148 Malcah Yaeger-Dror, Tania Granadillo, Shoji Takano, and Lauren Hall-Lew
analyzed from speakers who were not both from the same region and only
one in which speaker-sex differed.
The columns on the right of Table 5.1 show the number of speakers in
each cell.
3.1.3 Individual speaker variables
For the Varbrul Analysis, each speaker’s unique code categorized him/her by
sex (MFG),^9 age (by decade) and dialect/region (as speci¿ ed before). Except
for two men from the deep south, all speakers were middle class; as already
noted, the canvassing strategy elicited calls from mostly computer-literate
speakers, many with graduate degrees, and most telephone dyads were lim-
ited to those with identical demographics.
Except for the cross-generational calls discussed earlier, all dyads were
symmetrical. To the analysts, even the exceptions appeared to be quite “sol-
idary” and “reciprocal” (Brown and Gillman 1960).
As Tables 5.1 and 5.2 show, each speaker was also coded for situation,
with the newscasters (N) isolated from casual conversationalists (CF). The
number of NEG tokens analyzed and discussed here is found on Table 5.2,
along with the number of tokens from situations which are alluded to in pass-
ing.^10 Note that there are fewer tokens for the Japanese CF corpus than for the
others, not because the speakers use fewer of NEG tokens, but because fewer
phone calls have been analyzed; the average number of tokens per speaker is
not surprisingly low.
Table 5.2 Number of Tokens for Each Situation
SITUATION SPA N I SH JAPANESE ENGLISH
News 100 161 100
Debate — 287 530
SWB — — 505
CF 450 299 1626
Mean NEG tokens / CF speaker 22 37 343.2 Coding the dependent variable
Acoustic measurements of fundamental frequency, amplitude, and duration
were used to determine the prosodic prominence of NEG tokens in each of the