42 Jen Hay and Margaret Maclagan
may be associated with a certain amount of stigmatisation (see, e.g., Lewis
1975; 1977, Pring 1976, Fox 1978; Brown 1988). Linking /r/ and intrusive /r/
also behave differently with respect to the nature of the following boundary.
Word-internally, linking /r/ “almost invariably occurs,” while there are “occa-
sional” instances of word-internal intrusive /r/ (Wells 1982: 224). Gimson and
Cruttenden (1994: §12.4.7) claim that in RP “the insertion of /r/ is obligatory
before a suf¿ x beginning with a vowel, where the /r/ is historically justi¿ ed,”
whereas “the insertion of intrusive /r/ before a suf¿ x is strongly resisted.”
Hay and Sudbury (2005) investigate how /r/-sandhi emerged in New Zea-
land English in the late 19th century, concurrent with the decline of rhoticity.
There was never a period of complete /r/-lessness in New Zealand English—
linking and intrusive /r/ arose before rhoticity was gone. Linking /r/ across
morpheme boundaries remained categorically present. Across word-boundar-
ies, it declined with the loss of rhoticity, but at a much slower rate than non-
prevocalic /r/. Intrusive /r/ across word boundaries arose gradually. Intrusive
/r/ across morpheme boundaries was a later innovation, which seems to have
largely postdated the disappearance of rhoticity in New Zealand. This historical
trajectory makes sense of the observations that intrusive /r/ is less frequent than
linking /r/ (it’s more recent), and that across morpheme boundaries linking /r/ is
categorically present (it always has been), and that intrusive /r/ is resisted.
The identity of the preceding vowel has also been reported to play some
role in conditioning the likelihood that /r/ will arise. Jones (1964) and Gim-
son (1980) claim that intrusive /r/ is more likely following schwa than other
vowels. It is also claimed to be less stigmatized in this context (Crystal 1984:
43; Brown 1988: 149). Wells claims intrusive /r/ after /ɬ/ is more stigmatized
than after other vowels because it was a later innovation (1982: 225). Hay and
Sudbury’s (2005) study of 19th-century New Zealand English shows the low-
est rates of intrusive /r/ after /ɬ/, supporting Wells’ claim.
In addition to linguistic conditioners, there may also be social condition-
ers in some dialects. The most comprehensive contemporary corpus-oriented
study was conducted by Foulkes (1997a, b), who studied /r/-sandhi in a strati-
¿ ed sample of speakers from both Newcastle upon Tyne and Derby. While
social categories appeared to play no role in Derby, there were clear class
and age differences in Newcastle, suggesting that /r/-sandhi may be socially
conditioned in some dialects.
No large-scale systematic study has investigated rates of /r/-intrusion
following different vowels. And while it has been claimed that /r/-sandhi is
more common across word boundaries than morpheme boundaries, no one
has investigated whether different types of suf¿ xes may give rise to different
rates of /r/-intrusion. Empirical data on the possible conditioners of intrusive