s errors in the absence of an children’s errors in the absence of an acquisition model, for, whether rare or legion, acquisition model, for, whether rare or legion, the pattern of errors remains unexplained. the pattern of errors remains unexplained.

A theory of language acquisition must explain A theory of language acquisition must explain what we get ‘right’ just as much as what we what we get ‘right’ just as much as what we get ‘wrong’. get ‘wrong’.

As the specific complexity of our competence As the specific complexity of our competence leads to a theory of UG, so the specific leads to a theory of UG, so the specific systematicity of our errors leads to the systematicity of our errors leads to the thought that we are not, in general, falsifying thought that we are not, in general, falsifying hypotheses. hypotheses.

33

The mere existence of errors doesn’t militate for The mere existence of errors doesn’t militate for empiricism, or, rather, some as yet unspecified empiricism, or, rather, some as yet unspecified learning regime based on general principles. learning regime based on general principles.

The crucial issue is how errors are explained, The crucial issue is how errors are explained, and there are many ways of classifying and and there are many ways of classifying and explaining errors that are perfectly consistent explaining errors that are perfectly consistent with the nativist stance. with the nativist stance.

E.g.: a child’s errors should be consistent with E.g.: a child’s errors should be consistent with some parametric value of UG, i.e., the errors are some parametric value of UG, i.e., the errors are only relative to the target language, not UG.only relative to the target language, not UG.

34

Motherese and EmpiricismMotherese and Empiricism..

It provides an initial framework from which It provides an initial framework from which the child may proceed to abstract statistically the child may proceed to abstract statistically syntactic categories. syntactic categories.

The unpopularity of the Motherese The unpopularity of the Motherese hypothesis has two principal sources:hypothesis has two principal sources:

1.1. Motherese is not a universal Motherese is not a universal phenomenon: some phenomenon: some cultures and cultures and communities either lack Motherese all communities either lack Motherese all together - parents speak to their children together - parents speak to their children with no with no peculiar prosody - or parents peculiar prosody - or parents actually tend not to actually tend not to talk to their children talk to their children much at all; even so, the much at all; even so, the children acquire children acquire their respective languages perfectly their respective languages perfectly well.well.

35

2.2. Differential exposure to Motherese is not Differential exposure to Motherese is not correlated with differential rates of correlated with differential rates of

language language acquisition. acquisition.

Whatever Motherese is Whatever Motherese is forfor, it does not appear , it does not appear to have a decisive role in language to have a decisive role in language acquisition.acquisition.

Prosody, especially that of Motherese, Prosody, especially that of Motherese, mightmight reflect word boundaries, but it is far from reflect word boundaries, but it is far from clear if phrasal boundaries are reflected (see clear if phrasal boundaries are reflected (see e.g. Pinker). e.g. Pinker).

36

In effect, then, what the child must be able to In effect, then, what the child must be able to do, if she is to progress from words to do, if she is to progress from words to phrases, is recognise that phrases, is recognise that DaddyDaddy, as it might , as it might be, is the be, is the headhead of a subject NP, but this is of a subject NP, but this is something that looks not to be either something that looks not to be either phonetically or morphologically marked. phonetically or morphologically marked.

The child may analyse (parse) its input The child may analyse (parse) its input stream, but to do so the child requires some stream, but to do so the child requires some structural constraints (phrase structural constraints (phrase bracketings/parsing) specific to language and bracketings/parsing) specific to language and there is no data to suggest that this is there is no data to suggest that this is encoded in the input.encoded in the input.

37

Semantic Bootstrapping: Abstraction Semantic Bootstrapping: Abstraction vsvs. . InnatenessInnateness

Semantic bootstrapping refers to the Semantic bootstrapping refers to the hypothesis that children utilize conceptual hypothesis that children utilize conceptual knowledge to create grammatical categories knowledge to create grammatical categories when they’re acquiring their mother tongue. when they’re acquiring their mother tongue.

E.g.: categories like “type of object/person” E.g.: categories like “type of object/person” maps directly onto the linguistic category maps directly onto the linguistic category “noun” while category like “action” onto “noun” while category like “action” onto “verb”, etc. “verb”, etc.

This helps children start on their way to This helps children start on their way to acquiring part of speech.acquiring part of speech.

38

The hypothesis received support from the The hypothesis received support from the experiments that showed that three-to five-experiments that showed that three-to five-year-olds do, in fact, generally use nouns for year-olds do, in fact, generally use nouns for things and verbs for actions more often than things and verbs for actions more often than adults do.adults do.

Theta-roles are understood to be innate. Theta-roles are understood to be innate.

If not the child would have to hypothesise If not the child would have to hypothesise along, ‘All along, ‘All objectsobjects are named by count nouns’. are named by count nouns’.

Where does Where does objectobject come from? (See Fodor come from? (See Fodor 1998. 1998. Concepts: Where Cognitive Science Concepts: Where Cognitive Science Went WrongWent Wrong, OUP. ch. 3)., OUP. ch. 3).

39

Since the bootstrapping mechanism need Since the bootstrapping mechanism need not be understood as a property of UG it not be understood as a property of UG it doesn’t challenge the nativist hypothesis. doesn’t challenge the nativist hypothesis.

Bootstrapping could be construed as a Bootstrapping could be construed as a separate mechanism that maps semantic separate mechanism that maps semantic properties onto the syntax proper. properties onto the syntax proper.

Bootstrapping offers no reason to favour Bootstrapping offers no reason to favour a statistical model of learning rather a statistical model of learning rather than a rule-constraint based one.than a rule-constraint based one.

40

MoralMoral::

Bootstrapping doesn’t seem to call into doubt Bootstrapping doesn’t seem to call into doubt the rationalist (anti-empiricist) claim that the rationalist (anti-empiricist) claim that syntactic categories are not learned by syntactic categories are not learned by abstraction.abstraction.

The poverty of stimulus argument doesn’t The poverty of stimulus argument doesn’t necessarily demonstrate the falsity of necessarily demonstrate the falsity of empiricism.empiricism.

It is not, though, a question of demonstration. It is not, though, a question of demonstration.

Like in any other science these are empirical Like in any other science these are empirical and theoretical considerations.and theoretical considerations.

41

It is not good enough to talk vaguely of a It is not good enough to talk vaguely of a mechanism that has a “preference for mechanism that has a “preference for rules stated in terms of unobservables over rules stated in terms of unobservables over those stated in terms of observables” those stated in terms of observables” (Cowie 1999. (Cowie 1999. What’s Within: Nativism What’s Within: Nativism ReconsideredReconsidered, OUP: 189). , OUP: 189).

It is not as if any old unobservables will do.It is not as if any old unobservables will do.

The constraint is quite specific. We want to The constraint is quite specific. We want to know specifically how the child can have a know specifically how the child can have a “preference” for ‘rules’ involving, say, “preference” for ‘rules’ involving, say, subject NP and matrix auxiliary verb.subject NP and matrix auxiliary verb.

42

The question is straightforwardly The question is straightforwardly empirical.empirical.

There There isis evidence that the child is able evidence that the child is able statistically to recover some information statistically to recover some information from phonetic streams, but there is from phonetic streams, but there is nono evidence that the child can statistically evidence that the child can statistically induce syntactic categories.induce syntactic categories.

43

RulesRules

Are epiphenomena: they are neither Are epiphenomena: they are neither formulated, nor represented, nor tested by formulated, nor represented, nor tested by the learner; nor are they theoretical the learner; nor are they theoretical postulates.postulates.

We can talk about rules, but only for We can talk about rules, but only for taxonomic convenience. taxonomic convenience.

It is thus simply false that Chomsky or others It is thus simply false that Chomsky or others think of a given grammatical rule as crucial; think of a given grammatical rule as crucial; it is a mere taxonomic effect, whose it is a mere taxonomic effect, whose interpretation and explanation can changed interpretation and explanation can changed radically with the development of linguistics.radically with the development of linguistics.

44

Linguistics Linguistics per se per se is not in the business of is not in the business of refuting empiricism.refuting empiricism.

Linguistics attempts to construct theories Linguistics attempts to construct theories that, as in any other science, have that, as in any other science, have universal scope, economy, and predictive universal scope, economy, and predictive success. success.

This is in itself independent of claims of This is in itself independent of claims of nativism. nativism.

45

The psychology proper begins when one The psychology proper begins when one construes the theories as answers to the construes the theories as answers to the question of what speaker-hearers question of what speaker-hearers knowknow; ; consequently, the questions are raised as consequently, the questions are raised as to how we acquire the information and to how we acquire the information and put it to use. put it to use.

Such a construal places constraints on the Such a construal places constraints on the theories (theories (explanatory adequacyexplanatory adequacy), but ), but these are quite innocent, for there is no these are quite innocent, for there is no a a prioripriori bar on empiricist answers to the bar on empiricist answers to the problems.problems.