Perceptions of /a/ fronting Across Two Michigan Dialects 231variation in F1 and F2 was very similar for the two talkers. It can also be seen
that corresponding vowels were generally very similarly placed within the F1/
F2 space, except in instances where an effect of NCCS would be expected.
A parametric speech synthesizer (Sensimetrics 1997) was used to gener-
ate the /a/-to-/æ/ synthetic vowel series. Each vowel item was synthesized
with a fundamental frequency contour that fell linearly from 120 Hz at onset
to 100 Hz at offset. With the exception of F2, formant frequencies were ¿ xed
at approximately the mid-point of the range between /a/ and /æ/, as produced
by both talkers (F1 = 750 Hz, F3 = 2500 Hz, F4 = 3500 Hz).
The frequency of F2 was varied from stimulus item to stimulus item. In
all seven different vowel stimuli were created, with F2 values ranging from
1245 Hz to 1443 Hz in 33 Hz steps. Pilot testing showed that listeners’ percep-
tions regularly transitioned from /a/ to /æ/ within this interval.
4.3 “Hot—hat” and “sock—sack”
The complete vowel series was embedded in each of two CVC syllable frames,
with appropriate formant transitions imposed at onset and offset. The ¿ rst frame
was /hVt/, which yielded a word series that varied from “hot” to “hat” along the
continuum. The second frame was /sVt/, which varied from “sock” to “sack.”
Figure 9.5 shows the endpoint stimuli from each of these word series.
Figure 9.5 Spectrograms of the Step 1 and Step 7 versions of “hot—hat” (below) and
“sock—sack” (above).