106 Zsuzsanna Fagyal
indicates that AF and EF speakers patterned again with the Romance group
and differed from the Germanic group that had overall higher ǻV values. The
ten speakers’ speech samples were more vocalic, however, than previously
measured for French. The EF group again appeared more vocalic, but vocalic
interval durations (ǻV) varied to the same extent in both groups.
Neither vocalic nor consonantal durations varied homogenously in this
sample.^32 While a log transformation of measurements rendered the vari-
ance of consonantal duration measurements more homogenous, and therefore
allowed the use of parametric statistical tests, no mathematical trick could
accomplish the same for vocalic durations.^33
The non-parametric version of one-way ANOVA, the Kruskal-Wallis test,
with speaker as the grouping variable indicated signi¿ cant inter-speaker differ-
ences with respect to the length of vocalic intervals (H=191.635, median=8.85,
df=9, p<0.001) and consonantal intervals (H=27.299, median=9.41, df=9,
p<0.001). The longest vocalic and consonantal intervals were measured in the
slowest speaker, Alain’s reading, followed by four other EF speakers: Chris,
Jacob, Karl, and Octave. They were followed by all AF speakers who tended
to have short vocalic and consonantal intervals. Yasin’s reading showed the
shortest vocalic and consonantal interval durations, and he was also the fast-
est speaker in the corpus. Notice, however, that this ranking only partially
overlapped with speakers’ articulatory rate: Octave from the EF group for
Figure 4.3 Ave r age %V a nd ǻV values for French speakers of North African (AF)
and European descent (EF) and languages representing main rhythm
types (Ramus et al. 1999: 273).