Chapter 4. Where and what is (t,d)? 131
To my knowledge no Firthian analyses yet exist of related English data.
However, working from a very different perspective, Steriade (2000) explores
critically the role of contrastivity in the categorisation of intervocalic flapping in
American English (a phonetic phenomenon by this criterion). Her driving agenda
is that, “the distinction between phonetic and phonological features is not con-
ducive to progress and cannot be coherently enforced. It is unproductive because
in order to understand phonological patterns one must be able to refer to the
details of their physical implementation, in perception and production” (Steriade
2000: 314). Tucker & Warner explore the contrast between this view and the alter-
native strict separation of phonology and phonetics in the light of their analyses
of the devoicing of nasals in Romanian. Having shown that the devoicing “derives
from both phonetic and phonological causes” they point out that this does not
necessarily entail the existence of two sharply delineated systems; it may simply be
that, “all sound patterns fall somewhere on each of several dimensions that make
up what we attempt to separate into phonetics and phonology” (Tucker & Warner
2010: 319). They argue that the answer to this is neither strict separation nor total
integration but the classification of sound patterns on several, mostly continu-
ous dimensions, “which all together make the phenomenon relatively phonetic
or phonological” (Tucker & Warner 2010: 320). This approach would seem very
promising for the analysis of word-final stops since it would obviate the need
for a sharp dividing line between cognitively and physically constrained phonetic
effects. We have seen the evidence for both here, and yet it is difficult to separate
the two: as Kühnert & Hoole (2004) show, they interact at a highly detailed level,
at least in assimilation, and their surface manifestations are often the same, and
this seems also to hold for (t,d).
One aspect of the variable behaviour of final stops that these models do not
cover, however, is the cooccurrence of CSPs. If it is the case that the pertinent
dimension of sociophonetic variation is not the lenition/assimilation etc. of par-
ticular word-final segments, but the manipulation of phonetic explicitness across
longer stretches of speech, a segmental based model would fail to capture the facts.
Simpson demonstrates in this volume and elsewhere how restricting the analysis
of variability to a single segment whose phonetics are governed by the immedi-
ate segmental context can obscure significant generalisations. In his analysis of
glottals in Suffolk English (Simpson 1992), he examines the insight of Trudgill
(1974) and Lodge (1984) that there are cooccurrence restrictions on glottalisation
in some East Anglian varieties of English, and demonstrates that even one of the
authors who drew attention to these misses some examples of the phenomenon
because the analysis is couched in terms of derivational reduction rules which
apply to individual segments. Simpson’s solution is inspired by the Firthian notion
of “prosody”, a phonological construct which has phonetic exponents across a