LINGUIST List 2.146

Thursday, 18 Apr 1991

Disc: Phonology

Editor for this issue: <>

Directory

1. , Re: Halle on the Phoneme
2. , Halle and Baudouin
3. , Re: (Morpho)phonology
4. Mike Hammond, Phonology
5. "Larry G. Hutchinson", Re: (Morpho)phonology

I sometimes think that Halle's argument is difficult to follow because most
of the arguers do not share Russian intuitions about how to pronounce things.
First of all, let us note that the voice assimilation process applies to 
Russian-accented English. "Nice boy" is pronounced 'ni[z]e boy'. This does
not mean that the Russian speaker thinks that the English word 'nice' ends in
/z/. Baudouin's phonological level (called "phonemic" by him) would allow 
him to analyze 'nice' with an /s/ phoneme (=Ulaszyn's morphophoneme) without
giving up the claim that the [z] derivative had some kind of perceptual 
autonomy. The whole point was that phonemes are supposed to be "janus-like"
in character--able to serve the speaker and hearer alike. So he didn't use
the term consistently to represent cases of neutralization, since the 
neutralization issue only addresses the listener's viewpoint, not the 
speaker's.

But it is extremely important to compare voice assimilation alternations
with those connected to Russian allomorphy--fleeting vowels, consonant
shifts involving historical waves of palatalization, other alternations
connected to historical loss of the yers, etc. None (zero) of those 
phenomena play any role in the pronunciation of English, because they
are not phonological (physiophonetic) operations that govern Russian 
pronunciation. They involve operations that govern how Russians relate
morphemes to each other. This fact lends a certain compelling plausibility--
and I dare to call it "intuitive" even--to the alternational dichotomy that
Halle did not address with his argument. He struck down the wrong concept
of the phoneme--the one that was shorn of its speaker-based functionality.

One last remark: most of my knowledge of Ulaszyn comes from Alexis. So I
don't challenge what he has said about this late Polish student of Baudouin's.
But my impression is that his work was closer to Moscow than Leningrad. My
reasoning is that the Leningraders really did confuse the physiophonetic/
psychophonetic alternations when it came to cases of automatic neutralization.
I seem to recall that Ulaszyn did not confuse his morphophonemes with cases
of psychophonetic alternations. This puts him closer to Moscow, which 
distinguished allophonic cases as 'varijatsija' and neutralizing cases as
'varianty' without giving up Baudouin's essential level of abstractness.

 -Rick Wojcik (rwojcikatc.boeing.com)

(1) Avery Andrews writes that he finds Halle's argument against
(Leningrad) phonemics uncompelling because, unlike other simplicity
arguments, he does not see the single mental structure behind the
phenomena which the L-phonemic theory separates and which Halle
saw as being a single phenomenon. Am I missing something? The
The point, although Halle did not originally make it explicit (I think),
is that we do find strong evidence that allophonic and neutralizing
processes are the same, viz., they never appear (or disappear) one
without the other. It is true, of course, that there is some (more
or less anecdotal) evidence that speakers' minds find some difference
between the effects of these processes, and indeed that is precisely
what made Ulaszyn introduce the L-phonemic theory in the first place.
However, the argument for L-phonemics based on speaker perceptions of
sameness and difference is not very strong, I feel, because there are
many cases on record of speakers perceiving subphonemic distinctions
(e.g., Polish and Russian i vs. y in the days when they were in complementary
distribution) and there is some (much less) evidence of cases where
phonemic distinctions are not perceived by speakers (Labov's work
on the vowels of saw vs. soar in NYC r-less speech, for example).
It seems to me entirely possible that what speakers are conscious of
is something orthogonal to the whole issue.

(2) Rick Wojcik takes some exception to my statements about Polish
and Russian i and y and their relevance to the fundamental questions
of phonology. The facts that he alludes to regarding the disputes
among Russian linguists about the phonemic status of this distinction
only confirm my original point that this is case where native speaker
perceptions are at odds with (many) linguistic theories. His contention
that speakers perceive these as different because of the spelling can
easily be refuted by the following arguments:

 (a) In Russian, [y] is sometimes spelled 'i', specifically after
 'c', 'sh', and 'zh'. Speakers find no difficulty realizing
 that these are cases of [y].

 (b) If spelling were enough to help speakers identify allophones
 as different, then teaching a speaker a phonetic alphabet
 in which other allophones are also spelled differently should
 have the same effect, yet Russian speakers have enormous
 difficulties perceiving the fronted allophones of back vowels
 after palatal(ized) consonants even after extensive training
 in phonetics. 

 (c) In the Middle Ages, the Latin-based Slavic writing systems
 did not have a consistent distinction between [i] and [y],
 and this was deliberately introduced by reformers (including
 Jan Hus for Czech, I seem to recall, since Czech had the
 distinction too at one time) who were bent on reflecting the
 "pronunciation" (i.e., their mental image of the pronunciation).

I would add that the fact that initial [i] goes to [y] in Russian
(but not Polish) after a word ending in a non-palatal[ized] consonant,
while perfectly true, does not have any bearing on the question of
whether these are allophones or phonemes. For the question is precisely
whether this alternation is allophonic or neutralizing.

(3) Bringing the two points I just made together, it seems to me that
instead of arguing about the "correct" phonemic analysis (as the Moscow
and Leningrad phonologists have all these years), we might reasonably
ask whether the notion of phoneme is all that wonderful. The point
is that the examples under discussion, far from being some exotic
recent find, were precisely the examples that PHONOLOGY WAS INVENTED
TO ACCOUNT FOR. It seems to at least equally plausible that the
right theory of what speakers perceive as same or different has to
do with some notion of salience and phonetic similarity rather than
with distribution. Stampe's natural phonology would be the KIND of
theory that we would then want, since it claims, for example, that
only alternants created by LENITING natural processes are perceived
as the same as the things they come from, whereas sounds derived by
FORTITION are not. If we knew exactly what lenitions and fortitions
were, we might then find that Polish and Russian [i] and [y] cannot
be derived from each other by any possible lenitions and this is
why they are perceived as different. 

I said "KIND of theory", however, because (a) I am not sure that
we are told clearly enough what lenitions and fortitions are and
(b) I am not sure that this is enough. As to (b), I think American
speakers perceive flaps as quite different from t's, and yet the
former would appear to be lenitions of the latter. But, even if
I am right, Stampe's is perhaps the ONLY model of phonology which
is of this KIND.

Finally, it is absolutely vital not to confuse various kinds of
evidence for sameness or difference of two sounds. As the i/y
business shows, sounds can rhyme w/o being perceived by speakers
as the same. Incidentally, the names of the letters in Polish
and (recent) Russian are [i] and [y], a minimal pair that several
Leningrad phonologists have noted. Yet to my ear the following
doggerel is perfectly rhymed in Polish (perhaps other speakers
will comment?):

 Pokazesz mi 'You will show me'
 Litere y 'The letter y'

and

 Pokazesz ty 'You will show (to someone unspecified)'
 Litere i 'The letter i'

(Please do not cite these w/o due care, because I have omitted the
diacritics.)

While we're considering 'kudo' and 'Hade', let's also consider 'bicep'
and 'quadricep', as in 
 My left bicep is stronger than my right one.
 My quadriceps are getting stronger. (The person who gave me
this told me that we have four quadriceps, and, of course, two biceps.)

I've also heard mention of 'quads' (and 'gluts' and 'abs') but I don't
know whether I can say that I have a left 'quad', but this is an aside.

Hmm. Do we have 6 ceps altogether? I don't think so. We also have
triceps, but I don't think many people know where they are. In fact,
I suspect more people know where quads are than where quadriceps are.

Christine Kamprath

Alexis Manaster-Ramer has recently objected to treating words like
_canasta_ as exceptions to the English stress rules. He maintains:

"For, exception marking is simply, in reality, a device for indicating
that something is unpredictable but not not quite admitting it. My
own position is that exception marking makes sense if there is factual
evidence for the special status of the exceptions. I know of two
kinds of evidence of this kind, but there may be others. One, the
exceptional forms form a closed, unproductive set. Two, the
exceptional forms are perceived as foreign. In the cases, we are
discussing, I think there is no such evidence, and hence to claim that
English stress is predictable is similar to saying that Mandarin has
only three tones (the fourth being due to exception marking), or that
there is no /s/ in English, it is really, say, /m/ but with exception
marking."

In fact, words like _canasta_ are exceptions in Manaster-Ramer's first
sense. That is, while it is possible to exceptionally assign stress to
a light penult of a noun, it is not possible, for example, to
exceptionally assign stress to a preantepenult of a noun. Hence, it
makes sense to treat words like _canasta_ as exceptions, rather than
treat stress in the entire English lexicon as lexically marked.

He goes on to argue against treating words like _industry_ as
containing a final /y/ because:

 (a) "the resulting consonant clusters are highly unusual 
 (b) there is no basis in alternations for such a contrast
 between /y/ and /I/, i.e., it is not true that words
 with /y/ resp. /I/ show up with these as phonetic values
 before some suffix (for example). 
 (c) there is also no basis in analogy to other alternations.
 For example, it has been suggested (but I forget whether
 this is SPE or someone else, maybe McCawley??) that final
 -er can be derived from either /r/ or /Vr/, that this
 can be used to account for the stress pattern
 of mInister (cf. ministr-y, so underlyingly /mInIstr/,
 or carpenter (cf. carpentr-y). However, this won't work
 either because of examples like pilAster (cf. pilastr-ade)."

I don't think any of these objections go through.

First, it's not clear what the underlying naturalness of these
clusters has to do with anything. There are other examples in English
where underlyingly unnatural clusters have to be posited as well, as
evidenced by the following alternations:
 hymn hymnal
 paradigm paradigmatic
 syntagm syntagmatic
 gnostic agnostic
 knowledge acknowledge

Second, the absence of surface alternations is also not to be taken as
criterial for positing underlying representations. For example, the
fact the the initial aspirated /t/ of <top> doesn't alternate doesn't
prevent us from positing a rule of aspiration.

Third, the other alternations involving syllabicity do seem to go
through. Manaster-Ramer cites alternations involving orthographic -er.
 minister ministry
 carpenter carpentry
 pilaster pilastrade
He cites _pilaster_ with penult stress as a problem for the
claim that the word is underlyingly /pilastr/. This is not a problem
at all, however, as nothing prevents us from treating _pilaster_ like
_canasta_, as being marked to attract stress. In /pilastr/, it would
be the ultima that's marked to attract stress, but that option must be
open in any event to account for contrasts like the following:
_narthex_ vs. _helix_.

mike hammond

Halle's argument against the phonemic level of representation can can very
easily be seen rather as an argument against the sequential application
of phonological rules.

Chomsky's argument is logically flawed, given his own definitions. This is
independent of the question of whether or not these definitions actually
fit the linguistics of any real phonologists.

An elderly paper of mine on these two topics has recently been reprinted
(not updated and slightly garbled typographically) in Linguistic Research,
Vol. 9, 1988. This is the "Mr. Chomsky on the Phoneme" paper reviewed in
Hymes and Fought, starting on page 199 (contrary to what the index says).

[End Linguist List, Vol. 2, No. 146]