232 Bartáomiej Plichta and Brad Rakerd
4.4 Target words in citation-Form: The qualifying test
A condition of participation in the main experiment (referred to below as the
sociophonetic test) was that a listener had to be sensitive to the acoustical
variations in F2 that were represented across the stimulus set. Accordingly,
prior to participating in the main experiment, all listeners completed a quali-
fying test in which they were asked to categorize the vowel heard in each item
of the “hot”—“hat” word series, and each item of the “sock”—“sack” word
series, with the words presented in citation form.
Over the course of this test each word item was presented a total of
four times. The order of these presentations was randomized and different
on every test run. After a word was presented, the respondent’s task was to
decide whether the target vowel sounded more like /a/ (as in “hot” or “sock”)
or /æ/ (as in “hat” or “sack”). The choice was then reported by pressing the
appropriate button on the computer’s touchpad. The entire citation-form test
was completed at a single sitting, in about 30 minutes time.
4.5 Psychometric functions and cross-over points
Figure 9.6 shows the results of the qualifying test, which was passed by all
18 of respondents. The panel at the top shows /a/ and /æ/ psychometric func-
tions for the UP respondents; the panel at the bottom shows them for the
LM respondents. Psychometric functions are plots showing the percentage
of trials on which a vowel stimulus was heard as /a/ or /æ/ at each step along
the continuum. Note that the seven steps of the vowel continuum are ordered
from right to left. This is to agree with “standard” vowel space plots, like
those shown in Figures 9.2 and 9.4, which order F2 frequency values from
right to left to show their consonance with the front-back dimension of vow-
els. The lowest value of F2 for the present stimulus series occurred at step #1
(1245 Hz). F2 incremented thereafter (in 33 Hz steps) to its maximum value
(1443 Hz) at step 7.
It can be seen that the UP and LM groups responded similarly across
the vowel series. Both heard /a/ nearly 100% of the time at steps 1 and 2,
and then a progressively declining percentage of the time thereafter. The
percentage of /æ/ judgments was complementary, starting at zero, and then
incrementing gradually to virtually 100% by steps 6–7. The cross-over point
between /a/ and /ae/ is marked in each panel with a dashed line. It occurred
at step 4.7 for UP listeners, and at 4.8 for LM listeners. These cross-over
points provide a best estimate of the positioning of the perceptual category