110 Zsuzsanna Fagyal
contribution to the model (r^2 =0.58). It was positively correlated with ethnicity,
which means that, individually, AF speakers contributed more to the observed
variations than EF speakers. The main question, however, still remained: did
the three rhythm type indices account for any variation in any way at all in
the sample?
Residu al va r iat ions of %V, ǻV, a n d ǻC not accounted for by the regression
model showed a normal but bimodal distribution, split in two parts respec-
tively at +1 and -1 standard deviation from the sample mean. Since %V, ǻV,
and ǻC were inversely correlated with articulatory rate (see Table 4.3), this
split could be due to some extent to fast speech processes, known to induce
the compression of segmental durations, and thus making AF speakers pat-
tern differently from the more “slowly articulating” EF speakers.
The general conclusion is that ethnic origin and related heritage language
use in this sample is tied to a large extent to differences in articulatory rate,
splitting speakers in two well-de¿ ned groups: the group of the younger, more
slowly articulating EF speakers who also tended to be better students, and the
group of the somewhat (one grade) older, fast articulating AF speakers who
also had overall lower average grades in school. As it turns out, these param-
eters proved to be meaningful within the adolescent male peer-group social
order (see Discussion).
What remains of heritage language inÀ uence? One can hypothesize that
AF speakers’ readings were slightly more consonantal (see negative correlation
with %V), because these speakers tended to elide more vowels, and therefore
had more complex onsets and codas than EF speakers. Since devoiced vow-
els were considered consonantal, the slight amount of devoicing could have
reduced the number of syllables in the AF group, and contributed to a decrease
in %V and an increase in ǻV values. This hypothesis was examined next.
3.3 Beyond global measures: Syllable structure
Figure 4.4 shows the number of different types of syllables in EF and AF
speakers’ speech. The magnitude of differences between the categories was
reduced on a logarithmic scale for easier reading. As one would expect it in
French, CV syllables were the most numerous (61%) in both groups’ readings.
They were four times more frequent than the next most frequent type of syl-
lable: CVC (14%). With the exception of complex codas (CVCC), found only
in the syllable /jabl/ of the word incroyable ‘incredible’ in two EF speakers’
readings, all other types of syllables showed virtually identical percentages: