11.155, Disc: Phonemic Analysis

[LINGUIST Network]

LINGUIST List:  Vol-11-155. Mon Jan 24 2000. ISSN: 1068-4875.

Subject: 11.155, Disc: Phonemic Analysis

Moderators: Anthony Rodrigues Aristar: Wayne State U.<aristar at linguistlist.org>
            Helen Dry: Eastern Michigan U. <hdry at linguistlist.org>
            Andrew Carnie: U. of Arizona <carnie at linguistlist.org>

Reviews: Andrew Carnie: U. of Arizona <carnie at linguistlist.org>

Associate Editors:  Martin Jacobsen <marty at linguistlist.org>
                    Ljuba Veselinova <ljuba at linguistlist.org>
		    Scott Fults <scott at linguistlist.org>
		    Jody Huellmantel <jody at linguistlist.org>
		    Karen Milligan <karen at linguistlist.org>

Assistant Editors:  Lydia Grebenyova <lydia at linguistlist.org>
		    Naomi Ogasawara <naomi at linguistlist.org>
		    James Yuells <james at linguistlist.org>

Software development: John H. Remmers <remmers at emunix.emich.edu>
                      Sudheendra Adiga <sudhi at linguistlist.org>
                      Qian Liao <qian at linguistlist.org>

Home Page:  http://linguistlist.org/


Editor for this issue: James Yuells <james at linguistlist.org>

=================================Directory=================================

1)
Date:  Fri, 07 Jan 2000 15:27:18 +0100
From:  Gilles Bernard <gb at ai.univ-paris8.fr>
Subject:  Phonemic Analysis

-------------------------------- Message 1 -------------------------------

Date:  Fri, 07 Jan 2000 15:27:18 +0100
From:  Gilles Bernard <gb at ai.univ-paris8.fr>
Subject:  Phonemic Analysis

Dear colleagues

I just read two months of Linguist delivery and I discovered this heated
discussion; I cant resist joining in, because I would like the discussion to
address some really problematic points in systematic phonology, related to the
dynamics of phonemic systems.

The original demand by Martin Selzmann was essentially about clarification of
the notion of complementary distribution.

First, generative phonology, which keeps creeping into this discussion, has
nothing to do here. It provides no methodology for describing new languages
from scrap, does not work backwards (from the sound to the phonemes, or, as is
the case, to the morphonemes), and has nothing to say about system dynamics and
evolution. So it sure has no problem at all (you can allways devise a rule
transforming something into something else), which fact is in itself, for me,
its biggest problem. I do not include Natural Generative Phonology in this criticism.

1. Let us begin with an idealized phonemic description of some unknown language,
and then point out some problems.

A- The linguist looks for words that seem to differ in only one point (minimal
pairs). He is thus able to propose lists of candidates-phonemes (they cannot be
called variants or allophones yet, and they arent phones either) in specific
environments.

The linguist does this work with trained ears: trained by his mother-language,
by his experience, by his theoretical views, and, last but not least, by the help
of a native speaker. All this will be a help as well as a hindrance.

B- If some candidates-phonemes are in complementary distribution, then the
linguist can choose to consider them as allophones of the same phoneme. Thus
complementary distribution (and allophonic relationship) is fundamentally a
methodological tool.

The global coherence of the proposed system also plays a part here. Thus, in
spite of their complementary distribution, h and ng are not usually conceived
as allophones of the same phoneme, because the phonetic difference between them
doesnt have any parallel in the English phonemic system (that seems to me more
to the point than the phonetic difference alone: I agree there with...), and
(phonetic counterpart) does not correspond to language-specific (or universal)
contextual influence (coarticulation and position influence).

Real problems begin when some candidate-phoneme is in complementary distribution
to two or more others candidates (as in the English flap case, or Chinese j, q, x).
In Prague phonemics, this is a case of neutralization: an opposition between
two candidates-phonemes does not appear in some context, where one one can be
found. This one can be one of the candidates-phonemes, as in German, where the
p/b opposition is neutralized at the end of words, and only p is attested, or
a third one, as in the cases discussed here. Some have introduced the concept
of archiphoneme - a very problematic concept.

In such cases, one can choose to merge the isolated candidate with one of the
others, or consider it as a third phoneme (the last choice is to be found in
Chinese official script, but not in the French Oriental School one, who has
taken a more etymological choice).

Two side-remarks:

- Some candidates-phonemes are always in complementary distribution: thus vowels
and consonants can only be opposed through rather artificial minimal pairs (and
only in some languages).

- I'd like to join with Dan against two much thingization of allophones: the
same phone can be an allophone of some phoneme in one context and be in opposition
to this phonem in another; for instance, if o/e are candidates-phonemes before w,
and e/i elsewhere, [e] is an allophone of /i/ before /w/ and and opposed to /i/
elsewhere. So a phone is not just a manifestation of a phoneme; the attribution
of some phone to a phoneme is the result of a recognition process (we should not
only look at things top-down, but also bottom-up).

2. Going further, it may well be the case that in language acquisition some
similar process takes place: on the basis of complementary distribution and
system coherence, different candidates-phonemes are constructed by the native
speaker as allophones of the same phoneme. Then complementary distribution would
be a cognitive basis for constructing more complex representations, and
candidates-phonemes would be an underlying strate in the recognition of phonemes.

As the native speaker is not at this stage influenced by another language, one
has to suppose that, if I'm right here, the candidates-phonemes are deduced from
general perceptual processes.

3. Evolution in phonemic systems may be partly explained as a difference of choice
in allophonic groupings.

In the case of French nasal vowels, at least three stages can be distinguished:
first (coarticulation), vowels were (partly) nasalized in front of nasal
consonants (/an/ => [aAn], where I take A to mean the nasal portion of the vowel);
second (assimilation), vowels were totally nasalized in this context
(/an/ => [An]); third, the nasal context disappeared, and the nasalized vowels
could then be found in the same context than other vowels. A fourth stage
included denasalization before nasal consonants, after simplification of the
nasal vowel system (E=> A): /femmX/ (X meaning schwa) [feEm-mX]=> [fEm-mX] >
/fE-mX/ > /fA-mX/ > /fa-mX/ > /fam/. But morphology and analogy introduced new
nasal consonants after nasal vowels (ennui /An/).

Was the dropping of nasal consonants triggered by the phonemic status of the
nasal vowels? Or is it this dropping that gave nasal vowels their phonemic
status? We'd probably have to consider contexts: at the end of words, there was
a general tendency to drop consonants, so in this context nasal vowels would
become phonemes earlier. (Note that Generative Phonology has no problem here:
it posits non-nasal vowels everywhere, and derives nasal vowels from them,
confusing diachronic rules and synchronic ones.)

The complementary distribution still holds in some cases: nasal vowels cannot
be found before vowels (neutralization).

Some distance limit (measured in terms of system coherence) must be taken into
account in such cases. When the allophonic relationship trespasses this limit,
a split may take place, as in the French case. In other cases the conditioned
realisation generalises to all values, yielding what appears afterwards as a
unconditioned change (as unconditioned palatalization of Arabic g): the integrity
of the phoneme is preserved.

4. A phoneme cannot be defined only by its oppositions.

In order to explain system dynamics, Martinet introduced the "case vide" concept.
He didnt insist on it as a new concept, partly (so I think) because its relation
with his Prague School conception of phonemic systems is not clear. It goes like
that: (a) suppose a phonemic system with a hole phonetically justified (eg
Indo-european had no /b/; within the laryngalist theory, this was explained by
the difficulty of glottalizing a labial consonnant /p'/.) (b) relevant features
change so the hole has no more phonetic justification. Then (c) this hole will
play the role of an attractor relatively to the phones that happen to be near,
and may induce a change in the phonological position of some phoneme (with or
without splitting it), which will fullfill the hole.

A phoneme being defined in the classic view by its opposition with the other
(existing) phonemes, there isnt any place for such notions as holes or nearness
in phonemic systems. It seems to me phonemes should be defined by their position
(given by their phonetic realizations), not only by their oppositions.


PS. Jakobson is a member of the Prague School, not a disciple of Sapir, even if
the work of Sapir and the Prague School had much in common, also because they all
had the same neogrammarian background (not only composed of fools: quite
intelligent ideas about all this had been expressed by for instance Verner, but
also Brugmann and the young Saussure). I admire the capacity of some to erase
Jakobson's European past.

---------------------------------------------------------------------------
LINGUIST List: Vol-11-155