50 Jen Hay and Margaret Maclagan
omitted the second syllable of the base, leading to productions such as Oprese
and Oprish. Because of its low overall rate of intrusive /r/, Oprah does not
show strong effects of social class. Interestingly, sofa also does not show
strong effects of social class. This would seem to reinforce the observations
from other dialects that intrusive /r/ may be somewhat less stigmatized fol-
lowing schwa than following stressed vowels (Crystal 1984: 43; Brown 1988:
149). Claw, and especially bra, ma and plough, show a stronger effect of social
class, with higher social classes using dramatically less intr usive /r/. This is not
surprising for plough—as this is a relatively new, incoming variant and is also
(as we will argue) associated with innovativeness in the phonetics of the /au/
vowel. It is perhaps also not so surprising for claw, given that intrusive /r/ with
/ɬ/ is rare in some dialects, and seems to have evolved later in New Zealand
English than with some other vowels. It is interesting to note, however, that the
overall rate of /r/ after /ɬ/ is actually relatively high in this data-set—which is
an interesting difference from work on early speakers of New Zealand English,
where rate of /r/ after /ɬ/ (Varbrul weight .290) was considerably lower than
the rate of /r/ after either /a/ (Varbrul weight .840) or /ۑ/ (Varbrul weight .502)
(Hay and Sudbury 2005: 813). It is dif¿ cult to know, however, whether the rela-
tive inÀ uence of the vowels has changed, or whether there is something about
this particular base word that makes it unusually facilitative of /r/.
3.2 Degree of /r/-ness of intrusive /r/s
Having established that there are social and linguistic factors affecting the
likelihood of intrusive /r/, we hypothesized that there may also be more subtle
differences in the articulation of the /r/ when it is produced. Namely, the same
factors that lead an /r/ to be more likely to be produced, may also lead it to be
more /r/-like when it is produced. Comparing Figures 2.1 and 2.2 reveals that
there is some variation in how low F3 is. In the utterance in Figure 2.1, F1,
F2 and F3 look approximately equidistant, whereas in Figure 2.2, F1 and F2
appear closer together. We wondered whether such variation was random, or
whether there might be some structured variation to uncover here.
There are a variety of differences in articulation which could lead to a
decreased F3—including increased retroÀ exion, increased “bunching” or
constriction in the palatal region, and increased liprounding (Fant 1968, Lade-
foged and Maddieson 1996, Guenther et al. 1999). We do not want to speculate
about the most likely articulatory correlates of a lower F3 in a NZE context,
but will simply refer to /r/s with lower F3s as having an increased constriction
(relative to /r/s with higher F3s, all else being equal), without being speci¿ c