LINGUIST List 11.616

Sun Mar 19 2000

Disc: Underlying Schwa?

Editor for this issue: Karen Milligan <karenlinguistlist.org>

Directory

1. benji wald, Re: 11.584, Disc: Underlying Schwa?
2. jakob, unstressed I vs

Message 1: Re: 11.584, Disc: Underlying Schwa?

List discussion of the problem of "underlying schwa" (in English) has had,
in my view, a tendency to become overly scholastic. By scholastic I mean
arguing "theoretical" points without any hope of resolving disagreements
about them on the basis of empirical data. Clearly, this does not apply to
the many discussants who have reported phonological distinctions in their
unstressed vowels, e.g., those who agree with Trager & Smith (and ME) that
"ROSES" and "ROSA'S" etc etc are distinct -- not that I claim I would be
able to consistently recognise the difference between these two vowels in
an ABX labelling test (though I'm not bad at that sort of thing -- so maybe
better I should say I don't expect that all, maybe even most, speakers who
make the distinction, can make use of the distinction in hearing/decoding
tasks; note, then, that I suggest it is possible for some distinctions to
be the result of "historical residue" without being fully "functional" in
the auditory performance system). Similarly, my charge of scholasticsim
does not apply to those speakers who have pointed out that not all
unstressed vowels are "reduced" -- or always reduced, e.g., the first vowel
of "repeat" etc, the last vowel of "window", or "Sunday", etc etc etc. I'm
having a difficult time appreciating Theo's point that "reduction"
automatically means "to schwa", maybe this is just a "dialect-difference"
in how we understand the term "reduction". I understand it to mean the
SPACE in which contrasts may STILL be made is reduced, such that there are
fewer distinctions, and, as the history of English (inter alia) informs us,
great potential for eventual merger and even loss.

Most scholastic to me are the kinds of arguments that revolve around such
considerations as whether certain unstressed vowel should be associated
with certain stressed vowels, e.g., the second vowel of "syllable" with the
stressed vowel of "syllabic". Now, of course, we all wonder what happened
to the "le" in "syllabic" (as in the French version of Aristotle's name),
but the point is: do we have EVIDENCE to associate those two vowels? The
most scholastic view is that behavioral evidence is unnecessary; that the
notion of invariant sound:meaning correspondence is sufficient. There need
not be behavioral consequences. Yeah, that gets pretty close to arguing on
an aesthetic basis -- literally, i.e., "I feel it", to which the response
is "yeah, well I feel you're wrong" -- but there's no possibility of
disconfirmation (-- and maybe I don't really feel you're wrong, but PROVE I
don't; since when is the ability to resolve such disagreements, even if not
in the name of scientific method, dependent on accepting that claims are
made in GOOD FAITH?)

Now Theo did allude to some kind of behavioral evidence but rejected it as
in an invalid kind. That's the evidence of the spelling. Maybe he's
suggesting that the spelling is a kind of "brainwashing". You think the
"a" (uh the "A"?) of "syllAble" is the same as of "syllAbic" because
they're spelled the same, and the words have a common meaning -- but you're
wrong. Well, I happen to agree with Theo, but he didn't tell you why
you're wrong. So he didn't finish the job of disagreement that he started.
I looked in vain for anyone alluding to more relevant behavioral evidence
for phonological identifications, but, of course, there is such, well known
to historical linguists, in the form of NONSTANDARD spellings. These
spellings contradict the "morphophonemic-as-phonemic" principle. I haven't
really seen the spelling "syll*i*ble", but it wouldn't surprise any more
than "passible" for "passable" or "computor" for "computer" -- or is it the
other way around? You scornfully say, "bah! what kind of evidence is that
- you haven't seen...." (throwing my own words back in my face). Well,
TRUE STORY, decades ago when I was very much excited by generative
phonology (and I'm still interested in sound change and internal
reconstruction) a colleague asked me "I forgot. How do you spell it --
"repetition" or "repitition"?" The most reliable way of verifying
spellings (short of the humiliation of looking them up in the dictionary)
is by writing them down and seeing if they "look right". But under the
influence (of what?), I said thoughtfully: well, let's see, it's
"repetitive", not "repititive" [SAYing a STRESSED short "e" rather than a
short "i"], so it must be an *E*, not an *I* [this time saying the names of
the letters]". Yes, these sound:meaning relations help a lot with
spelling, and also with reconstruction of earlier sound systems, but
misspellings and hesitations show a different level of organisation of
sound systems -- the one which Theo seems to have in mind, the one in which
spelling can indifferently represent a "schwa" (or an unstressed vowel) as
"i" or "e".

Conclusion: I don't know. Maybe there are various levels of phonological
or at least phonologically related organisation in English and various
other languages, maybe enough levels to please any scholastic (as long as
s/he doesn't mind if someone else is right too -- yeah, I know, there's
still gonna be the problem of which insight is more important, significant,
productive, even brilliant, less trivial, etc). OR, maybe the ability of
spelling to influence a speaker's "underlying phonological representation"
of a word or set of words is superficial, if at all meaningful, and easily
undone by the pressure of a more direct and earlier formed phonological
system (e.g., a system formed early enough to get something like the
alphabetic principle of letter or group of letters referring to a
particular linguistic "sound" -- you thought I would dare use the word
"phoneme"? , not until I'm assured that discussion is over -- but too early
to codify morphophonemic relationships).

My last parethetical comment brings us back to where I think the discussion
belongs, and indicates that there is still plenty to argue about, but the
thrust of my mention of nonstandard spelling is to suggest that the
discussion try to resolve disagreements by searching for evidence that
MIGHT resolve the disagreements, instead of despairing of such methods and
standing (or falling) on aesthetic judgments masquerading as "intuitions"
as "evidence".

PS. Theo allowed a PROCESS of reduction of vowels to "schwa" (by
definition "to schwa", if I understood his dialect of "reduction" right) in
the case of sentences but not words. The crucial battleground is then
COMPOUNDS, with such considerations as how "waistcoat" (or "forecastle") is
pronounced and whether "daisy" (< day's eye) ever rhymed with "rabbi".

The contributions to this discussion have been illuminating, particularly
the
latest by Wheeler and Vennemann. The former's list of contrasts are about
as one would expect, with a few exceptions. His variation in "palate" may be
due to analogical influence from "pallet". While skipping over the
de-roticized
syllables I feel unqualified to judge, there remain "chicken" (I) vs.
"thicken" ()
and "tendril (I) vs. "spandrel" (). Now, I am very curious: what is the
underlying difference here that causes this surface variation? Even for
the
"easy" cases which Wheeler and others have mentioned, that has not been
made very clear, let alone for the above "problem cases".

Meanwhile, Vennemann says: " A reduced syllable of a word is one whose 
nuclear position permits no individual vowel color."

According to this, there should be no room for a variation between short I
and  in these words (or am I misunderstanding something here?)
Nonetheless, this variation of course does exist in the minds and speech
of Wheeler and others. Please understand: I am not accusing some
speakers who have differrent behavior of being "phonetic phonies", but I
would say:
1) It is possible and frequently happens that people have notions about
their linguistic behavior which are not confirmed by objective observation.
2) This area of behavior under discussion here is particularly susceptible
to such divergence. (In Chinese there is a similar issue of actual
pronunciation
of reduced vowels vs. speakers' opinions on their own behavior).

I am also quite capable of pronouncing Wheeler's list with the two different
vowel sounds, but if I do so, I notice that the length-ratio between the
stressed
and unstressed syllables is not the same as in my "natural" pronunciation.
I.e. I can only hear a difference in vowel-quality if the length of the
so-called
"reduced" syllable is slightly longer than I would normally assign to such a
syllable, and the only way that would happen would be to give it special
emphasis. But is there then a whole set of speakers who have a naturally
occurring "semi-reduced" syllable-type in which this contrast, and only
this contrast, can surface? 

The question of precise length can be measured instrumentally, and
although it would be a good piece of work to do that on a
statistically significant sample of cases in UNMONITORED speech, until
then, all we have is some people attesting that their behavior is
such-and-such, or others such as myself grumpily saying
"Maybe....maybe not".

- -Prof. Jakob Dempsey
 Dept. of Applied Foreign Langs. & Linguistics
 Yuan-ze University, Taiwan