34 Bernard Laks, Basilio Calderone and Chiara Celata
seventy-six enquiries, thirty-three are now terminated, and are available in the
PFC database. In addition, thirty-three are still being processed (transcription,
coding, etc.) and ten are currently being completed. In all, the PFC base now
boasts recordings of 489 French speakers representing a total of around 730 hours
of digitally recorded and indexed speech currently available online. A total of
forty-one hours are transcribed, aligned and coded for liaison and for schwa.
Phonetic, phonological, lexical and discursive data have been collected by
means of a series of Labovian enquiries (see for more recent approaches Labov et
al. 2006), in which each speaker is classified according to gender, age, and socio-
economic status. For each speaker, different types of oral production have been
recorded: word list reading, text reading, guided one-to-one conversation, and
free/spontaneous conversation with peers.
With specific regard to liaison, 49,728 sites of potential realization have been
coded according to the relevant segmental and contextual information. This com-
prehensive and theoretically neuter coding allows the researchers to have direct
access to multiple sorting possibilities.
Located at the intersection of at least four contradictory historical dynam-
ics, liaison is a very complex phenomenon (Durand et al. 2011). The historical
dynamics leading to the fall of the final consonant since late antiquity Latin has
fostered open syllabification. In contrast, generalized linkage, insofar as it fosters
long phonological sequences, promotes the maintenance of these final consonants
before an initial vowel, as a support for CV linkages. Furthermore, some of these
consonants take on a role as markers for number and person, and thus tend to
elude the dynamics of erasure: they tend to be conserved not in the coda of the
final syllable of the word which they mark, but rather, as onsets of the opening
syllable of the following word, thus undergoing resyllabification. Finally, spell-
ing conventions demand that the etymological final consonant be consistently
transcribed, whether it is realized in pronunciation or completely lost (Laks 2005,
2006, 2011; Clédat 1917).
Although generative and post-generative phonology (Schane 1965; Dell
1973; Encrevé 1988; Tranel 1995a, 1995b) has proposed that French liaison is a
homogenous phenomenon that can be represented as a single process whereby
a variable surface erasure rule applies to abstract underlying consonants, more
recent work has shown that liaison is a multifactorial and multilevel phenom-
enon highly sensitive to frequency effects (de Jong 1994; Fougeron et al. 2001;
Laks 2007; Durand et al. 2011).
In particular, the PFC database has allowed large-scale investigations concern-
ing all major factors of sociolinguistic and geographical variability, including age,
gender, French spoken as a first vs. second language, diatopic variation of northern