294 Thomas C. Purnell
to non-Wisconsin controls.^7 Of interest is the occurrence of apparent post-vo-
calic obstruent devoicing (e.g., /d/ > [t]) among the Wisconsin English speak-
ers. The Wisconsin English speakers were divided into four groups based on
their birth date:
- group 1: 1866 to 1892 (N=5)
- group 2: 1899 to 1918 (N=9)
- group 3: 1920 to 1939 (N=9)
- group 4: 1966 to 1986 (N=5)
The control group consisted of subjects not from Wisconsin who were approx-
imately as old as the third group. The sound ¿ les of speakers in groups 1 and
2, and one speaker from group 3 come from the Wisconsin English Language
Survey (WELS; Cassidy 1948) and the Dictionary of American Regional Eng-
lish (DARE; Cassidy 1985). Most of the sound ¿ les of the youngest test group
came from the University of Wisconsin Xray Microbeam database (Westbury
1994). The remainder of the test subject recordings is from speakers living in
Watertown, Wisconsin.
The studies reporting this data make two claims, which, in essence, follow
the general pattern of linguistic change outlined by Labov (1972: 39). First,
German plus other devoicing languages located in one geographic area (Polish,
Dutch, some dialects of Yiddish, etc.) fostered interlanguage substrate proper-
ties initially. Since then, the distinctive acoustic characteristics were reallocated
as a regional feature, distinct from general American English. The saliency of
this dialect pattern as reÀ ected in comedy routines at the national level contex-
tualizes this phenomenon within the purview of perceptual dialectology.^8
Second, the acoustic characteristics of vowel duration and percent glot-
tal pulsing appear to hold a signi¿ cant trading relation such that an observed
weak acoustic characteristic, in vowel duration for example, is noticed to be
strengthened by a complementary characteristic, percent glottal pulsing. This
observation contrasts with reports in the literature of a more universal charac-
teristic of post-vocalic obstruents as a change in ¿ rst formant and fundamen-
tal frequency (Kingston and Diehl 1994, 1995; Stevens and Blumstein 1981;
Stevens 2000). What Purnell et al. (2005b) report instead is that for each of
their test groups, the degree of the relation holding between duration and puls-
ing distinguished one group from another. For example, the data set with the
more recent birth dates unexpectedly reveals reduced glottal pulsing for voiced
as well as voiceless tokens. This is a dramatically different pattern than that
observed for speakers born 30 to 40 years earlier, but somewhat like the oldest
group which displayed a preference for glottal pulsing as the dominant acoustic