reply to Brian MacW.

[Tom Givon]

Good comments, Brian. But you ascribe to me more than I said. It is not
that I think evolution involves no local-level changes. It involves all
levels of changes. And clearly complex structural organization evolves from
more simple precoursors (see general discussion in J.T. Booner's "The
Evolution of Complexity"). What I objected to only is the total obsession
of (at least the early) connectionist work with the low-level stuff. And
there are clearly exceptions to that. E.G., Jeff Hinton's rather inspired work
on modeling the interaction between individual learning behavior and
genetic evolution (reviving Lamarck but with a much more plausible mechanism
based on -- guess what -- the rise of automaticity) has been a great inspi-
ration to me in my own work. But -- even when adaptive steps are small and
subtle, when they involve highly-organized functional modules, they cannot
be explained merely by the lower-level steps (but rather by their interaction
WITHIN higher-level modules). So for example, you cannot explain in toto
the adaptation of the lower tip of the primary motor cortex (Broca's area)
into a grammar-related module by purely local low-level chganges. Nor can you
explain the extension of the hippocampus-based system of episodic memory
from visual to verbal this way. Nor the adaptation of portions of the
primary auditory center into a phonemic analyzer. Nor the adaptation of
the object-recognition module(s) and visual semantic memory into a
general verbal/visual semantic memory. Etc. etc. etc. In each case, local
and low-level changes do not happen in a vacuum, they happen within the
context of extant complex functional modules and the extention of extant
complex behavior to related but not identical new functions. If connec-
tionists would treat these realistic evolutionary problems seriously, they
will, I suspect, cease to become "connectionists" in the classical (and
to my mind rather restrictive) sense. So we should then welcome them back
into the large community in the muddled middle where science is practiced
(or at least where I think it can be practiced best).
Best, TG