[From: Bruce Nevin (Fri 93115 15:36:24 EST)]
Bill Powers (931103.0900 MST) --
Two ideas about what the controlled variable is if one controls "for"
contrast:
1. Contrast itself is the object of control
2. Changes in an external measure called contrast are a consequence of
controlling for something other than contrast.
<Discussion in each case in terms of contrast meaning difference between
pronunciations of words.>
This is how I understand it. I could be wildly off the track.
1. We control pronunciations with respect to articulatory targets, maybe
acoustic too.
2. The values of reference perceptions for pronunciation ("targets" in (1))
are set in such a way that (as it turns out) the perceived differences
between targets are maximized. This is termed contrast, as externally
observed.
3. It may be convenient to think of the reference perceptions or targets
as the phonemes.
4. The contrast observed from the outside in (2) might be a result of
control of the contrast between targets, as in the first proposal that
you paraphrase above. However, this control would not be brought about
by varying pronunciations directly, but by varying the reference settings
or targets for pronunciations (with an indirect effect on pronunciations).
5. The externally-observed contrast in (2) might be a result of the
learning process, and its extension into ongoing adaptation of one's
pronunciation norms according to one's social situation. Same caveat as
in (4) about controlling settings rather than controlling pronunciations
directly.
6. As gain is increased, actual pronunciations more closely approximate
their reference "targets". Such pronunciations may be said to be more
"contrastive". Conversely, as gain is decreased, pronunciations of
different phonemes (and of different words) are less distinct from one
another, and such pronunciations may be said to be less "contrastive".
This use of the notion of "contrast" should not be confused with the
contrast (2, above) of phonemes (3) and of words made up of phonemes.
The phonemic contrast between pen and pan does not increase with
hyperarticulated pronunciation or decrease with "hypoarticulated"
pronunciation of those words.
7. Just so, pronouncing the phonemes in a word does not depend upon
perceiving (even in imagination) phonemically different words and
controlling the contrast between the desired word and the alternatives.
The reference settings for phonemes are pre-set when you undertake to
pronounce a word. As to how they get set, and what that has to do with
contrast, reread (4, 5).
8. It appears to be a bit trickier when you recognize a word from someone
else's pronunciation, where their targets differ from your own. If your
targets and theirs are the same, we can see how an approximation to the
target is normalized, perhaps with something like Martin's "trajectory"
account. But if their targets differ markedly, some additional
recalibration may be needed. This is reflected in a learning interval
before one is able easily to understand someone with a "strong accent".
Perhaps there is some kind of mapping from "my pronunciation target for
phonemes a, b, c" to "their pronunciation targets for phonemes a, b, c".
There certainly seems to be little doubt that they are perceived as
different pronunciations of the same words, and even that they are
perceived as being made up of the same phonemes differently pronounced
(though evidence familiar to me here is shakier). In other words, my
sister in law's Georgia pronunciation of the vowel part of "can" is perceived
both by me and by her as the same phoneme as my pronunciation of it. Even
though if I now pronounced that sound it would be for me three phonemes
(two vowels with a y between them), it is clearly her pronunciation of a
single vowel phoneme when I hear it. Perhaps this is normalization on
the basis of word recognition, but it doesn't feel that way, and anyway
wouldn't that beg a question or two?
9. The targets used in speaking may begin to shift toward those used in
recognizing a different accent. This takes even longer. Refer again to
(5) above. During this kind of change through time, at any given point
in the process of change, it appears that targets for pronunciation are
always maximally differentiated. This looks like control maximizing
their distance from one another in the available phoneme-space.
It would be very nice to verify this experimentally. I could be dead
wrong. But this is my understanding.
Did cases like this show up in the contrast experiments?
Not sure what experiments you're referring to.
Bill Powers (931030.1230 MDT) --
Diphthongs: There's a timing contour that makes the difference between
the vowel part of a diphthong and the glide part. Compare:
we glide u, vowel i
phooie vowel u, glide i
you glide i, vowel u
eeuw! vowel i, glide u
Off on my afternoon marathon home. Trying slowly to catch up.
Bruce