64 Jen Hay and Margaret Maclagan
present across a word-internal morpheme boundary, and so should resemble
onset /r/s. Indeed Hay and Gibson (2005) show that these behave alike in per-
ception. Linking /r/ across word boundaries is variable, but occurs at higher
rates than intrusive /r/. This exemplar account would therefore predict that
intrusive /r/ should contain a weaker constriction than the other /r/s.
Further work will be needed to reveal whether an exemplar account of
speech production is responsible for the relationship between frequency of /r/
use and degree of constriction. Certainly, such an account would predict that
there should be other variables where there is a link between the frequency
of occurrence of the variable and its phonetic quality. For example /hw/ may
involve reduced aspiration in environments facilitating merger with /w/ in
dialects which contain this as a variable. Consonants in environments facili-
tating consonant cluster reduction may be reduced. In the early loss of rhotic-
ity in NZE, /r/ may have undergone a gradual weakening. As Thomas (2002)
points out, acoustic work on consonants has been sadly lacking in the history
of sociophonetics. We are being highly speculative here, but it is certainly
within the bounds of possibility that the sociophonetics of consonants may be
much more gradient and intriguing than has hitherto been thought.
The relationship between the phonetic realization of /au/ and the like-
lihood of /r/-insertion is also intriguing. Under a traditional phonological
account of the alternation, one would have to argue that the phonetics of
/au/ has now changed so much that it is being reclassi¿ ed, for some speakers,
as [-high] and so eligible for the /r/-insertion rule (or not eligible for the dele-
tion rule, depending on one’s analysis). We think it is more likely that a more
gradual analogical process is taking place. As the phonetics of /au/ changes,
it gradually starts to resemble phonetically other vowels which are associ-
ated with intrusive /r/—in particular, the offglide begins to resemble schwa,
which does participate in the alternation. It may be particularly relevant that
there is a related triphthong (as in À our, sour), which always has an associated
or thographic r and so attracts linking /r/. As the second target of /au/ begins to
more closely resemble schwa, the distinction between the second two targets
of the triphthong will be considerably lessened, increasing the overall resem-
blence between, e.g., À our and plough. Indeed, some of our students have
even reported informally to us that they think these words rhyme. Thus just as
intrusive /r/ arose after /a/ in the late 19th century (e.g., ma/r/-and by analogy
to ca/r/-and), the same type of analogical process may be leading intrusive
/r/ to arise after /au/ in contemporary NZE (e.g., plough/r/-and by analogy to
À ou/r/-and). In order to more directly test this interpretation, we will need to
record productions of the relevant “triphthongs” from speakers. It may also be
useful to elicit rhyming judgements.