132 Rosalind A. M. Temple
given stretch (or “piece”) of speech. This would seem a promising avenue for
exploration of the variability of word-final stops, although it should be noted that
the Firthian approach is declarative, with strict separation between phonology and
phonetics, and is therefore on the face of it not compatible with Steriade or Tucker
& Walker’s advocation of total or partial integration of the two.- Conclusions
This paper has, I hope, made a case for an answer to the first set of issues explored
in the discussion, relating to the “what” of the title. It is clear from the data exam-
ined that the behaviour of word-final /t,d/ in clusters is not qualitatively different
from that of other word-final consonants, either in their segment-specific physical
manifestations or in their interactions with common Connected Speech Processes
in this variety of English. This does not conclusively prove that a phonological analy-
sis is wrong: CSPs could be part of the post-lexical phonetic implementation pro-
cesses which interacts with the output of a variable phonological rule, as suggested
by Bermúdez-Otero (2010a, b). However, since the CSP account seems perfectly
adequate in accounting for the observed behaviour of word-final coronal clusters,
it would seem that there is no need to invoke such a rule in the absence of posi-
tive evidence for an unambiguously phonological effect. It appears, then, that what
(t,d) is is simply one manifestation of the general phenomenon that, in Browman &
Goldstein’s words, “in casual speech (...) segments are routinely elided, inserted and
substituted for one another” (Browman & Goldstein 1990: 359).
The other set of issues, that is where this situates (t,d) and associated phe-
nomena in the grammar, is less easy to resolve and depends on the place of this
and other CSPs which, “can neither be modeled adequately at a symbolic, phono-
logical level, nor left to be accounted for by the mechanics of the speech mecha-
nism” (Nolan 1992: 280). But some well motivated model is needed in order to
provide a sound basis for any sociophonetic/sociolinguistic analysis. The explora-
tion in §3.2 of potential different frameworks for analysis was necessarily brief and
far from conclusive, although it is clear that there are grounds for concluding that
some are not satisfactory. With respect to more promising frameworks, the data
reviewed here are insufficient to draw definitive conclusions about which approach
to the phonology-phonetics interface best fits with the empirical observations of
word-final stops. More articulatory data would be needed to implement a met-
ric of gradience/categoricity,^33 for example, whereas a Firthian-inspired approach- Although even articulatory data would not be able to disambiguate all tokens, for example
those with preceding /n/, which involve alveolar closure.