Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!hao!boulder!sunybcs!rutgers!iuvax!merrill
From: merrill@iuvax.cs.indiana.edu
Newsgroups: comp.ai
Subject: Re: Language Learning
Message-ID: <12400009@iuvax>
Date: Wed, 11-Nov-87 18:06:00 EST
Article-I.D.: iuvax.12400009
Posted: Wed Nov 11 18:06:00 1987
Date-Received: Sat, 14-Nov-87 04:32:29 EST
References: <1966@uwmacc.UUCP>
Organization: Indiana University CSCI, Bloomington
Lines: 61
Nf-ID: #R:uwmacc.UUCP:-196600:iuvax:12400009:000:3223
Nf-From: iuvax.cs.indiana.edu!merrill    Nov 11 18:06:00 1987


M. B. Brilliant writes:

>  In article <1399@tut.cis.ohio-state.edu>, zwicky@ptero.cis.ohio-state.edu (Elizabeth D. Zwicky) writes:
>  
>  > ... Children
>  > do learn second languages; the interesting question is whether or not
>  	...
>  
>  What was once believed to be quick language learning in children is now
>  believed to be only quick pronunciation learning.  Children who seem to
>  know a language well, because they speak it fluently without an accent
>  in social situations, are not necessarily ready to use it to learn
>  things they don't already know.
>  
>  Even pronunciation learning may have nothing to do with crystallization.
>  The onset of identity crisis might make an adolescent less eager than a
>  child to adopt a new language.  And an adolescent might be more likely
>  to recognize that superficial fluency is not enough, and adopt a
>  halting pronunciation to signal that the language has not been mastered.

Although the syntactic evidence is evidently less solid than I had
realized, the phonetic data about production and recognition of second
languages is fairly clear.  A child can gain mastery of both the
production and perception of difficult phonemes, but an adult can not
(it seems) become proficient in recognition even if he is trained to
produce a distinction correctly.

The best data along these lines come from Japanese adults trying to
learn English as a second language.  Japanese, like related Asian
languages, does not contain the [r]/[l] pair; thus, speakers of
Japanese do not learn to discriminate between these two phonemes very
well.  Even if adults are taught artificially to make the distinction in
speech, no matter how patiently---thus getting around the "see doggie
run" kind of argument---they *do not* acquire any statistically
significant skill in recognizing these two phones.  It seems to me
that this fact indicates that there is a real crystalization effect.

I seem to remember another experiment, which I will relate with the
proviso that I may have to retract my claims about it.  It seems to me
that I remember an experiment in which adult speakers of Japanese were
trained to discriminate same/different along the [r]/[l] continuum for
short intervals.  The experimenters hoped to find that the subjects
would then be able to generalize these discriminations to learn [r]
from [l].  My memory is that the subjects could learn to make the
narrow discrimination of side-by-side tokens, but could not
generalize.  If so, this would also indicate that there was a real
crystallization effect, and might lead one to hypothesize that it lay
in the ability to generalize.

Incidentally, language is not the only area in which a crystallization
effect has been hypothesized.  I've heard of claims for such a
phenomenon in musical performance (specifically cello,) and other
motor behavior.  In the cognitive realm, the general similarity of
many of the stages of concept acquisition might also reflect stages of
a crystallization hierarchy, although I don't know enough
developmental cognitive psych. to be able to argue that one either
way.  Does anybody out there know more of the details of the
situations in either of these fields?

--- John Merrill