Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!swrinde!ihnp4.ucsd.edu!news.cerf.net!pagesat.net!internet.spss.com!markrose
From: markrose@spss.com (Mark Rosenfelder)
Subject: Re: What's innate? (Was Re: Artificial Neural Networks and Cognition
Message-ID: <D3nIwA.5CA@spss.com>
Sender: news@spss.com
Organization: SPSS Inc
References: <D38qGn.H6L@gpu.utcc.utoronto.ca> <D3DsyK.1tt@cogsci.ed.ac.uk> <D3Fpv3.8o2@spss.com> <D3LGHJ.LpM@gpu.utcc.utoronto.ca>
Date: Tue, 7 Feb 1995 22:38:33 GMT
Lines: 102

In article <D3LGHJ.LpM@gpu.utcc.utoronto.ca>,
Andrzej Pindor <pindor@gpu.utcc.utoronto.ca> wrote:
>In article <D3Fpv3.8o2@spss.com>, Mark Rosenfelder <markrose@spss.com> wrote:
>>There's some stages of language acquisition where this might be true-- 
>>infants learning names of things, for instance-- but I don't see that it's
>>even approximately true of the process as a whole.  Children start speaking
>>by babbling, for instance, which is hardly copying anything they hear.
>>A two-year-old's sentences are almost entirely novel, not copied from
>>any adult utterance, and indeed violating many rules of adult grammar.
>
>I can only speak from my own experience with two-year-old's (and it has been
>quite a few years since) but their sentences are mostly ad-hoc combinations
>of the elements of the language they have heard. They are able to parse
>the speach they hear into words and they combine them according to patterns 
>they have discerned. What is your definition of "almost entirely novel"?
>Children are copying the words they have heard and phrases they have heard,
>what is the novelty you are talking about?

Children, like adults, can create words, phrases and sentences they have never 
heard before.  Certainly the elements they build them out must be learned
from others, but to call the whole process "copying" because of that is
like saying that writing a computer program is "copying" because the names
of the commands are copied from a manual, or that composing a piece of music
is "copying" because all the notes already exist on the piano.

Let's look at some of the things English speaking children generally learn
by about their third birthday:

* They can express a wide variety of concepts, by no means limited to 
ones they have heard before.  The major purpose of language, after all, is
to communicate; and one of the things a theory of language must explain,
it seems to me, is how a speaker can use it to frame and express utterances.
Putting the emphasis on "copying the words... and phrases they have heard"
seems to me to miss the point.

* They can handle such syntactic features as:

-- morphological changes (plurals, verbal inflections) 
-- morphological irregularities
-- proper use of pronouns (including proper use of 'I' and 'you', and
   use of object pronouns, which implies an understanding of case roles)
-- subject-verb inversion in questions, which in turn presupposes knowing
   that sentences are composed of *groups* of words that go together 
   (syntactic constituents), not just strings of isolated words
-- correct placement of modifiers within a noun phrase (before or after
   the noun-- compare "strong mens", "something funny")
-- proper use of pronouns in place of repetition of noun phrases
-- insertion of contentless pronouns (e.g. it) where English requires them 
-- demonstratives, articles, prepositions, particles, and other form words
-- past, present, and future tense
-- aspect (unmarked vs. progressive)
-- negatives, including the insertion of the auxiliary "do", and the 
   negation of just part of the sentence (different in English than
   negating the main verb)
-- subordinated clauses (including the recognition that the main clause,
   not the subordinate clause, must undergo S-V inversion in questions)
-- auxiliaries and other multi-verb sentences (including which ones require
   'to', and again including proper usage in questions)
-- imperatives (including dropping of subject)
-- quantifiers (some, many, another, etc.)

(This is not simply a summary of English grammar; important omissions,
presumably learned later, include passives, reflexives, and the perfect tense.)

* They've grasped the conceptual and semantic features underlying all that, 
including developing a system of categorization (cf. _Women, Fire, and 
Dangerous Things_), an understanding of time, causation, and possession, 
understanding of reasons and goals, acquaintance with various speech acts and 
their effect, and of course an enormous amount of real-world knowledge.

Considering all this, most linguists don't find an explanation in terms of
copying what is heard very compelling.  They see the child building a
complex conceptual system, linked to sophisticated grammatical and pragmatic
knowledge, and, importantly, a procedure for perceiving and integrating new 
grammatical insights; copying things heard is only one part of the process. 

(If they're Chomsky, of course, they think that much of this is a) inbuilt
and b) language-specific.  But this is by no means accepted by all linguists.)

>>And children are often highly resistant to grammatical correction, which
>>they shouldn't be if they're just copying what they hear.
>>
>They are resistant to grammatical corrections precisely _because_ they are
>copying what they hear, with most obvious patterns imprinitng on them most
>strongly. They listen to the language all the time, most of which you are not
>aware that they are listening. Occasional corrections rarely have enough 
>weight to influence the strongest patterns. As they grow up, they start to
>notice more and more subtle patterns and then corrections have more chance to
>have influence.

In broad outline this is true; but the acquisition of new patterns (or rules,
in linguist-talk) is more complex than that.  What linguists have noticed,
and seek to explain, is why *certain* patterns are generalized, and not
others.  Certain classes of errors are very common, and others are made
hardly at all.  (See Pinker, p. 272, for some examples.)

It's not enough to say that a given error is not spoken because it's not
heard.  Children say plenty of things that are in error, and that they have
not heard.  This is one reason why Chomsky hypothesizes UG; he posits, in
effect, that children already know not to make certain errors.  I myself
am willing to hear other explanations (but I want to see the problem
explained, not dismissed or ignored).
