Re: Human language acquisition

in article [EMAIL PROTECTED], Fred Mailhot at
[EMAIL PROTECTED] wrote on 12/2/03 9:32 AM:

> On 11/29/03 10:00 PM, "Glen M. Sizemore" <[EMAIL PROTECTED]> wrote:
> 
>> 
>> GS: Well, I was careless. In modern times, we do, literally, sometimes apply
>> rules when we speak our native language. But a child who says "goed" or
>> "hitted" is not following a rule literally, and people spoke grammatically
>> for thousands of years before anyone knew that the notion of rules could be
>> metaphorically applied. Clear it up for you Mike? I didn't think so.
> 
> Hi Glen...
> 
> Obviously, I have no grasp on the behaviourist (I never know whether or not
> to capitalize that) literature concerning language acquisition (I've yet to
> read Verbal Behaviour - or Chomsky's review, for that matter -) although I
> know that Catania discusses it in his text...
> 
> Anyway...what's the behaviourist view on *why* (English-speaking) kids
> overregularize and say things like "goed" and "breaked", when those forms
> are *never* in their input experience ??  In particular, why do (English-
> speaking) kids *always* do this ??
> 
> Incidentally, most generative linguists (I can hear your eyes
> rolling...*grin*) regard Marcus (1993)[1] as the final nail in the coffin of
> the "negative evidence" issue. I'm not quite so sure about it, but there are
> some interesting points raised...
> 
> 
> Marcus, Gary (1993). "Negative Evidence in Language Acquisition." in
> Cognition, 46 (1993), 53-85.
> 
> 
> Cheers...
> 
> Fred.
> 
 Hi, Fred

See Rohde and Plaut "Language Acquisition in the Absence of Explicit
Negative Evidence: How Important is Starting Small?" for a different point
of view (basicaly that Gold's theorem on learnability does not apply under
the assumption that language presentation may be modeled as a stochastic
process - "implicit" negative evidence in the form of the distributional
properties of the input allows prediction failures to drive learning without
the need for explicit negative evidence).  See also, however, Nowak,
Komarova, and Niyogi's "Computational and evolutionary aspects of language"
review paper for a treatment that includes the deep results of statistical
learning theory due to Vapnik and Chervonenkis - a VC-dimension aware
analysis: "In summary, all extensions of learning theory underline the
necessity of specific restrictions."

-- Michael