40 Bernard Laks, Basilio Calderone and Chiara Celata
Langacker 1987). In particular, the phenomenon of the ‘long tail’ asks for a refine-
ment of the view that part of the liaison process must be inscribed as a repository
of ‘constructions’ in the mental lexicon (Bybee 2005; Bybee and McClelland 2005).
According to the latter view, ‘irregular’ liaison is stored in the lexical repository
of constructions as a list of types, and “a certain level of token frequency is neces-
sary to maintain these irregularities” (Bybee 2005: 24). This analysis fits well with
the head of the curve: tracing the 13 lexical environments (see Figure 2) which
account for 30% of the overall data back to a set of derivational rules cannot in
any case be shorter than providing the exhaustive list which accounts for them.
However, the tail of the distribution cannot be dealt with by the same cognitive
mechanism of full-form storage: most of the types that we find here have a globalTable 1. Number of occurrences (frequency), percent and cumulative percent for the 10
highest ranked the 13 lowest ranked liaison types in the PFC corpus.
Realizations of liaison in the PFC corpus
16,805 tokens and 3,105 types
Rank Liaison Frequency Percent Cumulative percent
1 on_L_a 1318 7.84 7.84
2 on_L_est 653 3.89 11.73
3 ils_L_ont 555 3.30 15.03
4 en_L_a 401 2.39 17.42
5 on_L_avait 339 2.02 19.43
6 on_L_était 290 1.73 21.16
7 quand_L_on 268 1.59 22.76
8 dans_L_un 240 1.43 24.18
9 deux_L_ans 210 1.25 25.43
10 dans_L_une 209 1.24 26.68
... ... ... ... ...
3093 vous_L_apporte 1 0.01 99.93
3094 vous_L_apprenait 1 0.01 99.93
3095 vous_L_apprend 1 0.01 99.94
3096 vous_L_apprendre 1 0.01 99.95
3097 vous_L_approvisionner 1 0.01 99.95
3098 vous_L_arrêter 1 0.01 99.96
3099 vous_L_arrêtiez 1 0.01 99.96
3100 vous_L_arriverez 1 0.01 99.97
3101 vous_L_arriviez 1 0.01 99.98
3102 vous_L_assurer 1 0.01 99.98
3103 vous_L_attendez 1 0.01 99.99
3104 vous_L_auriez 1 0.01 99.99
3105 vous_L_autres 1 0.01 100.00