The purpose of this study, which focused on the word taco as spoken in Spanish and English, was to explore the word as a unit in second language (L2) speech acquisition. As expected, acoustic measurements revealed that Spanish and English monolinguals' renditions of taco differed systematically. It was also shown that the extent to which Spanish/English bilinguals approximated English phonetic norms for any one segment of taco was correlated with their approximation for the other three segments, and that early learners differentiated Spanish versus English taco more than did late learners. It also appeared that the bilinguals produced /t/ with less English-like voice onset time (VOT) values in English taco than in other English words without a cognate in Spanish. In a perception experiment, listeners were able to identify the native language of Spanish and English monolinguals on the basis of their production of taco. The listeners heard larger differences between Spanish and English taco tokens spoken by early than late learners of English L2. Two additional perception experiments assessed further the phonetic dimensions that listeners use to determine language identity and to gauge bilinguals' speech production accuracy. Listeners assigned to language identification and goodness rating tasks responded to acoustic information distributed over all four segments in taco, although the VOT of the word-initial /t/ appeared to be the single most important phonetic dimension. Taken together, the results of this study suggest that (a) bilinguals' accuracy in producing the various segments of a second language word may be interrelated and (b) in judging L2 speech, listeners respond to phonetic errors distributed over the entire word.

This study examines the role of grammatical prerequisites on code-switching in young bilingual children. It is proposed that code-switching is constrained not only by grammatical properties of the languages involved; it is also regulated by principles and mechanisms of language use. Constraints on code-switching are therefore defined as processing principles that, however, depend on grammatical knowledge. They ensure that switching does not result in a violation of grammatical coherence, defined in terms of both linear sequencing and structural configuration. Some of these claims are tested empirically, analyzing the speech of two bilingual children acquiring French and German simultaneously. It is argued that even in the earliest uses of mixing, constraints are not violated; in many cases they do not apply because the relevant grammatical relations do not yet hold. Code-switching is nevertheless used from early on in accordance with these constraints, as soon as a certain kind of grammatical knowledge is accessible. Most importantly, functional categories have to be implemented in the child's grammar.

Reading in a second language (L2) can produce inefficient processing in otherwise proficient readers. This paper argues that mental translation during L2 reading may facilitate the generation and conservation of meaning by allowing the reader to represent portions of L2 text that exceed cognitive limits in a familiar, memory-efficient form. Fifty-one intermediate-level French students, in high, middle, and low reading ability groups, participated in think-aloud interviews while reading French texts. The relative frequency of translation use among these groups is compared at the beginning and end of a semester and is found to decrease with level of reading ability. The specific contexts in which students relied on translation are identified, and functional benefits and strategic uses of translation are discussed. Finally, hypotheses and questions are developed for future research.

The present paper addresses the argument that access to UG in SLA implies an asymmetrical knowledge of ungrammaticality, which is evidenced in superior accuracy and uniformity in learners' judgments of ungrammatical structures versus grammatical structures (Felix, 1988; Zobl, 1992). I attempt to show that the asymmetry position is conceptually defective, that the evidence for it is inconclusive and inappropriate, and that the evidence adduced is contradicted by other data.

This paper responds to David Birdsong's discussion of the positions and the data presented in Felix (1988) and Zobl (1992). We attempt to clarify a number of misconceptions that we feel characterize his discussion and elaborate further on the performance and competence issues contained in the earlier papers.