Abstract

This exploratory study pursues the construct of interactional oral fluency during peer interaction by (a) investigating raters' perceptions (verbal protocol) and comparing them with those of individual performance, (b) examining the rated scores (Likert-scale) in two performance conditions using empirically-based scales that were newly developed based on the verbal protocol data, and (c) conducting correlation and regression analyses with the rated scores and the temporal aspects of speeches (pruned and unpruned speech rates). By employing competence-based and performance-based measurements, the study also explores conceptual gaps regarding oral fluency between Second Language Acquisition and Language Testing research. The participants were 56 Japanese university-level learners of English and four native English speakers who rated the learners' oral fluency. The verbal protocol data showed that temporal aspects (e.g., pauses) and interaction-specific features (e.g., turn-taking) were interwoven, indicating that joint performance between the interactants is a constituent of the construct of interactional oral fluency. The statistical analyses revealed that learners' individual performance does not necessarily predict how they will perform in an interactional context, challenging the assumption that L2 specific processing is stably manifested in performance regardless of context. (C) 2014 Elsevier Ltd. All rights reserved.