Structure-dependency & the Wall St. Journal

Daniel L. Everett dever at VERB.LINGUIST.PITT.EDU
Sat Apr 19 20:26:04 UTC 1997


One thing Brian said in his posting I agree with:
structure-dependency/rules-kinds of issues are not going to be resolved on
a b-board. There is a rich literature on this.

On the other hand, there are some simple questions to ask. If Chomsky
claimed that children are not exposed to data on structure dependency, he
would be wrong, as Pullum might have shown in his BLS article (I doubt
that Chomsky can be accurately said to have claimed this). But let's say
that such a gross factual error was made. It turns out, then, that
children *are* exposed to sufficient data to draw the correct
generalizations about structure-dependency. But what does this mean?

What is it that children learn about structure-dependency? They learn
embedding, based on notions of main clauses and subordinate clauses (or
any label for those things you care to give). How do they do this? By
trial and error? If that were the case, then Chomsky really would be
wrong. Is there evidence that children err in acquiring their language on
the side of linearity, rather than configuration? Chomsky's point was and
is that configuration is a much more abstract and difficult concept than
linear order. Therefore, it is curious, so Chomsky's argument runs, that
children do not err in placing subordinate verbs first in yes/no
questions, rather than main verbs, since subordinate verbs would be
favored by linearity.

Of course, there is an potential response (which readers of this list
will be aware of), namely, that children use semantics, not syntax in
forming questions, so that the linearity/configurationality issue Chomsky
raises might be a red herring. But this is a hard issue in itself - to
show, for example, how the semantics and syntax match up.

The crucial issue in Chomsky's story is that a behavioristic approach is
wrong. That is, there seems to be no convincing evidence that children
learn syntax, morphology, or phonology by trial and error (they *do* learn
discourses this way, though - one of the significant differences between
discourse-level generalizations and sentence-level generalizations). There
is something they know when they start learning grammar. And whatever this
is, it keeps them from making some fairly easy to imagine errors. So it
seems pretty specific. (But maybe it's not. It is something more specific
than what a rock, tree, or bat knows, but less specific than what an adult
knows. There is a lot of room there. Some of Hume's footnotes make the
same points, with very unChomsky-like conclusions.)

In any case, Pullum's reading of the Wall St. Journal has nothing to say
about any of this, so far as can be told from Brian's report of it.

-- DLE

More information about the Funknet mailing list