108 Rosalind A. M. Temple
Word-internally, deletion is probably lexicalised in most cases, occurring nearly
categorically in words like grandmother, grandfather and Christmas, but it also
occurs in less frequent compounds, such as landmarks (28) and second-hand (29):(28) one of the local landmarks [lɔːklˠ̩ːanmɑːks] was this brickyard chimney
(29) they bought things in second-hand shops [sɛʔ͡n̩ˈhant͡ʔʃɒps]As with (t,d), most, though not all, of these examples are pre-consonantal and so
an unsurprising outcome of “phonetically natural” CSPs. Indeed Nolan (1992)
gives a hypothetical example of the total lenition of word-internal /d/ in the word
hundred (“[ˈhʌndɹəd] (? → [ˈhʌnd̞ɹəd]) → [ˈhʌn̞ɹəd]”; Nolan 1992: 23), which he
classifies under “Target Undershoot” at the phonetically natural end of his scale.
Is there, then, any independent evidence that word-final (t,d) clusters are quanti-
tatively or qualitatively different from Examples (24) to (29), which would justify
their treatment as a special variable rule?2.2.2 Sociolinguistic variation in lenition
One type of evidence for the special treatment of (t,d) would be sociolinguistic
effects not applying to other cases of word-final lenition/deletion. Such effects
have been found for AAVE and some southern US dialects, but not for other
varieties of English. Gimson’s classic text on English pronunciation (as re-edited
by Cruttenden) is peppered with what are essentially sociolinguistic judgements,
such as the comment that, “the elision of one of a boundary cluster of only two
consonants sometimes occurs in casual speech but is usually characterised as sub-
standard, e.g. He went away /hɪ wɛn ə ̀weɪ/ (...) Let me come in /lemɪ kʌm ̀ɪn/”
(Cruttenden 2008: 302). Interestingly, where word-final clusters are concerned,
these contexts correspond exactly to the prevocalic cluster reduction noted as a
qualitative and quantitative sociolinguistic difference between African American
and other varieties of North American English. By contrast, it is striking that
(a) word-final clusters are grouped in the above quotation with cross-boundary
sequences, suggesting no special status, and (b) no such evaluative judgements
are proffered in comments on the deletion of cluster alveolars before consonants,
which Gimson/Cruttenden seem to treat as straightforward, socially unmarked
CSPs, entirely to be expected in RP:
[...] sounds may be elided in fast colloquial speech, especially at or in the vicinity
of word boundaries (...) In addition to the loss of /h/ in pronominal weak forms
and consonantal elisions typical of weak forms, the alveolar plosives are apt to be
elided. Such elision appears to take place most readily when /t/ or /d/ is the middle
one of three consonants. (Cruttenden 2008: 303)