Left to their own devices: exploring interactional practices in an online group speaking task

Abstract

While Interactional Competence (IC) has been a facet of models of competence for as long as communicative competence has existed (Hymes, 1974), a psycholinguistic approach (Van Moere, 2012) has tended to dominate speaking testing, particularly online speaking test delivery (Roever & Dai, 2021). The psycholinguistic approach views speaking abilities as residing within the individual. Online tests are often monological and poorly reflect interactional abilities (Roever & Ikeda, 2021). In contrast, a sociolinguistic-interactional view emphasises speaking as a contextualised, social act. This perspective “relocates speaking abilities in the interaction between people” (Van Compernolle, 2021, p. 193). This emphasis on interaction creates a problem however: how to measure individual contributions to a shared action (McNamara, 1997). This dilemma is seen in the rather abstract, detached nature of assessment criteria which could be enhanced via detailed Conversation Analytic (CA) interactional insights (Kley et al., 2021; Youn, 2023), specifically how orderliness pervades talk (Pekarek Doehler, 2021a, 2021b, 2018; Hall, 2023, 2018; Thompson & Couper-Kuhlen, 2005). This mixed-methods study aimed to investigate interactional practices used in an online, triadic, scenario-based task involving 56 candidates of varying language proficiency (below-B2, B2, and C1). Expert raters assessed candidates, and their ratings were analysed using Many-facet Rasch Measurement (MFRM) to account for rater tendencies. Qualitative analysis produced a glossary of 12 turn-taking and 10 repair practices, based on established CA studies (Wong & Waring, 2021). CA-derived coding, Mann Whitney mean comparisons, and Spearman Rho correlations were used to explore the relationship between interactional practices and proficiency level. The findings show quantitative differences in certain practices between proficiency levels (e.g., preliminary moves as a B2 and C1 differentiator), alongside general qualitative differences in how these practices were used across proficiency levels. The findings raise interesting questions as to how technology can be better leveraged to assess IC in future and how we might gain a better level of granularity in assessing the “observable artefact” (Waring, 2018) of IC. Several suggestions are made regarding future exploration of test discourse using a CA-derived coding approach to provide constructive engagement with scale development, rater training, task design and empirical elaboration of the CEFR for online interaction.