The Frequency and Degree of “intrusive /r/” in New Zealand English 49The effect of the identity of the af¿ x is shown in Figure 2.4. As reported
in Hay and Warren (2002), the rate of /r/ insertion differs across af¿ xes, with
af¿ xes marking more word-like boundaries more likely to attract an /r/. Hay
and Warren explain: “-ing is the af¿ x with the strongest (most word-like)
boundary and attracts the highest rate of intrusive /r/. -ish is also a relatively
productive, relatively separable af¿ x. This contrasts with af¿ xes like -ese and
- ify, which are less frequent, less separable, and more likely to occur with
bound roots (see, e.g., Hay and Baayen 2002).” They demonstrate that the rate
of /r/ use with these af¿ xes correlates with various measures of their pars-
ability. The more word-like the af¿ x boundary is, the more likely intrusive /r/
is to occur.
Finally, Figure 2.5 shows the predicted effect of social class, which inter-
acts with the identity of the base. Con¿ dence intervals have been left off the
graph to reduce the visual clutter. Note that the y-axis spans a much greater
range than the previous ¿ gures—reÀ ecting the fact that the effects of base and
social class are of a greater magnitude than the effects of gender and af¿ x.
A number of things are apparent from the graph. First, intrusive /r/ is rela-
tively rare after Oprah and much more common after sofa and claw. The rela-
tively low rate of intrusive /r/ after Oprah reÀ ects the fact that many participants
Figure 2.5 Model predictions showing interaction between Social Class and base
type. A score of 3 represents a relatively higher social classes and 11 a
relatively lower social class.