Advances in Sociophonetics

44 Bernard Laks, Basilio Calderone and Chiara Celata

for /z/ and /n/ than for /t/, as was the group of lexical environments occupying

the head positions. Table 2 shows that the alveolar stop is the least represented

consonant if one considers the 30 most frequent environments; however, the same

holds true even if one considers the tail of the curve. In sum, the body and the tail

of the curve do not reveal any significant difference in the distribution of liaison

environments across consonant categories.

From this analysis we can conclude that the tendency towards the power-laws

distribution that we have observed over the whole repository of forms remains sta-

ble overall even if we consider different phonological contexts of enacted liaisons

separately. There are however some idiosyncrasies in the head zone related to the

fact that the nasal is the liaison consonant occupying the five highest positions in

the frequency curve, and in particular, the lexical environments including on + a,

est, avait, était stand out because of their enormous number of occurrences in the

corpus. This distributional bias accounts for the particular behaviour of the nasal

consonant in the head zone, compared to /z/ and /t/.

3.3 Distributional analysis of liaison types according to age

and educational level

By considering the productions of different groups of speakers we wanted to ver-

ify whether any of the liaison environments is specific to any group of speakers

in such a way that the general picture of the power-laws distribution disappears

when liaison is analysed with respect to particular subsets of linguistic use. These

subsets are defined here in terms of some basic, broadly conceived sociolinguistic

factors, that could in principle influence the linguistic behavior of the speakers

with respect to the production of liaison (see above, §1, for some background

on sociolinguistic variability in the production of French liaison). As specified

above, we took age as a typical sociolinguistic factor generally considered to influ-

ence non-obligatory liaison production. On the contrary, educational level was

included in the analysis because it represents a factor which is still poorly investi-

gated. As for age, we divided the PFC speakers into two groups: those aged 50 or

below, and those older than 50. We are conscious of the fact that this bipartition is

not as fine-grained as a sociophonological inquiry would require, but there were

limitations in the distribution of the PFC speakers across the different age classes

that forced us to adopt such a neat bipartition. As for educational level, we created

three groups of speakers: those with up to 14 years of scholastic education (‘low’

educational level henceforth), those with up to 18 years (‘intermediate’ educational

level) and those with up to 24 years (‘high’ educational level).