Resyllabification

			William Labov
		(University of Pennsylvania)


	One of the most fruitful areas of English phonology has been the
study of the simplification of final consonant clusters in spontaneous
speech.. One of the most challenging problems in recent years is to relate
these findings to the development of phonological theory, for the mutual
illumination of both fields of study.
	One of the most consistent constraints on (t,d) deletion in English
and comparable processes in other languages is the effect of the following
segment. Studies from many areas of the English-speaking world have shown
that retention of the final consonant is favored most by following vowels,
then glides, liquids, fricatives and stops. It has long been recognized
that this sequence is ordered by decreasing sonority. Guy (1991), Reynolds
(1994) and others link this pattern current phonological theory by arguing
that consonant cluster simplification is the result of delinking of the
second member of a complex coda followed by stray erasure when that
delinked segment is not re-associated with a following syllable. Thus the
possibility (or probability) of resyllabification should be directly
correlated with retention of the final consonant in speech.
	Whether this mechanism can in fact explain constraints on (t,d)
deletion can be tested since many English consonants change their phonetic
character radically when they shift from coda to onset position, as in the
contrast of night rate and Nye trait. The analysis of a collection of 763
natural misunderstandings suggests that the number of cases where this type
of resyllabification takes place is minimal. A Varbrul analysis of one
speaker recorded for six hours was undertaken, yielding 767 tokens, and
each retained cluster was examined auditorily for the presence of phonetic
features characteristic of final or initial allophones. Some evidence of
resyllabification was found, particularly with following /y/ onsets as in
last year. As the resyllabification mechanism predicts, /t/ or /d/ is
retained more often before words with initial /r/ than initial /l/. But on
the whole, the frequencies of phonetically attested resyllabification are
far too low to account for the observed retentions, and some predictions
run counter to observations, particularly in the case of glides.
	In Kahn's original rules for the resyllabification mechanism
(1980), the proposal for re-syllabifying clusters across word boundaries
was limited to vowel-initial words, where the re-syllabified consonant was
ambisyllabic, and any conversion to onset-type allophones was rejected.
This constraint on resyllabification is retained in the cross-linguistic
view of Kenstowicz (1994). Although resyllabification does not account for
(t,d) deletion in English, the exploration of this possibility has led to a
number of unexpected findings which further illuminates our view of this
process.

References:
Guy, Gregory 1991. Contextual conditioning in variable lexical phonology.
Language Variation and Change 3:223-239.
Kahn, Daniel. Syllable-Based Generalizations in English Phonology. New
York: Garland Publishing, 1980.
Kenstowicz, Michael 1994. Phonology in Generative Grammar. Oxford: Blackwell.
Reynolds, William 1994. Variation and phonological theory. U. of
Pennsylvania dissertation