Effects of L2 use on the production of L1 vowels in early Spanish-Catalan bilinguals

A considerable amount of research on early bilingualism investigating the plasticity of the perceptual system in L2 learning has focused on the Catalan mid vowels /e/ and /ε/ [1,2,3]. This research has shown that even early Spanish-Catalan bilinguals who are highly proficient in Catalan exhibit an asymmetric behaviour in lexical decision tasks in which Catalan-dominant bilinguals obtain higher error rates in the identification of /e/-nonwords than /ε/-nonwords. The source of this asymmetry may be attributed to the coexistence of native and mispronounced Spanish-accented /ε/-words in the mental lexicon of Catalan-Spanish bilinguals (/gaλεdə/ vs.*/gaλedə/) [4,5] or to the phonetic " weakness " of the mid-vowel contrasts due to various factors, including the pervasive influence of Spanish in a bilingual environment [6,7]. Evidence of this trend is provided by production data from children whose mid-vowels have been found to vary as a function of amount of exposure to Spanish in Barcelona [8]. In the present study we examined vowel production data from a group of 82 Spanish-Catalan (S-C) bilinguals born and raised in Barcelona (selected form a larger pool of 719 S-C bilinguals), first exposed to their L2 by school age (3-6 years), who had been found to perceive the mid-The participants read a passage in Catalan at normal speed containing target words with the 7 stressed vowels of Catalan. 5-8 words per vowel were selected according to consonantal place of articulation. Vowel frequency measurements (F0, F1, F2) were converted to Bark (B) and then normalized by computing the difference between Bark-converted F1 and F0 (B1-B0) and between Bark-converted F2 and F1 (B1-B0) for vowel height and frontness-backness, respectively [9]. Euclidean distances between B1-B0 and B2-degree of overlap in production; greater spectral distances on a two-dimensional vowel space representing greater robustness of contrasting vowel pairs in production [10,11]. as the within-subjects factor yielded a significant main effect for vowel (F(3, 76)=71.3; p<.001) and a non-significant main effect for subject group (F(3, 78)=71.3; p=.457). However, a significant vowel x subject group interaction (p=.005) indicated that differences in vowel production across subject groups were dependent on vowel type, as /ε/ and /ѐ/ varied consistently as a function of subject group: the more frequently Catalan was used, the more open and less fronted (less Spanish-like) the mid vowels were. Similar analyses with spectral distance scores as the dependent measure revealed significant differences according to subject group for Tukey post-hoc tests revealed that MostlyC subjects produced …