Re: CHAT: Parallelism

FWIW, I don't think I ever said "phonemes don't exist" or even "phones
don't exist."  I think it was you who said that in reply to me (or was
it Tom Weir?).  Phonemes, and phones, are elements of theories, and
one can't say that elements of theories don't exist; they exist in
that theory by definition.

I just thought that it was interesting that some people were
questioning the assumption that language works like:

    sound <---> phone string <---> phoneme string <---> morph

And perhaps substituting something like:

    soundshape* <---> morph (or higher level? or multi-level?)

where a soundshape is an articulartory gesture or the auditory
reflection thereof, built out of phones but not necessarily reducible
to them, as properties emergent from and unique to the particular
combination of phones may be vital to its makeup.

It seems to me as if this is one instance of the recognition that
contrary to the assumptions of many linguistic theories, the brain
doesn't work like a computer, and one very important dissimilarity is
that in many computers, computation is 'cheap' and storage is
'expensive', whereas in the brain, computation is 'expensive' and
storage is virtually free.

That's why suppletion and inflection (as opposed to agglutination)
and irregularity are so rampant in human languages: it's literally
much more efficient in language *use* to store a set of single
morphemes that encode "first person, singular, past tense, verb 'to
be'" and every variation thereof than to re-analyze an agglutination
of such values each time it is used.

Suppletion makes language *learning* more difficult, but anything
that involves access to stored, pre-made patterns rather than repeated
computation is preferable as far as the brain is concerned.

Now, this is *exactly the opposite* of the assumptions that guided
the entire Transformational Grammar project and most of its subsequent
incarnations.  (Indeed, what's the title of the latest theory from
Chomsky?  "The Minimalist Program")

It's also the opposite of what any theorist who likes neat, clean,
simple, spare, powerful formal theories of grammar and phonology would
prefer.  But it's how we really work.

So it seems to me eminently *likely* that we store unified gestalts
of sound-shape for each word (or phrase!), and modify them as little
as possible to say what we want to say, despite the fact that so much
intellectual energy has been spent separating everything into lots of
little levels and building everything out of interchangeable discrete
units.


Ed Heil ------ edheil@postmark.net
--- http://purl.org/net/edheil ---

Nik Taylor wrote:

> dunn patrick w wrote:
> > for instance, we
> > remember a phone number like 666-1369 better than, say, 238-9820.
>
> Well, 666 is a number that tends to be remembered.  :-)  But seriously,
> that doesn't mean that those numbers don't exist, does it?  They're
> three identical numbers, subject to "compression", so to speak, as I see
> it, similar to the way that file-compressors work.
>
> --
> Happy that Nation, - fortunate that age, whose history is not diverting
> -- Benjamin Franklin
> http://members.tripod.com/~Nik_Taylor/X-Files/
> http://members.tripod.com/~Nik_Taylor/Books.html
> ICQ #: 18656696
> AIM screen-name: NikTailor
>