This study investigates the generalizability of claims by Logan (Klapp, Boches, Trabert, & Logan, 1991; Logan, 1985, 1988a, 1988b; Logan & Klapp, 1991) about the development of automaticity in the adult learning of alphabet arithmetic problems to the context of adult second language acquisition. Logan's proposal is that as individual solutions to problems accumulate in memory a transition in problem-solving procedures takes place. This transition involves the shift from an algorithm-based procedure for deducing correct solutions to direct retrieval of individual solutions or instances from memory. In the present study, second language learners of English were presented with a rule for understanding the morphological constraint on the dative alternation (Mazurkewich & White, 1984) and asked to judge the acceptability of 36 sentences presented in a training set. The sentences were controlled for frequency of presentation, one being presented eight times, one seven times, and so forth. When presented together with novel instances of the same type in a transfer set, reaction times to old instances were significantly faster. Reaction times to repetitions of the previously presented verbs in new frames and novel verbs in old frames were compared as a test of hypotheses about strategy switches in processing alternating and nonalternating verbs.

Variability, both systematic and nonsystematic, has been the subject of much debate in recent years in the study of learner interlanguage. This article presents empirical evidence from a longitudinal study of a small group of advanced learners of French. Variability is noticeable throughout the period of learning of a particular structure in French. Nonsystematicity observed in individual learners' performance is explained in the light of developmental stages such as those put forward in Gatbonton's diffusion model. When looked at in terms of developmental processes, nonsystematic variability becomes an essential element of progress. Patterns of development are observed across the subjects, and learning can thus be seen to follow a systematic route.

During the past decade and a half, a great deal of research has posited a developmental sequence approach in second language acquisition. Much of this research has proposed that certain linguistic structures are acquired in a natural immutable order while other linguistic structures are acquired variably as a result of a learner's orientation (Clahsen, Meisel, & Pienemann, 1983; Meisel, Clahsen, & Pienemann, 1981; Pienemann, 1984, 1985, 1987; Pienemann & Johnston, 1987). Proposals have been made for extending the model into language assessment and pedagogy (Clahsen, 1985; Pienemann, 1984, 1992; Pienemann & Johnston, 1987). The present study reexamines the original social-psychological research upon which the multidimensional model is based and shows that it is incorrect due to faulty analyses. Further, it examines the limited applicability and generalizability of the developmental sequence approach for assessment and pedagogy.

The study reported here extends Tyler and Bro's (1992) investigation of the sources of native speakers' perceptions of incoherence in English text produced by nonnative speakers. Using paper-and-pencil tasks, the original study examined two competing hypotheses: (a) The primary source of interference was the order in which the ideas were presented versus (b) the primary source of interference was mismatches in discourse structuring cues. They found no effect for order of ideas but a strong effect of discourse structuring cues. In the present study, 80 subjects were tested on the same texts as those used in Tyler and Bro (1992) but using microcomputers. Subjects rated the text for comprehensibility and answered three questions concerning the propositional content. The computer format represented a more sensitive measure of subjects' reactions to the text because it did not allow looking back and because it provided information concerning differences in reading time for each manipulation. Once again, the results of the comprehensibility ratings showed a strong effect for miscues and no significant effect for order of ideas. Results of the true/false questions indicated that presence of miscues affected subjects' comprehension of the propositional content but that order of ideas had no discernible effect. Finally, reading time results also showed a strong effect for miscues and a mixed effect for order of ideas, suggesting that order of ideas does make a minor contribution to comprehensibility.