[From: Bruce Nevin (Tue 931012 15:01:48 EDT)]
Rick Marken (931008.1530) --
What is the relation between phonemes and contrasts? That seems to be a
critical issue in our discussion.
I suggested modelling control of phonemes by their mutual contrast within
a "space" that is defined in two ways: (1) Physically by the
environmental properties of the vocal apparatus and the atmosphere
between mouth and ear. (2) Culturally by some useable parts of that
"space" being ruled out of bounds. As an example of (2), German and
French allow use of parts of the vowel trapezoid that English does not
allow (for front rounded vowels).
This is not the whole story. Less contrastive (less careful, less
emphatic, etc.) speech appears to reduce the envelope all around. People
seem to control a perception of themselves sounding like the way such and
such kind of person talks, a matter of phonetic "style" below the level
of the phonemes. The number of points of contrast may differ between two
dialects of the same language (I distinguish pin and pen, and I
distinguish pod and pa'd as in pea pod and "pa'd go if he could", many
midwesterners don't). The phonemic shapes of individual words may differ
from one dialect to another, over and above the differences in
pronunciation of the phonemes. Control of a given phoneme in a word
leaves articulators in a position which amounts to environmental
disturbance of control of a subsequent phoneme in that word
("coarticulation effects"), and anticipation of the latter may even
disturb control of the former
But the basic idea of phonemes as contrasts is pretty straightforward, I
think.
I'll stop there, and not send the longish reply that I came up with at
first. I'm having trouble keeping up, and feel that I have not been able
to read recent posts in this thread as carefully as I need to. Maybe
later.
Bill Powers (931010.1130 MDT) --
Yes, identifying precise reference signals for phonemes would require
more than the pair test. If the above is correct, going about it as you
suggest would not be successful. The reference signals for lower-level
phonetic control would vary depending upon other factors such as those I
have identified. The reference signals for vowels identified in the
experiments reported in _Language_ would I strongly suspect be
stigmatized as unnatural sounding if hearers were asked for their
judgements of words or syllables containing the vowels, instead of
picking the vowels in isolation from a rich palatte of vowels (F1/F2
combinations separated near the limit of discriminable differences). But
the pair test does give you the contrasts, from which (if the converse of
the Gather program works) the specific reference signals are predictable.
The variation of reference signals in actual speech (as opposed to the
isolation of hyperarticulated targets) brings in the other factors.
Ditto re what I said to Rick about a longer reply now on the shelf.
How's that? Short and clear, I hope. I haven't been doing at all well
with the long and groping around kinds of posts.
Bruce
bn@bbn.com