science article

[Gary Marcus]

<Please accept my apologies if you receive multiple copies of this message.>

Hi, Brian,

Thanks for spreading the word about new results, and also for the
oportunity to clarify what we think our results show. (Both the article and
a Perspective written by Steve Pinker appear in the January 1st issue of
Science;  both are available electronically through
http:/www.sciencemag.org.)

I would certainly agree with you that there is huge difference between
tracking patterns of syllables and understanding sentences; we surely would
not want to claim that seven-month-old can understand sentences or that
they have "acquired syntax." (I tried to be clear about this both with
reporters and in the article itself.)

Our view is that what we have identified is a device for extracting rules,
and that this device is just one among many that may play a role in the
acquisition of language. Our results do not at all obviate the need for
other tools, such as those that detect statistical relationships, or those
that help us to understand the communicative intentions of others.
Different kinds of learning devices excel at different kinds of learning
tasks; our hope is that work developing an inventory of tools that infants
have _available_ can help constrain our hypotheses about what tools are
actually used in acquiring language.

The rule-extraction device that we have identified may or may not be
specific to language. We are planning to test whether babies and non-human
primates can recognize similar sorts of patterns in sets of tones or visual
stimuli.  My hunch is that we will find that the rule-extraction device is
not specific to language. Rather, I see the rule-extraction device as one
kind of  mental "circuit", of which we may have many copies, some of which
might participate in domain-specific devices, other of which might
participate in domain-general devices.

Another point that was obscured in one news report was what we think of the
idea of testing artificial "sentences" that include words that contains
sound contrasts outside the native language of a child. While we find such
a study interesting, we also told the reporter that we do not believe that
the question of rules vs. associations rests on it:  An infant who learned
a rule that was couched in terms of relations between words might not
interpret items containing non-native contrasts as words, and hence would
not take those items to be subject to that rule.

Thanks again for your interest.

Best wishes and Happy New Year,
Gary

------------------------------------------
Gary Marcus
Department of Psychology
New York University
6 Washington Place, Room 306
New York NY 10003-6634
e-mail: gary.marcus at nyu.edu
http://www.psych.nyu.edu/~gary/home.html