4 Chiara Celata and Silvia Calamai
function operated by the linguistic community with respect to that variability. As
the acquisitional evidence reviewed in the paper demonstrates, the language learner
has a “compulsion [...] to turn outward” (p. 18) in the direction of the community
patterns and successive contacts experienced in life after the formation of the initial
competence may erode parents’ influence in a gradual but continuous way.
From this analysis the author concludes that the individual patterns are not as
informative as is normally suggested by proponents of different types of “linguistic
individualism”; it is rather the case that the larger the picture, the more informa-
tive the data therein.
The paper by Labov constitutes an excellent introduction to the rest of the
papers collected in the volume. The emphasis on the accommodation mechanisms
that individual learners carry out in the construction of individual grammars is
but one of the many possible responses to the question of how the vast amount of
speech variation with indexical meaning is cognitively represented – an issue that
is directly or indirectly addressed by all of the papers in the volume.
In a similar vein, and taking its inspiration from usage-based models of gram-
mar and exemplar theory (Goldinger 1998; Bybee 2001), the paper by Bernard
Laks, Basilio Calderone and Chiara Celata “French Liaison and the lexical reposi-
tory” starts from the well-known lexicalist hypothesis that liaison is more fre-
quently realized in those word groups that have strong internal cohesion and high
frequency of co-occurrence (Bybee 2005), and shows that this hypothesis, whose
substantial correctness must be confirmed, should be refined in some respects if
we take the procedures of corpus analysis seriously and investigate a very large
amount of actual productions realized by different groups of French speakers.
The materials of the analysis are the 16,805 sites of realized liaisons coded in
the PFC corpus (Durand et al. 2002). PFC is the largest database of spoken French
currently available, and has been collected over many years according to the
“Labovian” paradigm of sociolinguistic enquiries. Based on such large repository
of actual uses, the paper shows that the liaison distribution is similar to a power-
law distribution in which a few word junctures are ranked high for productivity
and account for approximately one-half of the total observations, while a very long
list of less productive or unproductive junctures accounts for the remaining half
of the realizations. The authors argue that this statistical distribution goes beyond
some traditional views of linguistic storage, according to which part of the liaison
process must be inscribed as a nucleus of stored “constructions” in the mental lexi-
con while low frequency constructions tend to be lost (Bybee 2005). Quite on the
contrary, liaison shows that storage is limited to a relatively short (but cognitively
heavy) list of occurrences, while “a productive process of generalization” (p. 39)
must account for the long tail of dispersed, low-frequency realizations.