Chapter 4. Where and what is (t,d)? 99
question the case for treating (t,d) as a phonologically categorical^4 variable rule
within any framework needs to be made anew, since there remain no obvious
grounds for treating it in this way.^5 Moreover, the phonetic issues highlighted in
that paper suggest that there are good grounds for treating (t,d) as a function of
common Connected Speech Processes (CSPs) observed by many phoneticians in
English rather than a particular variable rule restricted to these coronal clusters.
The present paper will attempt to make the case for the CSP view of (t,d)
through a qualitative re-examination of data from some of the 38 speakers an-
alysed by Tagliamonte & Temple (2005), together with comparable data from the
same corpus containing other underlying coda consonant clusters and singleton
consonants. The data are all taken from audio recordings of sociolinguistic inter-
views collected for the York Corpus of British English under the direction of Sali
Tagliamonte and described in Tagliamonte (1998).^6 As Stuart-Smith et al., this
volume, demonstrate in their analysis of the complex social indexicality of the
detailed phonetics of rhotics, even cutting edge articulatory techniques cannot
in isolation give us a full picture of sociophonetic variability and need to be tri-
angulated with auditory and acoustic analyses, which are themselves imperfect
representations. Articulatory data are not available for the York recordings, so the
analysis in this paper will draw on acoustic and auditory observations, illustrated
by detailed phonetic transcriptions and a small sample of illustrative spectro-
grams; however, since the issues raised also crucially concern articulations which
are not necessarily audible or observable from the acoustic signal, reference will be
made throughout to the literature reporting relevant articulatory studies.
- There is a mismatch between the use of the word ‘categorical’ by variationists on the one
hand and general phonologists on the other: the former oppose ‘categorical’ rules, which always
apply (as in cases of regular allophony) to ‘variable rules’, which apply probabilistically (e.g.
coronal stop deletion is more likely to occur before consonants than vowels); the latter differ-
entiate between ‘categorical’ processes (e.g. the ‘replacement’ of a voiced stop by a voiceless one
under assimilation) and ‘gradient’ ones (e.g. the partial devoicing to various degrees of a voiced
stop under the same conditions). Both dichotomies apply to the discussion of (t,d) but the term
‘categorical’ is used here to mean non-gradient, since all analyses of the variable in question
agree that it is probabilistic. - Some scholars (e.g. Bermúdez-Otero 2010a; Myers 1996) argue for a dual view of (t,d) as
both a categorical and a gradient rule, as explicitly allowed for in Kiparsky’s (1985) view of LP.
The positive case for the categorical rule still needs to be made under this view. - The data collection was funded by a research grant (#R000238287) from the Economic and
Social Research Council for the United Kingdom. Digitisation of a subset of the data for the
present paper was funded by the John Fell fund of the University of Oxford. I am grateful to
Damien Mooney for his efficient assistance with the digitisation.