Enabling interaction: A study of examiner behaviour in an interactive speaking test

Trinity's Richard Harris described an investigation of the interactive task in GESE in his recent presentation at the 5th Future of English Language Teaching Conference.

Reliability is a primary concern for high-stakes speaking assessments though this can be at the expense of authenticity. Nevertheless, authentic interaction and representation of the complete construct in speaking assessments remains an ongoing aspiration.

This workshop described an investigation of the Interactive Task in Trinity College London’s Graded Examinations in Spoken English (GESE). This task transfers responsibility to the candidate to take control of their interaction with the examiner. The workshop described empirical research into the practices examiners use so that candidates take and keep responsibility for the interaction.

Using transcripts from the Trinity-Lancaster Corpus, grounded theory methodology was employed to identify patterns of examiner behaviour in Interactive Task performances. A range of examiner behaviour, or strategies, were identified, such as ‘accepting’, ‘rejecting’, ‘parsimony’ and ‘steering’. The data yielded 886 examples of performance that followed these strategies. Nine indices of candidate performance were selected, including measures of fluency, lexical sophistication, grammatical accuracy and syntactical complexity. Using automatic text analysis and statistical methods, the study then quantitatively compared the indices of candidate performance that had been elicited by each examiner strategy. The small and medium effect sizes discovered in the differences between some strategies indicate the authenticity achieved through the freedom afforded examiners to interact spontaneously does not bias performance.

However, there are behaviours that examiners might be advised to avoid in order to exhibit ‘bias for best’ in interaction. The workshop will close with an exploration of some of the strategies in more detail and a discussion of how the study findings can inform the way interaction is perceived and managed in a test environment.