A number of studies on second language (SL) instruction point to the hypothesis that instruction is effective, but determining the effect experimentally has been problematic. Overall, three difficulties with previous attempts to demonstrate a causal relationship between SL instruction and second language acquisition can be identified: (a) inappropriate or inadequate research design, (b) failure to operationalize or even to describe the instructional treatment, and (c) choice of SL assessment measures. This article presents the findings of a recent empirical study which (a) show that SL instruction is effective; (b) show that attention to form, either via detailed analysis of structure or highlighting of target language (TL) structures in context, promotes acquisition of interlanguage (IL) grammar, but that only the latter comes hand-in-hand with comprehension of input; and (c) replicate earlier findings suggesting an important role for markedness theory in instructed IL development.

Recordings of naturally produced stop-vowel English words with a wide range of voice onset time values were used to investigate nonnative perception of voicing categories. L2 learners of English from nine language groups and native American English speakers served as subjects. The responses of the language learners suggested that they were using a hybrid perceptual system, one in which the English voicing categories were not yet fully established.

There are indications that a determined effort toward theory construction is currently being mounted in SLA (Gregg, 1989, 1990; Long, 1990; Spolsky, 1989). This raises a number of epistemological issues, one of which is the question of multiple theories and lack of convergence (and the related question of distinguishing science from nonscience–and from nonsense). In a field where there are many different therorise, is the relationship between them one of complementarity or of opposition? Are there criteria for appraisal and choice? Issues relating to theory construction have received considerable attention in the literature (e.g., Lightbown, 1984; McLaughlin, 1987). This article aims to build on such initiative asking: (a) Are there multiple theories of SLA? (b) Is it a problem? and (c) what criteria have SLA researchers used to appraise and choose between theories, and how compelling are they? Following this, consideration is given to top-down and bottom-up approaches to SLA theory construction. I conclude by advancing the modest claim that rationality in SLA therory construction is possible, in spite of insistent problems.

Beck and Eubank (1991) criticize our recent SSLA article (Tomasello & Herron, 1989) on both theoretical and methodological grounds. While we appreciate their attempt to discuss and clarify important issues—and while they do make several sound and very interesting points—in a number of cases they seriously misrepresent our study. We will attempt to address the criticisms in roughly the order in which they were raised.