Chapter 2. French liaison and the lexical repository 51
the words included in the constructions such as their “relations sémantiques,
fonctionnelles ou statistiques particulières lorsqu’ils [= the words] sont utilisés
ensemble” (Bybee 2005: 26). One very typical example of liaison construction
is [ determinant + noun], as e.g. in
instantiation being a relatively high-frequency lexical environments, according
to our own analysis, as it occupies position 13th in Table 1). However, syntactic
cohesion and frequency of co-occurrence sometimes conflict. One such case is
reported in Bybee (2005: 35): the construction [verb + determinant] in the
specific case of
liaisons than any other contexts in which
initial words, including those with a higher degree of syntactic cohesion (such
as [auxiliary + past participle]
is therefore forced to conclude that “[c]ette tendance suggère fortement qu’il
existe une construction dans laquelle est [t] un est un constituant qui precède
un nom” (Bybee 2005: 26).
Our analysis by making an appeal to the notion of lexical environment and
completely disregarding the convention of referring each lexical environment
to a general morpho-syntactic label, overcomes this impasse and proposes a sta-
tistical approach in which liaison environments are exclusively specified by the
frequency of their usage in the corpus. On this basis, our analysis emphasizes the
statistical distribution of actual usages in a very large repository of forms, with
no superimposed predictions about how and to what extent syntactic cohesion
produces secondary assemblages within the classes of exemplars (Foulkes 2006).
- Conclusions
Starting from the acknowledgment of corpus phonology as a truly sociophono-
logical practice, we have proposed a frequency analysis of French liaison focusing
on the specific contexts of liaison realization attested in the PFC database. All
attested combinations of two consecutive words generating liaison at their junc-
ture were analyzed to uncover the distributional aspects of French liaison in its
actual lexical instantiations.
The results of the analysis, indicating the existence of a power-laws distribu-
tion in the production of liaison, suggested that there is a complex function regu-
lating the number of liaison environments and their token frequency in a corpus.
The phonological component, intended as the nature of the liaison consonant,
does not add much information to the analysis of such distribution, since very
few inconsistencies are found across different consonants. External factors, such
as the age and the educational level of the speakers, neither provide substantially