Re: Pinker-Bloom Natural Language Selection

From: Stevan Harnad (harnad@coglit.ecs.soton.ac.uk)
Date: Thu Dec 09 1999 - 14:48:12 GMT


Toth Eva <tothe@edpsy.u-szeged.hu>
& Iván Zsuzsanna <manka1@freemail.hu> wrote:

t&i> *Comment1: Beeing biologists we agree with the authors, that evolution,
t&i> and natural selection can be the only explanation for the development
t&i> of the language. We think, there is no point in disputing this.

The trouble with this ready acceptance is that it misses the reason that
it had to be explicitly stated in the first place, by Pinker & Bloom
that language must have evolved by natural selection:

If we were discussing the origin of upright (bipedal) walking, would a
target article on "Natural Walking and Natural Selection" be necessary?
If not, why not? And if so, what is it about natural language that makes
the question of its evolution more problematic than the evolution of
walking?

The answer is: Universal Grammar (UG). UG is a very complex set of
internal constraints -- even in its newest minimalist formulation. It, or
rather the neural structure and function in which it is implemented,
constitute a kind of language "organ," and it is the evolution of that
organ that we are inquiring about.

So far, that organ might just as well have been the locomotor system, as
in walking, and there would be no problem, for we could easily imagine
and even test an evolutionary scenario in which the adaptive advantages
of this organ were "shaped" by the usual blind-watchmaker processes of
natural selection. It is easy to tell an adaptive story about both the
advantages of bipedalism, and how it could have evolved out of
quadrupedalism.

Is the problem that UG is cognitive whereas locomotion is motor? No, for
one could tell the same sort of adaptive story, without any special
problems, for other cognitive capacities, such as pattern recognition,
learning, memory, and even reasoning and problem-solving. In fact, there
are aspects of natural language that are not problematic in this way
either: phonology, vocabulary, and primitive syntax. All of these could
have evolved in the usual way without raising any special problems.

But UG is different for two fundamental reasons. The first is that it is
NOT adaptive! It is not true that UG is a better or simpler or more
probable grammar than others that would have done the job just as well.

I believe this is a fact: The methodology of UG theory has always been
to invent other ways that one might have said things, especially ways
that look like simple generalizations from the way we actually say them,
yet the grammatical intuitions of all of us say that it is not possible
to say things that way, in ANY language. UG is then gradually discovered
through theoretical construction of the rules that are needed to INCLUDE
the ways that are grammatically possible and EXCLUDE the ways that are
logically possible but not natural-grammatically possible.

But notice that there is nothing here about UG being in any way optimal,
"saticeficeable", or anything like that. It is simply THERE. It is not
better or worse than the logical alternatives that would accomplish the
same thing (namely, to be able to express any possible proposition).

That is problem number 1.

Now the solution to problem number 1 (the non-adaptiveness of UG) could
have been that UG did not evolve, it was simply invented by us, and
passed on by learning and instruction. There are many such complicated
systems of rules, including, for example, mathematics, for which no
natural-selection story is needed.

But now we come to problem number 2: UG is not learnable! Yes, we learn
English and Hungarian, and the non-UG grammatical aspects of languages
(such as that is not correct to say "I ain't be coming"). But those are
not at issue. Insofar as UG is concerned, all we ever learn is whether
to set the dials on this big machine (or organ) to the Hungarian channel
or the English channel. The rest of the machinery we already have: we
never learn it.

How do we know we never learn it? Because none of those logically
possible sentences that the UG linguists have been inventing for
decades, and discovering (by consulting their intuitions) that they were
ungrammatical -- none of them is ever heard, spoken, or corrected by
anyone else, at any age, but especially not from age 1-6, when the child
becomes a full (grammatical) speaker of the language.

So that means that we do not LEARN UG, we already KNOW it.

And that is very bad news for an evolutionary theory of the origin of UG.
For exactly the same thing that makes UG unlearnable (it is called "the
poverty of the stimulus") also makes it unevolvable (problem 2),
especially since even if the variation had been there for selection to
operate upon, there is no adaptive advantage to UG over other logical
possibilities, including much simpler ones (problem 1).

Do you think Pinker & Bloom1s argument about the need for "parity"
conventions (which is equivalent to agreeing that it will be "1" that
means "yes" and "0" that means "no") are sufficient to cover the case of
an awkward monster like UG (a considerably more complicated convention
than parity!)?

t&i> *Comment2: It is not obvious, that the larger brain size is connected
t&i> to better communication skills. The H. S. neanderthalensis had bigger
t&i> brain than the modern Homo Sapiens, and nobody think that they were
t&i> better in communication than us.

You may be right, but we are talking about cognitive skills, not just
communicative ones, and who knows how these two strains were in
cognition (I'm not sure how they were in communication either, relative
to one another...)?

t&i> *Comment3: A question connected to the process of language evolution:
t&i> If there is only a man with spontaneous mutation, who is using a higher
t&i> grammatical step, it is not obviously an advantage, because the other
t&i> 100 understands him more difficult, although he can express himself
t&i> better, and more clear.

Correct. And in general, a mutation story for the "shaping" of UG is
much more difficult even to imagine, than one for shaping walking. And
that is the problem!

t&i> *Comment4: We don't agree, and don't understand, why will have a mother
t&i> with "m" language rule a baby with "m+1" language rule?

I agree; but note that after the general ideological argument on behalf
of the evolution of language like anything else -- arguments in which
the easy and the hard were inextricably wrapped together, simple,
syntactic conventions and vocabulary being trivial cases, and
UG being very difficult, if not impossible to treat in the same way --
Pinker & Bloom were also quite lax about whether they were talking about
genetically inherited rules or learned ones.

--------------------------------------------------------------------
Stevan Harnad harnad@cogsci.soton.ac.uk
Professor of Cognitive Science harnad@princeton.edu
Department of Electronics and phone: +44 23-80 592-582
Computer Science fax: +44 23-80 592-865
University of Southampton http://www.cogsci.soton.ac.uk/~harnad/
Highfield, Southampton http://www.princeton.edu/~harnad/
SO17 1BJ UNITED KINGDOM



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:06 GMT