COGNITION Cognition 72 1999 111 123 www elsevier com locate cognit Online processing of native and non native phonemic contrasts in early bilinguals Nu ria Sebastia n Galle s Salvador Soto Faraco IDIBAPS Departament de Psicologia Ba sica Universitat de Barcelona P de la Vall d Hebron 171 08035 Barcelona Spain Received 12 October 1998 accepted 13 April 1999 Abstract There is considerable debate about whether bilinguals can distinguish L2 phonemic contrasts as ef ciently as rst language speakers can To test this issue a group of highly pro cient Spanish dominant Catalan Spanish bilinguals who had been exposed to Catalan between the ages of 3 and 4 but who previous to this age had been exposed only to Spanish and another group of Catalan dominant bilinguals who had been exposed to Catalan from birth were compared in a gating task We developed a variation of the gating procedure that included a two alternative forced choice test after each fragment was played The differences between the two alternatives consisted of phonemic contrasts existing in Catalan but not in Spanish Four contrasts were tested two vocalic contrasts e e and o O and two consonantal contrasts S Z and s z The results showed that Spanish dominant bilinguals even the subset who were able to make correct identi cations at the last gate systematically performed worse than the group of Catalan dominant bilinguals needing longer portions of the signal to be able to correctly identify the stimuli We argue that these results support the hypothesis that L1 shapes the perceptual system at early stages of development in such a way that it will determine the perception of non native phonemic contrasts even if there is extensive and early exposure to L2 q 1999 Elsevier Science B V All rights reserved Keywords Speech perception Bilingualism Catalan Spanish Gating 1 Introduction To what extent can second language phonemic contrasts be mastered This is a question that has lead to heated debate among psycholinguists and has important Corresponding author Tel 134 93 402 1363 fax 134 93 402 1363 E mail address sebastia psico psi ub es N Sebastia n Galle s 0010 0277 99 see front matter q 1999 Elsevier Science B V All rights reserved PII S 0010 027 7 99 00024 4 112 N Sebastia n Galle s S Soto Faraco Cognition 72 1999 111 123 consequences for models of language acquisition Early phonologists Polivanov 1932 Trubetzkoy 1969 Trubetzkoy 1969 already noted that native phonemic categories exert an important in uence on the way we perceive non native speech sounds Indeed previous research has established that it is during the 1st year of life that our speech perception system is tuned to perceive phonemic contrasts existing in the language of our environment reducing our capacity to perceive non existing ones Werker Tees 1984a Probably the best known example is the dif culty that Japanese speakers have in distinguishing the phoneme r from l Goto 1971 A large number of studies have addressed acquisition of L2 phonemic categories and several models have been proposed to account for the different patterns of results encountered see Strange 1995 for an excellent review of the eld Among the factors considered to explain the dif culties in acquiring a second language contrast an important role is attributed to age of acquisition and duration of exposure In a few words the sooner and the longer the exposure the better L2 will be mastered Some authors Lively Pisoni Yamada Tohkura Yamada 1994 Flege 1995 propose that given an early and or intensive exposure very ef cient acquisition of foreign contrasts should be possible According to these models the perceptual system does not lose its capacity to distinguish new non native contrasts Miyawaki Strange Verbrugge Liberman Jenkins Fujimura 1975 Werker Tees 1984b Mann 1986 but compared with the expert native speaker the latter will focus on a more abstract level linguistic of processing freeing attentional resources to process language in a more ef cient way Mayberry Eichen 1991 Jusczyk 1993 Lively et al 1994 For instance in the rst postulate of his speech learning model Flege 1995 argues that the mechanisms and processes used in learning the L1 sound system including category formation remain intact over the life span and can be applied to L2 learning p 239 What these models propose is that listeners weight the acoustic features of the input in an optimal way to perceive their native language i e in a language dependent way When learning a second language the speech signal will be sieved through this L1 tuned weights As listeners receive more exposure to L2 initial weights will be realigned to better classify the new phonemic categories including the creation of new L2 categories Obviously this realignment will be better in L2 processing the longer and the earlier in life the exposure takes place But is there any limit to this malleability in the formation of new phonemic categories In the present study we address this question by analyzing second language perception in a group of Catalan Spanish bilinguals who began to be extensively exposed to L2 Catalan at or before the age of 4 years old The comparison group will be a sample of Catalan Spanish bilinguals who learned Catalan as their rst language These two samples have had a similar linguistic input from 4 years onwards the only difference being their linguistic experience during the rst 4 years of life Few studies have examined the phonological systems of early bilinguals e g Mack 1989 Pallier Bosch Sebastia n 1997a and those studies that exist suggest that in naturalistic circumstances without highly specialized training the speech perceptual system does not seem to be prone to modify initial phonemic categories In the study of Pallier et al 1997a Spanish dominant bilinguals individuals who N Sebastia n Galle s S Soto Faraco Cognition 72 1999 111 123 113 learned rst Spanish at home and before the age of 6 then learned Catalan performed rather poorly in categorical perception tasks with the Catalan phonemic contrast e e in Spanish there is only one e category see below for a more detailed description of the differences between Catalan and Spanish at the segmental level However it is wrong to assume that all Spanish dominant bilinguals in the Pallier et al study were not able to perform the e e contrast Indeed while performance of most Spanishdominant bilinguals on categorical perception tasks for L2 materials was erratic some of them performed as