language learning

[Brian MacWhinney]

Dear Funknet,

  Thanks to Geoff Nathan for picking up the ball on Rob Freeman's
question about the logical problem of language acquisition. Although
Chomsky floated some ideas in "Syntactic Structures" and "Aspects" about
language being unlearnable because of degenerate input, that idea was
abandoned when Newport, Gleitman, and Gleitman showed that input to children
was, if anything, even more grammatically well-formed than adult
conversation.
  Attention then turned to a particular version of "Plato's Problem" couched
in terms of question formation in English, just as Geoff has noted. In the
most publicized version of this discussion (from Piatelli-Palmarini,
1970), Chomsky discussed the unacceptability of questions like

Is the boy who __ standing in line is tall?

as opposed to

Is the boy who is standing in line tall?

The ungrammaticality of these is said to hinge on the Structural Dependency
condition.

This issue is discussed in great and clarifying detail by Geoff Pullum and
Barbara Scholz in a target article in The Linguistic Review 1o from 2002.
The whole issue is devoted to commentary on the Pullum-Scholz analysis.

I have also been working on this topic, hoping to use the CHILDES database
as a way of testing Chomsky's claims empirically. The results of my
analysis will be coming out in the next issue of The Journal of Child
Language, along with 12 additional commentaries.

It would not be easy to summarize all of these discussions here.  However,
perhaps the most important outcome of my analysis, stimulated in many ways
by the Pullum-Scholz analysis, is that there is good positive data
available to the child for the learning of the Structural Dependency
condition.

In the end, I find myself agreeing with Chomsky on one basic point.  This
is the fact that children have a basic capacity to associate words into
conceptual clusters that then can be treated as "chunked" wholes.  This
operation is enough to allow for the emergence of structure.  There is no
reason to believe that this type of structural dependency is exclusively
linguistic, but I think it is reasonable to imagine that the evolution of
language allowed our species to develop this ability beyond that of our
closest biological relatives.

The particular examples that Geoff Nathan cites involve a somewhat
different constraint on raising.  For this constraint, an account based
not just on
the availability of positive evidence, but also on the conservative nature
of the learning of wh-question patterns seems more appropriate.  Data
providing evidence for this conservatism this comes from:

Kuczaj, S., & Brannick, N. (1979). Children's use of the Wh question modal
auxiliary placement rule. Journal of Experimental Child Psychology, 28,
43-67.

--Brian MacWhinney