Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.sprintlink.net!uunet!in1.uu.net!psinntp!scylla!daryl
From: daryl@oracorp.com (Daryl McCullough)
Subject: Re: What's innate?
Message-ID: <1995Feb22.033435.26696@oracorp.com>
Organization: Odyssey Research Associates, Inc.
Date: Wed, 22 Feb 1995 03:34:35 GMT
Lines: 160

For what it's worth, I would like to apologize to Neil Rickert for my
hot-headed comments in past postings in this thread.

jqb@netcom.com (Jim Balter) writes:

>>Neil, I don't know where you are coming from with these analogies,
>>since they seem to me to work against your argument. *If* children
>>learned juggling or integral calculus the same way they learned
>>language, then I would agree that the same kind of learning was
>>involved. But children don't just "pick up" integral calculus from
>>being exposed to it, the way that they do language.

>I have been exposed to very little juggling in my life, and not a whole lot
>of calculus, certainly before the age of 18.  I know of children, both of 
>whose parents are physicists, who *do* in fact develop very early familiarlity
>with mathematics.  I know of children who develop incredible abilities to
>juggle, skateboard, kick ball, throw frisbees, etc. etc. from no more than
>"exposure". In fact, I suspect that a young child could learn to juggle
>with no exposure at all beyond merely knowing that activity can be done.
>What are your grounds for your claim?

There are several different points to be made here:

1. In the case of language skills, we are not talking about what *some*
children accomplish, we are talking about what *all* children who are not
severely handicapped accomplish.

2. In the case of physical skills such as driving a car, juggling,
etc. there is much more than simply exposure. There is immediate
feedback.  If you do something wrong in juggling, you drop the
ball. If you turn the steering wheel the wrong way, you go the wrong
way. There is absolutely no "poverty of stimulus" argument for (a
large number of) physical skills, since you immediately learn that you
have made a mistake.

3. In the case of complicated mental skills like calculus, people don't
have to simply "pick it up". They can be *taught* it using language. (Or
they can read about it on their own, again using language.) The big
advantage to having language is that it makes learning almost everything
else so much quicker and easier.

Having said that, I realize that little kids actually have language to
help them learn language. Once they learn a core language, that core
language can be used to "bootstrap" them to the full language. For
example, they can be *told* "No, not `Me want bannana', `*I* want a
bannana.'"  However, I don't know whether this kind of correction is
necessary for children. Very young children of immigrants seem to be
able to pick up English just by watching TV, where there is no such
correction.

Anyway, whether or not there is a UG, it sure seems to me that language
is very much a special case. In many ways, it is much unlike juggling or
anything else that kids learn.

>>However, there is the case of "prodigies", kids who learn piano
>>or math or whatever without being taught, but simply by being exposed
>>to it. I don't know whether there is any good theories about what
>>makes prodigies, though.
>
>Surely you don't think that the facts that Mozart was continuously exposed to
>harpsichord music in the home and that his father wrote manuals on the
>teaching of music are irrelevant, do you?  "Simply being exposed" fails to
>note that such prodigies had unusual exposure.  Now, not everyone equally
>exposed develops equal abilities.  There are matters of temperament and other
>contingencies.  And of course people do have different aptitudes, but that
>does not seem to mean that some have an appropriate "module" and some do not.

I wasn't using the case of prodigies to argue either for or against the thesis
of UG. I was saying that their existence is an interesting piece of data, if
I knew what it showed.


>What we can say about language is that nearly everyone seems to have 
>sufficient aptitude to a pretty good job.  But there is the major contingency 
>that language aids in communication, and we do seem to have a builtin desire 
>and need to communicate, which the child does from the moment of the
>first breath.  To get from loosely analyzed anecdotes about exposure
>to "a language module" seems quite a stretch.  I'm not saying that
>that's all that Chomsky does, but it seems to be all that *you* are
>doing.

What anecdotes did I give, and where did I argue that they showed the
existence of a language module??? I really don't know what you are
referring to. I was simply objecting to Neil's lumping language
learning in with driving and juggling. I think that language is a
special case, whether or not it is innate.

>You complain later about weak arguments; how strong is yours?

Weak as water. Chomsky's arguments are weak. Neil's arguments against
them are weak. My arguments against Neil are weak. Let's face it,
we're all a bunch of weaklings in this newsgroup. 8^)

>And what is your argument that Neil doesn't understand what Chomsky means
>bu POS?  Even if you can show that he doesn't, it is quite a leap, and
>a bit of an ad hominem one, to claim that he hasn't *bothered* to.

Well, I apologize to Neil for that and other hot-headed remarks. What
bothered me about Neil's posts on the poverty of stimulus was that he
never made explicit what he understood POS to mean. His numerous
ridiculing articles never made clear exactly what position he was
ridiculing. Now, maybe that is because Chomsky was very vague about
what *he* meant. But in that case, it seems to me that Neil should be
attacking the vagueness, instead of saying that the poverty of
stimulus argument works as well for driving or juggling.

>Um, the main argument presented against it [UG] in this group has
>been that the arguments presented for it in this newgroup are pretty
>weak.  What we have is agnostics arguing that a claim isn't
>well-supported.  Agnostics don't need a particularly strong argument
>for that.

I agree with *that* argument. But Neil has gone beyond simply saying
that he finds the evidence unconvincing. He has gone to great lengths
to ridicule the idea of a UG. (With his posts about a "driving"
module, etc.) I'm pretty tired of that kind of argument by ridicule.
Searle uses against AI, Neil uses it against UG. I don't think it
accomplishes anything, other than making communication between people
who have differences of opinion much more strained and difficult.

Of course, maybe Neil thinks that the UG hypothesis is so hopeless
that there is nothing to be learned from those who support it.

>>>>          It seems to me that stochastic arguments actually support
>>>>Chomsky, because (1) sentences with the right word frequencies still
>>>>sound like nonsense, and (2) sentences that are clearly sensible can
>>>>use quite rare word combinations.
>>>
>>>It seems to me that this misses the point.  When the words are chosen
>>>on the basis of the meaning that is to be conveyed, the number of
>>>possible words drops to a very small number.  The "quite rare"
>>>word combinations may be highly probable given that semantic
>>>restriction.
>>
>>If the words are chosen according to meaning, then why is a knowledge
>>of statistical likely necessary? I think that this statistical approach
>>is a pretty weak alternative to using grammar.
>
>I agree. Unlike Neil, I think that it is reasonable to analyze natural speech
>in terms of grammar, although I think it is a grammar heavily constrained by
>linear production, not the sort of recursive phrase-structured grammar that
>would lead to the silly claims of Pinker and Curry that deeply nested phrases
>are "grammatical" despite the fact that no human produces them and no human
>recognizes them (apparently these folks are unfamiliar with ad hoc grammars
>like FORTRAN_identifier = letter letdig letdig letdig letdig letdig).
>
>But of course that does not imply a UG.

I don't think the claim that deeply nested phrases are grammatical is
silly. I agree that there are processing limits to what we can produce
and understand, but I think that grammar is *not* the processing
limitations.  There is no reason to have a "rule" saying "Don't do X"
when X is something you couldn't do, anyway.  Grammar is structure and
redundancy in language that makes understanding language in the
presence of noise possible. It's a code, like Shannon's
error-correcting codes, but more closely tied to semantics.

Daryl McCullough
ORA Corp.
Ithaca, NY
