(www.bangortalk.org.uk). We thank Jorge Valdés Kroff for his assistance with the
statistical analysis. We are also indebted to Fraibet Aveledo, Lofti Ayali, Manuel
Delicado Cantero, Gwyn Lewis, Esther Núñez Villanueva, Rocío Pérez-Tattam,
Alberto Rosignoli, Carmen Silva-Corvalán, Gerald Stell, Enlli Thomas, Jorge Valdés
Kroff, two anonymous reviewers and the audiences of the 7th International
Symposium on Bilingualism (Utrecht, July 2009) and the 39th International
Symposium of the Spanish Linguistics Society (Santiago de Compostela, February
2010) for useful discussion of earlier versions of this paper.
(2) The glosses in examples contain abbreviations to be understood as follows: 1P = 1st
person plural, 1S = 1st person singular, ADJ = adjective, ADV = adverb, CONJ = con-
junction, DEF = definite, DEM = demonstrative, DET = determiner, IM = interac-
tional marker, M = masculine, NT = neuter, PRES = present, PRESPART = present
participle, PRON = pronoun, REL = relative, SG = singular, V = verb.
(3) Syntactic categories (i.e. words) can combine with other syntactic categories to form
constituents.
(4) Within the MLF model it is assumed that subjects originate outside VP.
(5) Bilingualism in Wales can be dated from the introduction of universal (English-
medium) education in 1870. Deuchar (2005) shows that about 35% of the population
in Wales was bilingual in 1901 while 50% were English monolinguals and about 15%
Welsh monolinguals. By 2001 just over 20% of the population were Welsh-English
bilinguals with the rest English monolinguals and no Welsh monolinguals.
(6) See Gathercole and Thomas (2007) for evidence that language proficiency self-reports
match the actual proficiency of the speaker.
(7) See Gal (1978) for evidence that there is a strong correlation between language choice
and social network.
138 Part 3: Bilingual Language Use