I hesitate to continue this thread too long on Corpora List, but since you
asked :-()
Dr Hatch wrote:
> And the paradigm is eminently refutable. In my paper "Chomskian
> linguistics: God's truth or hocus pocus", Ethnographic Studies, 4, 1999
> (available from myself at the above e-address) I point out the
> following:
>
> a. The hypothesis that the 'corpus babies=>toddlers are exposed to
> is insufficient is not in any sense an empirical statement. [In fact
> it is as much dogma as hypothesis.]
I believe the argument here from Chomsky and his pupils is in part the fact
that we don't get negative evidence for certain things, yet we wind up
knowing them anyway.
The lexical gaps in the English dative alternation are a common (and
debated) example. So for instance, I can say "She gave the book to the
men" or "She gave the men the book", and I can say "She distributed the
book to the men" but not "*She distributed the men the book." One can
argue over the * on the last sentence, and I'm sure you can find examples
like it in corpora, but the point remains that (for most of us native
speakers) "distributed" is definitely marginal in this context compared
with "gave".
So the argument would be, why do children fail to make the obvious
generalization:
NP1 to NP2 NP2 NP1
give OK OK
distribute OK *
Of course, there are a lot of answers one might suggest, not all of them
relying on a Chomskian analysis. I'm not saying this is the best argument,
I'm just saying that it is the _sort_ of argument that Chomsky used. Not
so much "children don't hear enough data", but "why don't children make the
obvious generalizations?"
That said, you have to remember that even children who are exposed to very
little language (maybe their parents are abusive, or they live in an
isolated place) still pick it up. Maybe less vocabulary, but the grammar
is there (apart from the very very extreme cases, like Genie).
(BTW, Dave Graff--hi, Dave!--suggested that there's a lot of variability in
what children learn, even within a dialect. I'm sure, but I think most of
that is (1) lexical, as in this example, or (2) willingness to accept
marginal cases. But I'll try to stay on track. Sorry, Dave!)
> b. A child's 'mastery' of its native language including parts of
> the grammar is often deficient until it is 5 or 6.
This was documented by Chomsky's wife, Carol Chomsky, back in the 1950s (or
early 1960s). In fact she went further, arguing that some parts of grammar
don't stabalize until age 8 (IIRC). I don't think that this argument is
damaging to Chomsky's point (nor does Noam), any more that the supposed
refutation I heard some years ago when it was found (or at least argued)
that children could hear their mother speaking in utero. What's a couple
more months...
> c. The hypothesised UG, or whatever, can be replaced by an
> hypothesised facility (found in at least all mammals) that may be called
> the active pattern recognition facility or somesuch. This would
> enable us to account for more than just linguistic ability.
Chomsky's argument, back in Syntactic Structures, was that grammar is not
just patterns, but s.t. considerably more complex in a well-defined
mathematical sense. (An early example was the fact that in yes-no question
formation in English, it is not the first auxiliary verb that fronts, but
the root clause's aux V--and this requires a parsing of structure, not just
patterns.)
Of course, Chomsky might be wrong about the mathematical form of grammar...
Also, there are the well known gaps in the patterns, like the dative
movement case above, or even more interesting, the "COMP-trace filter"
sorts of things.
And of course Chomsky's response by the claim that our language learning is
done with the same faculty found in other mammals is, why can't those
mammals learn language under the same circumstances that humans do (or even
with special training, which humans don't require)?
> All the evidence suggests that languages have to be learned. It's just
> that none of us learn like dim Pavlovian dogs.
No one--Chomsky included--would disagree, I think. The question is how we
do learn, and how much of what we know is built in (and how much that
differs from what other mammals have).
> But, as I flunked my phonology course as a fresher, perhaps someone
> explain what the hell all this has to do with that subdiscipline in
> NORMAL language.
I think the point is that all of us know a normal language, and no
computers or other mammals do.
-- Mike Maxwell maxwell@ldc.upenn.edu
This archive was generated by hypermail 2b29 : Wed Oct 19 2005 - 04:14:54 MET DST