Phonetic Detail in the Perception of Ethnic Varieties of US English 295characteristic of post-vocalic voicing. The middle two groups were observed
having greater balance between vowel duration and pulsing. This pattern of
a younger group adopting the forms of an earlier group reÀ ects the, now old,
observation by Labov that “[g]roup A is adopted as a reference group by group
B, and the feature is adopted and exaggerated.. .” (1972: 39). In the context of
the present study, group A is represented by older Wisconsin English groups
(groups 1 and 2), and group B by the youngest group (group 4). In short, this
pattern of younger speakers exaggerating a form of an earlier reÀ ex of the voic-
ing distinction should not be unexpected by sociolinguists.
Within the context of the acoustic-perceptual interface identi¿ ed on the
Hoenigswald-Preston observation (Figure 13.1), the trading relation between
duration and pulsing found in the Purnell et al. data raises the following over-
lapping questions:
Are the signi¿ cant articulatory or acoustic characteristics also perceptually
signi¿ cant cues? That is, would we expect a change in fundamental fre-
quency to be the strongest perceptual cue as opposed to vowel duration even
though it is not a signi¿ cant measure across test subject groups? Another
question is, are there latent factors underlying the articulatory or acoustic
characteristics, e.g., a “low frequency” hypothesis, that listeners might be
paying attention to rather than to the speci¿ c values of the characteristics?
Within each speaker group, if latent factors are present, are these underlying
factors identi¿ able as perceptual cues to the voicing distinction? Or, is the
acoustic-perception link in Figure 13.1(d) transparent?The present chapter answers these questions by testing whether or not
there is variation across groups with respect to the variables selected or the
weight of the variables contributing to the model.^9 One test hypothesis is that
there is variation. Given the reported acoustic variation already across the test
subject groups, it is hypothesized that the variation reÀ ects a split between the
Wisconsin English speakers and the control subjects. Another test hypothesis
is that acoustic signi¿ cance does not necessarily imply perceptual signi¿ -
cance nor does perceptual signi¿ cance imply acoustic signi¿ cance.
3.2 Methodology
Seven main acoustic measures were collected as potential predictors in this
model: vowel duration; the duration of the consonantal gap (for stops) or frica-
tion (for fricatives); the duration of glottal pulsing in the gap; the percentage
of glottal pulsing in the gap (% glottal pulsing); the ratio of the vowel duration