318 Thomas C. Purnell
components identi¿ es the factor to which a speci¿ c measure likely falls for
the interpretation of the latent factors. An interpretation of Table 13.11 reveals
that there are different and similar latent factors to acoustic and perceptual
data. The factors appearing active for the acoustic data are by and large
divided by overall vowel quality: ¿ rst MID-BACK VOWEL, then INTEN-
SITY, then MID-FRONT VOWEL. In contrast, the factors appearing active
for the perception data are by the two features of vowel quality: ¿ rst VOWEL
HEIGHT, then /ܭ/ BACKNESS, and /o/ BACKNESS. That HEIGHT precedes
BACKNESS should be unsurprising in light of the primacy of vowel height
across languages (Ladefoged and Maddieson 1996). BACKNESS would then
provide a secondary division of vowels within each vowel height.
Three graphic representations of the “hello” data are shown in Figures
13.9 through 13.11. Since the three strongest measures (Table 13.9) for both
the acoustic and perception data fall into distinct latent factors (Table 13.11),
the raw values for those three measures were plotted against each other. In the
plot of acoustic features (Figure 13.9), the space discriminating the three dia-
lects is somewhat unclear. However, in the plot of the perceptual data (Figure
13.10), the data is more distinct across dialect. One reason for the ef¿ cacy of
two /ܭ/ variables is seen in the need to distinguish the SAE /ܭ/ which display a
wider backness variation than the other two dialects (Figure 13.11).
Table 13.11 Eigenvectors for Multivariate Results of the First Three Principal
Components
12 3
Acoustics
/ܭ/ F3-F2 (Z) -0.396 -0.335 0.580†
/o/ F2 (Z) 0.475† -0.049 0.117
RMS Standard Deviation 0.448 0.565† 0.040
/ܭ/ F1 (Z) 0.357 -0.121 0.703†
Maximum Flow Declination Rate -0.493† 0.359 0.020
RMS Initial Value -0.218 0.650† 0.392
Perception
/o/ F2 (Hz) -0.383 -0.310 0.662†
/ܭ/ F1 (Z) -0.562† 0.310 -0.085
/ܭ/ F3-F2 (Z) 0.113 0.894† 0.223
Change in RMS 0.393 0.026 0.694†
/o/ F3-F1 (Z) 0.609† -0.091 -0.152
† Identi¿ es the eigenvector used for plots.