where does grammar come from

Elizabeth Bates bates at CRL.UCSD.EDU
Fri Jan 29 18:06:33 UTC 1999

Subject: Re: Request for good syntactic examples

Yes do keep looking for semantically opaque syntactic phenomena, that is
an important issue.  But I think it is important to keep in mind that
semantic content is only PART of a functionalist argument for why
grammars look the way they do.  As Brian MacWhinney and I have noted
for a few years, a functionalist account is going to need to take into
considerably both the semantic content AND the large set of information-
processing constraints that serve as the "Darwinian space" in which
grammars have evolved.  A metaphor: line orientation detectors,
center-surround cells and ocular dominance columns do not "look like"
the content of vision, and yet they arise over and over again.  two
possible explanations:

(1) These features of visual cortex are entirely
arbitrary, mutations under strict genetic control,
innately insured.

(2) These features are necessary to get the job of vision
done, and will arise even if they are not innate.

There are now multiple neural network simulations showing that
many of these features DO arise, inevitably, in a neural network
with no innate "knowledge", as a function of trying to solve
(a) the dimension-reduction problems involved in mapping from a
3-dimensional world through a 2-dimensional retina, (b) the
special problems posed by competing inputs from two eyes with
overlapping visual fields, (c) waves of plasticity that pass over
cortex due perhaps to maturational gradients (think of plants being
periodically sprayed by nutri-grow....).  In other words, opaque-looking
features can be the product of processing and development, so inevitable
that they do not have to be innate above and beyond the innate factors
that get the animal to try to solve that problem in the first place
(e.g. eyes that happen to be placed on the head so that the visual
fields overlap -- frogs don't have that problem, and don't have
ocular dominance columns as a result, but they GET ocular dominance
columns if you implant a third eye right in between the usual two
in an embryonic frog.....).

So, it is useful to collect semantically opaque grammatical phenomena,
but these cannot constitute, ipso facto, evidence for innate knowledge.
The pathway may be very indirect, heavily determined by processing
facts and by the severe dimension-reduction and constraint-satisfaction
problem involved in mapping from a high dimensional meaning space (with
a lot of competing material) onto a low-dimensional channel.  -liz bates

More information about the Funknet mailing list