LINGUIST List 3.867

Fri 06 Nov 1992

Disc: Phonetics

Editor for this issue: <>

Directory

1. "Bruce E. Nevin", phoneticians
2. John S. Coleman, 3.839 Final Devoicing

The discussion of differences missed by ear and caught by instrumental
analysis reminded me of

Labov, William. 1974. On the Use of the Present to Explain the Past,
 Proc 11th Int. Congr. of Linguists, 1974. (Reprinted in a
 collection edited by Baldi and Werth.)

Labov (1974) presents evidence that the linguistic intuitions of native
speakers may not be reliable, even given the idealization of the pair
test or the availability of native speakers trained in phonetics and
(other aspects of) linguistics. Native speakers may not hear systematic
phonetic (subphonemic) differences that are apparent upon instrumental
analysis. And yet native speakers must unconsciously hear those same
phonetic differences and must control their audibility in their own
speech, or they would not persist in the systematic way that they do.

The particular data in question show that the unconscious behavior of
speakers of the English of Norwich, England, reflects a contrast between
/ay/ and /oy/ that native speakers and linguists alike had agreed had
merged and were indistinguishable.

This research relates to a merger two or three centuries ago that was
reversed or was "undone" in the 19th century. The usual explanation is
that this reversal of sound change was due to the influence of spelling,
borrowings from the hypercorrected dialect of literate speakers.
Labov's data suggest a different explanation, namely, that a distinction
that is no longer contrastive can nonetheless still hang around in the
language, perhaps to reemerge later. An analogy suggests itself to a
recessive phenotype emerging from the gene pool of a biological
population.

Even though this is a difference that makes no difference to native
speakers in distinguishing lexical items such as "line" and "loin", and
it makes no evident difference on the level of presentation of self (a
usual function of dialect differences), it must make a difference to
speakers of this dialect on some level in order so to be maintained.

This suggests limitations on what may be claimed for any scheme of
representation for the phonological contrasts of a language. The
question exactly which type a given sound change may be is less
important than that a range of alternatives be known and considered in
the work of analyzing and understanding a particular situation in
language change. If an alternative interpretation for a given
correspondence or other datum is not raised for consideration, its
plausibility can never be evaluated relative to other interpretations.
This, I believe, was Hoenigswald's intent in _Language Change and
Linguistic Reconstruction_.

	Bruce Nevin
	bnbbn.com

More on the incomplete neutralization controversy.

Two papers, Fourakis and Iverson (1984), and Jassem and Richter (1989),
have been cited as showing failure to replicate earlier incomplete
neutralization results in German and Polish respectively. Using more
natural and spontaneous test materials than earlier experiments,
F&I, and J&R performed two experiments each:

F&I 1. Measurements of duration of vowel duration and consonant closure
duration. The result: failure to replicate earlier experimental results
showing preservation of the voiced-voiceless distinction as a systematic
durational distinction.

F&I 2. In less natural speech (a reading task), the incomplete
neutralization effect was replicated.

F&I conclude that incomplete neutralization is related to how natural
the speech is. In more natural speech, neutralization is complete;
in less natural speech, neutralization may is incomplete. Note that
only their first experiment is the crucial one to this debate.

J&R 1. Using randomized test-data, native Polish speakers cannot
distinguish underlying voiced from voiceless categories, showing
that the neutralization must be complete.

J&R 2. Like F&I 1, there is no statistically significant durational
distinction between devoiced and underlyingly voiceless categories.

 -------------
Comment: The three critical experiments, F&I 1, J&R 1 and 2, suffer
from two major problems which cast doubt on the conclusion that they
do not support incomplete neutralization. F&I 1 and J&R 2 suffer from
the same problem, that by concentrating on duration measures alone,
they fail to demonstrate the possibility that the word-pairs in
question are distinguishable on the basis of other properties,
such as spectral or amplitude distinctions. Lisker (1986) identifies
at least 16 possible cues to the voiced/voiceless distinction, of which
11 are relevant to word-final position. 4 of these are non-durational.
It is well known that spectral and amplitude distinctions are as
important as, if not more easily perceivable than, durational distinctions.
Scott (1984) shows that non-durational cues are critical to the
resolution of the incompletely neutralized t/d contrast in American
English. Consequently, measurements of durational distinctions is
sufficient to establish that a putative neutralization is incomplete,
but is not sufficient to show that neutralization was complete!

The remaining critical experiment, J&R 1, employed randomized test
stimuli, rather than permitting subjects to make same-different
decisions on devoiced vs. voiceless stimuli. While randomization is
relevant during collection of the data, to avoid explicit contrastive
effects in the cues, it is not necessary in administering the test
stimuli. In fact, a phonetician wishing to check a pair of
words to determine whether a particular distinction is perceptible
or not is at liberty to explicitly focus on the two words, played
over and over again many times, before forming a judgement. The
experiment was thus much harder than necessary for the purposes
of validating or refuting the incomplete neutralization effect.

--- John Coleman