LINGUIST List 8.736

Fri May 16 1997

Disc: Functionalism

Editor for this issue: T. Daniel Seely <seelylinguistlist.org>

Directory

1. Rick Wojcik, Re: 8.732, Functionalism
2. Rick Wojcik, RE: Functionalism 8.732
3. Geoffrey S. Nathan, Re: 8.732, Disc: Functionalism

Message 1: Re: 8.732, Functionalism

> From: benji wald <bwaldHUMnet.UCLA.EDU>
 [responding to some statements I made about Stampean Natural Phonology]
> >an initial [pt] target ... For Natural Phonology,
> >the fortitive epenthesis gives rise to the 'intuition' that you can't
> >have initial /pt/ clusters.
> 
> In discussion now these have been called, as above, /pt/
> clusters (English "*pot*ato" with a "reduced vowel"). I don't
> get it. To me, an "English" (if not universal) initial
> "cluster" only *releases* the last consonant of the cluster
> (cf. the "p" in the non-initial /pt/ "cluster", as in "aptitude"
> is not released). But the /p/ is released in "potato" -- or
> does someone claim otherwise?

My statement only refers to an attempt to pronounce an initial
tautosyllabic sequence [pt], not what might loosely be transcribed
as [pt] in the pronunciation of "potato". We can probably agree
that the initial syllable of that word is pronounced rather 
inconsistently by speakers. The important point is that speakers
sometimes produce what phoneticians describe as [pt] for the 
initial syllable of "potato". The question, if I remember correctly,
was why speakers should be able to pronounce the reduced [pt]
if they couldn't pronounce the unreduced one. I was merely trying
to explain how the theory of Natural Phonology would account for that.
Epenthesis, as a fortition, would not naturally apply to the output
of lenition processes. This point is made over and over again in
Stampe and Donegan's 1978 paper "The Study of Natural Phonology" in 
the Indiana U. Press Dineen volume.

> In a "real" English cluster, like
> /st/, not only is nothing but the final consonant released, but
> there is no delay in onset of vowel voicing following it, e.g.,
> in "stay" (i.e., there is no what used to be called "aspiration"
> of the consonant before the stressed vowel). In /pt/ato, the t
> release does lead into delayed vowel voicing, as it should for a
> SINGLE consonant before a stressed vowel. So talking about a
> "cluster" seems to be loose talk, and I can't tell if that is
> leading to a pointless discussion expressed in the following
> apparent paradox:
> 
> >The fact that you can't pronounce the cluster in no way means that the
> >cluster is banned as a side-effect in the pronunciation of some other
> >phonetic target.

Ouch! Pointless? I'll agree to your argument that "cluster" is being
used in a "loose" sense. I don't think that more precise wording would
have changed the point, although I apologize for any lack of clarity
on my part. 

> >Traditional phonological theory captures constraints
> >of this sort as static 'phonotactic' conditions on clustering, not as
> >behavioral constraints on linguistic articulation.
> 
> Indeed, and we find that most (?) English speakers treat the
> /kn/ in "knish" like the /pot/ (NOT /pt/) in "potato", and say
> "knish". They don't say, e.g., /a-knish/, putting an
> anaptyctic vowel in front, as if they associated the initial
> /kn/ cluster with medial /kn/ as in "acne". With respect to
> this, we note that Spanish speakers learning English do tend to
> supply an anaptyctic vowel to pronounce "stay" as "e-stay".

Quite true. BTW, I have made the point in the past that initial [zb]
in foreign names (e.g. Zbigniew) is subject to *either* epenthesis
[zbIgnu] or devoicing+assimilation [spIgnu]. In Natural Phonology,
the ban on initial [zb] clusters is not a static constraint, but
one based on articulatory substitutions of this type. I have been
told that Hindi speakers sometimes prefer "i-school" and sometimes
"si-chool" for the English words. I can't really tell you why English
speakers systematically avoid the anaptyctic solution, but it is a
very interesting question for any phonological theory. I'm not sure
that the point is relevant to this discussion, unless it would somehow
favor markedness-based theories of acquisition over process-based ones.
I do believe that the answer lies in a discussion of prosody,
and this thread has tended to concentrate on the level of segmental
representation.

> Meanwhile, Swahili speakers take Arabic loans like stahili
> "deserve" and English stimu "steam" and converts the /st/
> clusters into /sit/, an extra syllable, where the vowel tends to
> be voiceless, a common Swahili phonological process for as vowel
> to undergo in between two voiceless consonants (and also between
> a voiceless consonant and pause). Devoicing of vowels in
> certain contexts seems to be as "natural" as voicing consonants
> between vowels, leading to a contradiction for a theory which
> thinks it has to choose between C -> voiced /V_V and V ->
> voiceless /C_C (where both C's are voiceless)...

OK, but why would language learners have to make such choices in
a vacuum? Wouldn't they be able to observe what other speakers of
the language do in those situations and act accordingly? English
learners have to resist both speech impediments (i.e. processes)
if they want to pronounce English correctly, don't they? (Under
the assumption that both processes afflict the speech of the learner.
This assumption holds in Natural Phonology for pristine learners.)
The theory says that learners strive to eliminate impediments to
desired pronunciations. Processes are essentially speech impediments.
When you eliminate the ones that impede proper articulation, the 
residual system is your adult phonology. This point is fundamental
to understanding what the theory is about. Stampean theory predicts
that very early language learners will have both devoicing of 
vowels between voiceless consonants and voicing of consonants between
vowels. (In fact, such processes do occur in L1 acquisition, but they
are not observed--may not even be manifested--in all learners. Having
them doesn't mean that they actually have to become dominant during
phonological acquisition. They do in some learners, but not all.)

I know that Stampean theory is very different--quite strange to some
people. It requires one to treat acquisition as a kind of "loss" of
pre-existing rules. Phonological derivations are much more complex in
children and aphasics than in proficient speakers. It is
reasonable to view the existence of contradictory processes as a true
"contradiction" in a theory of markedness/competence. However, the
existence of such contradictory processes is actually "a feature, not
a bug" in Natural Phonology. After all, as Wald stated, such 
processes occur quite naturally in the world's languages. If markedness
theory were really correct, then why didn't the human race converge on
the same phonology ages ago? What keeps phonologies different if there
is some kind of gold standard that we all use to arrive at grammatical
analyses with? Natural Phonology simply takes the view that we put
naturally-occuring, but chaotic behavioral constraints, into some kind
of order, depending on what challenges the target language poses for
the articulators. What you do with contradictory processes depends on
what you have to say.

- 
Richard H. Wojcik
Boeing Information & Support Services Natural Language Processing
P.O. Box 3707, MS 7L-43, Seattle, WA 98124-2207 (phone: 425-865-3844)
richard.h.wojcikboeing.com (fax: 425-865-2965)
- --Opinions expressed above are not those of The Boeing Company----

Sorry to reply to my own post, but I need to correct a reference in
it:

Richard Wojcik wrote:
> I was merely trying to explain how the theory of Natural Phonology
> would account for that. Epenthesis, as a fortition, would not 
> naturally apply to the output of lenition processes. This point 
> is made over and over again in Stampe and Donegan's 1978 paper "The 
> Study of Natural Phonology" in the Indiana U. Press Dineen volume.

The correct reference is:
 Donegan, P. and D. Stampe. 1979. "The Study of Natural Phonology"
 in D. Dinnsen, ed. Current approaches to phonological theory.
 Bloomington, Indiana U. Press.

Message 3: Re: 8.732, Disc: Functionalism

At 10:05 AM 5/15/97 -0400, Benji Wald wrote:
>
>In discussion now these have been called, as above, /pt/
>clusters (English "*pot*ato" with a "reduced vowel"). I don't
>get it. To me, an "English" (if not universal) initial
>"cluster" only *releases* the last consonant of the cluster
>(cf. the "p" in the non-initial /pt/ "cluster", as in "aptitude"
>is not released). But the /p/ is released in "potato" -- or
>does someone claim otherwise? In a "real" English cluster, like
>/st/, not only is nothing but the final consonant released, but
>there is no delay in onset of vowel voicing following it, e.g.,
>in "stay" (i.e., there is no what used to be called "aspiration"
>of the consonant before the stressed vowel). In /pt/ato, the t
>release does lead into delayed vowel voicing, as it should for a
>SINGLE consonant before a stressed vowel.

	While it is true that syllable-final stops are unreleased when
followed by other consonants in English, it is by no means clear that
all first consonants in clusters are unreleased--stop plus liquid and
glide clusters are indeed released, with the aspiration spreading over
the liquid/glide. Spencer's Phonology text has an extensive
discussion of these cases: please, cry, quick, cute (assuming one
thinks the /y/ in 'cute' is in fact a part of a cluster). Whether a
language releases a consonant in contact with a following one is a
language specific, and even level-specific question. The [pt-]
cluster in 'potato' may not be a *phonological* (that is underlying)
cluster, but on the surface it's indistinguishable from a real cluster
(modulo aspiration facts) in languages that have them (such as Greek).
And the reduced version of 'I dunno' which is pronounced [dnou] with
nasal plosion is in fact the same as the Russian [dno] 'bottom'
(except the Russian /d/ is fully voiced).
	So, the English [pt-] is *phonetically* a cluster, but
phonologically not--native speakers perceive a vowel there, even if
they don't make one. That's the whole point Sapir was making in his
famous article about 'psychological reality of the phoneme', and
similar claims have been made by Natural Phonology (as Rick Wojcik
pointed out) and by various versions of Lexical Phonology--the output
of post-lexical processes are by and large not perceived by native
speakers, while lexical processes (which are in general
structure-preserving) ARE perceived, because they amount to switches
between phonemes.

Geoff

Geoffrey S. Nathan
Department of Linguistics
Southern Illinois University at Carbondale,
Carbondale, IL, 62901 USA
Phone: +618 453-3421 (Office) FAX +618 453-6527
+618 549-0106 (Home)