LINGUIST List 4.356

Tue 11 May 1993

Disc: Velar palatalization

Editor for this issue: <>

Directory

1. H.Stephen Straight, Velar>palatal: phonemic change,yes; allophonic alternation,no
2. , 4.330 Velar palatalization

In his 4.330 response to Andy Spencer's questions regarding the apparent
contradiction between the fact that k > ch is an attested (and frequent?)
historical change and the fact that k > ch is unattested (or rare?) as an
allophonic alternation, John Coleman observes that k and ch (or s) mark
"the start- and end-points of a CHAIN of natural phonetic/phonological
changes". Coleman walks us through a series of steps, which appear to be
summarizable as (1) palatal assimilation of k before i or j followed by
(2) strengthening of aspiration and (3) reanalysis of this highly
aspirated palatalized velar as a voiceless palatal affricate (or
fricative).

Step 1 is an assimilation that is well attested in many allophonic
alternations (in English and other languages). Step 2 is not discussed
but might result from a general fortisization driven by forces external
to the specific k>ch>s development.

Step 3 is of course the crux of the development . Coleman observes that
"The distinction between an [i]-coloured aspiration portion [of a
palatalized velar plosive] and a voiceless palatal fricative [C] (IPA
c-cedilla) is largely a matter of duration and air pressure. Otherwise,
they are acoustically practically identical. So it is phonetically
natural for an aspirated [k,] or [c] allophone of /k/ to come to be
perceived and pronounced as a voiceless palatal affricate [cC]."

The critical juncture, therefore, appears to be this shift of an
allophone of k from being merely auditorily indistinguishable from a
voiceless palatal affricate (vpa) to being articulatorily distinguishable
as a vpa. And the evidence suggests that this particular shift in
articulatory target, unlike others that appear in allophonic
alternations, necessarily results in (or is motivated by) the emergence
of a new vpa phoneme, employable in the underlying articulatory
representations of lexical items in just the way the phoneme k is
employed.

We have now rephrased the original question, "Why no k/ch allophonic
alternation?", as the question "Why are Velar and Palatal incompatible
allophonic articulatory targets, even though they are occasionally
associated with identical (and attested) allophonic auditory cues?" What
we need to do now, I suppose, is look for other examples of auditorily
similar allophones to see how they differ in their susceptibility or
resistance to merely allophonic articulatory redefinition.

Study of the differences between auditory and articulatory phonology may,
BTW, eventually result in raising a more fundamental question for
phonological theory: "Do the auditory representations of lexical items
differ in content from their articulatory representations?" For example,
are the underlying articulatory representations of _tab_ and _bat_
identical except for the order in which articulatory targets appear,
while their auditory representations differ in numerous ways because of
differences in voice onset timing, vowel duration, formant transition
slopes, and other such acoustic cues for their recognition? In other
words, syllable-initial and syllable-final occurrences of consonants may
have unitary articulatory representations but be represented as separate
auditory units.

Confirmation that these two kinds of representation must for other
reasons be treated separately could help explain an otherwise
inadequately explained fact: Conduction aphasics retain their ability to
understand what is said to them but cannot reproduce it. In a bifurcated
phonological theory we could claim that such aphasics have lost the
connections between auditory and articulatory representations that would
have been provided by their damaged arcuate fasciculus (the band of nerve
fibers that links secondary auditory cortex [Wernicke's area] to
secondary motor cortex [Broca's area].)

Query: Do conduction aphasics also exhibit the difficulties in
articulatory compensation that would be predicted by a loss of auditory
representations available for monitoring as a result of
articulatory-to-auditory (Broca's-to-Wernicke's) connections? (I'm not a
neurolinguist even if I do sometimes try to sound like one.)

Whew, four screensful! Sorry. Thanks for your patience.

H. Stephen Straight
Anthropology and Linguistics, Binghamton University (SUNY)
 E-mail: <sstraighbingvaxa.bitnet>
 <sstraighbingvaxa.cc.binghamton.edu>
 Voice: 607-777-2824; Fax: 607-777-2477

John Coleman says in reply to Andy Spencer's queries:

>> since each step along the way is a phonologically
>> natural assimilation, phonological theory cannot help but
>> characterize K > CH as a natural assimilation.

This strikes as a bizarre conclusion. Phonological theory should
in fact say that K > CH is NOT a natural synchronic assimilation,
while the single steps (e.g. K > K') are. Otherwise, almost EVERY
chain of sound changes will have to be treated as a natural synchronic
process.