6 Dennis R. Preston and Nancy Niedzielski
as frequently, although, as documented in Chapter 7, research evidence for
greater Kanto devoicing is slim. In this study, Yonezawa Morris asks respon-
dents from Tokyo and Osaka to listen to voiced and devoiced tokens of lexical
items and phrases as well as items that have distinctive and shared pitch accent
patterns and identify the city of the speaker. She shows that respondents from
both areas are not only able to identify devoicing but also attribute voiced and
devoiced tokens to the areas indicated by the stereotypical beliefs outlined
above. This chapter also hints at the intriguing possibility that many tokens
of voiced vowels were identi¿ ed as devoiced (and attributed to Tokyo) on the
basis of their occurring between voiceless consonants that have a particularly
high promoting effect on devoicing. If this is so, it is an interesting linguistic
parallel to the social redirection of acoustic facts outlined in Chapter 11.
In Chapter 8, Clopper investigates how well (or how poorly) nonlinguists
identify the regional provenience of US dialect samples. She provides speech
samples of group classi¿ cation (with no labels or regions suggested by the
experimenter) and a second test in which paired samples were used, again
with no suggested identi¿ cation, and the respondents rated how much one
sample was like the other. Although gross differences were fairly well recog-
nized (Northern and Southern US varieties were distinguished, for example),
more subtle aspects of dialect differences were not detected, although both
an enhanced ability to recognize one’s own variety and an experience factor
(based on geographical mobility) of familiarity with other varieties were sig-
ni¿ cant. Clopper concludes by suggesting that an exemplar model of phonol-
ogy is best equipped to account for her ¿ ndings.
Chapter 9 explicitly tests the ability of participants in a vowel shift (the
Northern Cities Chain Shift—NCCS) and nonparticipants to identify an
intended vowel along a continuum of signals from an unshifted to shifted posi-
tion. Plichta and Rakerd recruited listeners from urban southeastern Michi-
gan (shifted) and the Upper Peninsula of Michigan (nonshifted) and presented
them with the second step of the NCCS—the fronting of /ľ/. The seven-step
continuum tested the location of F2 that caused a respondent to switch his or
her identi¿ cation of the vowel from that of “sock” or “hot” to that of “sack” or
“hat.” One would expect the shifted NCCS speakers to tolerate a much higher
(fronter) F2 and still report “sock,” and the nonshifted respondents to identify
a much lower (backer) F2 boundary for the switchover to a “sack” interpreta-
tion. The results were somewhat more complicated; when the test words were
presented in isolation, the two sets of respondents exhibited a crossover point
at the same step (mean scores of 4.7 and 4.8). When the same seven-step items
were embedded in carrier phrases which contained other examples of NCCS-
shifted or nonshifted vowels, the nonshifted listeners were not inÀ uenced by