58 Jen Hay and Margaret Maclagan
3.4 Likelihood of /r/-insertion after /au/
Finally, having established that both the frequency and the degree of intrusive
/r/ act as sociolinguistic variables, we then turned to a closer investigation of the
plough paradigm. This vowel is a relatively recent addition to potential environ-
ments for intrusive /r/, and it is not a vowel that facilitates intrusive /r/ in other
dialects. Theoretically, the high /u/ offglide of the vowel should prevent intrusive
/r/ from appearing. The vowel /au/, however, has undergone relatively extreme
phonetic change in New Zealand English. Where many New Zealand speak-
ers born in the late 19th century still produced /au/ as a closing diphthong with
the second element approximating their /ɬ/ vowel (Gordon et al. 2004), mod-
ern NZE speakers usually have a very central second element, approximating
schwa (Maclagan 1982). In addition, the ¿ rst element has raised from [a] or [æ]
to closer to [͑], so that for many speakers /au/ is now realized as [̸͑]. One effect
of this change is to lessen the difference between the two targets so that the vowel
is relatively less diphthongal and relatively more monophthongal in production.
We therefore speculated that the likelihood of intrusive /r/ appearing may
be partially predicted by the phonetic quality of the /au/ vowel produced by
that speaker. This would be intriguing—providing an interesting insight into
the interface between phonetics and phonology.
Assessing this link is not straightforward, as the presence of /r/ will itself
affect the formants of the preceding vowels, so directly comparing the target
words of /r/-inserters and non-/r/-inserters is not practical. Any difference
found during the preceding vowel could be due to the presence of the /r/ rather
than to any inherent difference in the quality of the vowel.
Fortuitously, the test sentences included four instances of the word plough
in non-intrusive environments (see Appendix 2.1). Using Praat, we took mea-
surements of F1 and F2 at two target points during the diphthong. The ¿ rst
measurement was taken at the ¿ rst steady state portion of the vowel or where
the F2 was highest, and the second at the second steady state or where F2 was
lowest. The second target was measured before F2 started to lower for the fol-
lowing consonant, which was /m/ in all cases. For two tokens, the vowel was
in fact, a triphthong, /̸ࡱ /, and we took measurements at the ¿ rst and the last
target of the vowel.
In attempting to relate these raw values to the quality of the vowel, the
normalization problem rears its ugly head again. Raw formant values will, of
course, be affected by the length of the vocal tract. Rather than compare raw
values, then, we calculated the Euclidean distance, in F1/F2 space, between the
two targets of each vowel. This value would be zero for an entirely monoph-
thongal vowel, and it would be high for a vowel which contained a high degree
of internal movement. Figures 2.11 and 2.12 give examples of tokens with