easily imagined errors

[Brian MacWhinney]

Dan,
  I agree.  It is true that the fact that Chomsky was wrong about the facts
concerning the distribution of data to derive the structure-dependency
generalization does not mean that the rest of Chomsky's argument is wrong.
It is true that, as you and Chomsky say, there is something that "keeps
children from making some fairly easy to imagine errors."  But these "easy
to imagine errors" are not actually ones that ever occurred to the child.
The child never tries to derive questions from the corresponding
declaratives (as several previous email messages have noted).  Because of
this, the linear movement or transformation generalization was not one that
the child was considering in the first place.
  I agree that the question is how the child accesses semantic structure in
a disciplined enough way to avoid egregious errors.  To explore this, we
don't need the hard examples.  We can just look at a sentence like "Is
Daddy coming?"  There is a pretty rich child language literature on the
development of questions.  For this type of question, there appears to be a
stage when the aux is missing and we have just "Daddy coming?"  The
intonation is there, as is the verb and the subject.  Only later, it
appears, does the child add the aux.  I think this path makes sense.  The
most uniform, reliable marker of the question across types in English is
the intonation.  That gets mapped first, along with the core proposition.
Then the embroidery gets added later.  The aux wasn't moved, it was just
added.  When we get to the harder examples, the story is the same, since
the complex-NP subject is a cognitive unit the child doesn't look to it for
the required aux.

--Brian MacWhinney