Advances in Sociophonetics

36 Bernard Laks, Basilio Calderone and Chiara Celata

2. The distributional analysis of French liaison

This analysis deals with 16,805 enacted liaisons produced in either free or guided

conversations.

Each enacted liaison corresponds to a liaison environment, comprising a left

word ending in a consonant and a right word beginning with a vowel. Although

in very rare cases the liaison consonant may be realized as unlinked (that is, with

the consonant realized as the coda of the first word; Encrevé 1988), the present

analysis deals with linked consonants only. The total repertory of liaisons therefore

includes all liaison environments with their frequencies of occurrence. As our aim

here is to analyze the productivity power of each type of environment by calculat-

ing the number of realizations produced, we concentrate here on enacted linking;

we postpone the ratio analysis of realized vs. virtual liaison to a forthcoming paper

on the social stratification of French liaison.

Figure 1 provides an example of a liaison environment. Recording ID 75cv11

corresponds to the performance <trois> linked <ans> [trwazɑ̃]. The lexical con-

text for this particular occurrence enables us to construct the environment type

<trois> linked <ans>. If the combination <trois> linked <ans> appears in other

recordings of the PFC corpus, the frequency of occurrence (or token frequency)

of this liaison environment will be > 1.

ID: 75cv11 ...Et depuis deux trois ans depuis que je paye mon appartement j’habite...

LIAISON

le context trois ans right context

Figure 1. A liaison environment.

We ranked all liaisons’ environments attested in the PFC corpus according to their

token frequency. Next, the frequency cumulative percent of all environments was

calculated to evaluate the productivity of each individual environment and of the

subgroups of these environments.