[Sugar-devel] Phonology (was Re: Fwd: Summer of Code Proposal: Furthering Speech Recognition in Sugar.)

[C. Scott Ananian]

On Tue, Mar 24, 2009 at 10:57 AM, Sean DALY <sdaly.be at gmail.com> wrote:
> :-)
>
> I'm no linguist, but we speak 3 languages in the house and I have been
> fascinated to observe that my kids seem to "borrow" phonemes between
> languages. I met somebody at a party once who said that has been
> documented in bi- and tri-lingual children, apparently as a way of
> "economizing" brainpower when small (<6 y.o.) Not sure how
> authoritative a party person is, though.

My second-hand "oral story from a linguist" is similar, and ties in
with Noam Chomsky's theories of universal grammar: very young children
apparently are born with a remarkable ability to distinguish phones
from all known languages, and as you age the brain starts to learn
which distinctions are meaningful and to "forget" the unmeaningful
distinctions, presumably in order to reinforce recognition of the
remaining meaningful distinctions.  Hence a small child will be able
to distinguish between 'r' and 'l' and between 'k' and 'kh', an
ability which grown up Japanese- and English-speakers (respectively)
have lost.

Your story seems to deal with the flip side of this: the ability to
articulate (or not) those distinctions.  There's an interesting bit of
optimality theory/phonology about the "markedness" of various sounds:
universally across languages some sounds (and syllable structures!)
are "more basic" and others "more complicated".  Children usually
progress in a straightforward order from unmarked through marked
structures as they learn: for example, vowel-consonant syllables like
"ma" as less marked than consonant-vowel-consonant syllables like
"dad", which are less marked than syllables with complex consonant
clusters like "schmooze".  Perhaps your kids are borrowing less-marked
alternatives from their other languages as a handy substitute for the
marked element they can't quite get their tongues around yet.
 --scott

(As an obligatory comp sci aside: one of my favorite projects when I
was a "comp sci graduate student studying phonology" was a set of
processing scripts which grunged through the entire contents of
Project Gutenberg analyzing poetry "slant rhymes" for quantitative
evidence to support the markedness of certain phonological structures.
 The theory went that, given a non-exact rhyme, a poet should prefer
those which differ in "unmarked" qualities over those which differ in
"marked" qualitities.  http://cscott.net/Publications/proj4.pdf )

-- 
                         ( http://cscott.net/ )