Talking to Avatars: The Computer as a Tutor and the Incidence of Learner's Agency, Feedback, and Grammatical Form in SLA

Abstract

Increasingly, SLA research is exploring models of hybrid learning where computers are used not only as tools for information and communication purposes but as autonomous electronic tutors. Incipient evidence has shown that e-tutors can facilitate L2 grammar development and even supersede other instructional technologies (e.g., Nagata, 1996; Nutta, 1998; Watts, 1989). However, very few theoretical arguments have been presented to explicate these findings. Quite possibly, the edge of e-tutors may arise from their ability to engage learners in interactive practice (Nagata, 1996), while other multimedia technologies may only, at best, showcase pre-recorded interactions. According to the Interaction Hypothesis (Long, 1996), participation in interactive practice can promote SLA because it allows learners to notice, comprehend, and process input, produce output, form and revise hypotheses in response to feedback, and produce modified output, processes which have been shown to be developmentally helpful (Mackey, 2007). However, the IH does not specify whether or not the effects of these processes are determined by the type of agency of the learner; that is, whether it is necessary for learners to engage proactively in the aforementioned processes, or whether mere exposure to interactive practice by others (by e.g., watching a video) may suffice. On this note, existing empirical research is scant and contradictory, with Mackey (1999) showing a benefit for practice over exposure to it, Muranoi (2000) showing no significant differences, and Hsieh (2007) yielding mixed results.

Arguably, the relative effects of learner's agency may be moderated by the type of practice that is being performed or observed. One usual component of practice is corrective feedback, since its presence or absence draws the line between tasks for learning and testing purposes (Loschky & Bley-Vroman, 1993). Hitherto, the facilitative effects of feedback have been largely validated (Li, 2010; Lyster & Saito, 2010; Russell & Spada, 2006). However, there is no consensus as to which type of feedback yields the largest effects. These conflicting findings may be explicated by an insufficient isolation of three components of feedback, i.e., positive evidence, negative evidence, and prompting for error repair, so research must address whether they make separate contributions to L2 grammar development (Leeman, 2002; McDonough, 2005). In the realm of CALI this issue remains unaddressed, and additionally, no study to date has explored the potential of oral (as opposed to written) feedback, thus under-exploiting the potential of computers to simulate authentic communication (Warschauer, 2004).

Finally, the effects of both practice and feedback may be mediated by the type of grammatical form under instruction. However, while several studies have corroborated this, there is little research on how the difficulty of a form exerts its mediating role (DeKeyser, 2007; Ellis, 2007).

Using an innovative e-tutor and a pre-test/post-test/delayed post-test design, this study empirically investigated whether intermediate-level university students of L2 Spanish (N=127) learned more grammar when given opportunities to practice or when given audiovisual exposure to others' practice. Some engaged in computer-simulated conversations with pre-filmed real people, whereas others observed videos of those interactions. The second variable under investigation was the type of corrective feedback provided, which varied in terms of the explicitness of the negative evidence (whether the error was merely spotted or explained) and the requests for error repair (whether or not students were prompted to reformulate their ill-formed productions). The effects of agency and feedback were investigated on two different grammatical forms varying in complexity, Spanish prepositional relative clauses (easier) and present subjunctive (harder). An exit questionnaire was administered to tap into participants' perceptions of the learning experience. Results showed that practice had beneficial effects over exposure to it only under more demanding conditions (less explicit feedback) and for one of the targeted structures (prepositional relative clauses). In terms of feedback, prompting yielded no significant effects, while the explicitness of negative evidence did, with more explicit feedback yielding better results for both structures. However, these effects were more evident for the easier form, prepositional relative clauses. In terms of learners' perceptions, performers were significantly more satisfied than observers, and explicit feedback generated opposite reactions. From a theoretical perspective, these findings contribute to the growing SLA literature on the effects of practice and feedback, while from a pedagogical perspective they prove that computerized oral feedback can work, and help to elucidate precisely when computers may have an edge over non-interactive audiovisual media.