| This dissertation compares the quantity and quality of dyadic interaction and learner discourse in three discussion formats: (1) oral discussion among non-native speaker (NNS) classmates, (2) computer-assisted discussion (CAD) between native speakers (NS) and NNS learners, and (3)CAD among NNS classmates. The investigation departs from previous research in that earlier studies have almost exclusively examined the effects of CAD in full class discussions among NNS classmates (Kelm, 1992; Kern, 1995; Beauvois, 1997). This design, however, fails to incorporate the work of interactionist theorists who have shown that small-group work is better than full-class discussion for promoting target language practice and interaction (Long 1983; Varonis & Gass, 1985; Pica 1994). Furthermore, discussion among NNS classmates does not take full advantage of the power of CAD to engage learners in authentic interaction with NS.; In addition to the question of how CAD can enhance foreign language methodology, the present study also explores how the data can inform current second language acquisition (SLA) theory, specifically the Interaction Hypothesis (IH) and Sociocultural Theory (SCT). The quantitative analysis reveals significant differences between the three discussion formats with regard to the quantity of learner production and interactional modification. The qualitative analysis, however, suggests that the quality of the interaction is more important than the quantity of interaction for SLA. Furthermore, the conclusions indicate that the IH framework does not allow for a sufficiently detailed investigation of the quality of interaction. For example, the results reveal that simply counting the occurrence of interactional features provides very little information about the usefulness of the interaction in making input to the learners more comprehensible or encouraging the learners to modify their output. Therefore, SCT is proposed as a richer model that provides a better framework for the investigation of learner discourse. Using this framework to examine the degree of learner involvement in the discourse, the results show that involvement is related more to interlocutor than to discussion format and that learners are most involved, regardless of task, in the NS/NNS CAD environment. Finally, the pedagogical implications of the research are discussed. |
