110 Rosalind A. M. Temple
in Temple (ms). They are far less problematic when (t,d) is analysed in terms of
CSPs, so we once again turn to the qualitative data to confirm whether this non-
phonological analysis can be justified.
I present here just a small sample of typical (t,d) tokens with different combi-
nations of preceding and following consonants, where the variable rule analysis
would state that deletion has applied, beginning with cases where the preced-
ing and following consonants are pronounced in their unlenited citation forms.
Examples (30)–(33) are typical of target undershoot in continuous speech:(30) oh I’d booked my [bʊk ̚mə̃] ticket, yes
(31) but we still kept corresponding [kʰɛp ̚kʰɔɹɪspɔndɪŋ] all the time
(32) so of course I left school [lɛfskuˑlˠ] at fourteen
(33) ... whether I spent the first few [fɜːsfjuʷ] months of my lifeIt is not necessary to assume here that the speaker has deleted the (t,d) consonant
in the phonology and therefore produces no alveolar closing gesture; rather, it is
perfectly plausible that these are cases where the hypothetical target for the /t/ orTable 1. Results of GoldVarb analysis of the effects of following and preceding
phonological context on deletion of /t,d/.
Factor weight % deletion Total N
following consonant
nasal .918 70 69
stop .890 66 93
fricative .887 62 101
glide .690 38 106
/r/ .605 28 29
/l/ .496 25 24
/h/ .354 11 62
vowel .291 8.3 507
pause .200 5.5 127
range [72]
preceding consonant
/s/ .690 41 303
/ʃ/ .565 31 64
nasal .497 21 329
stop .382 16 169
liquid .374 21 126
non-sibilant fricative .298 12 127
[39]
total 1118