Interaction in a paired oral assessment: Revisiting the effect of proficiency

Abstract

Paired oral assessments have gained increasing popularity as a method of assessing speaking skills (East, 2015; Galaczi, 2014). Several advantages have been associated with this method, including practicality and authenticity (Taylor, 2003). Nevertheless, concerns have also been raised in terms of the interlocutor effect in paired speaking tests, particularly in regard to the interlocutor’s oral proficiency (e.g., Norton, 2005). The present study reports on an approximate replication of Davis (2009), who looked at the effect of interlocutor proficiency on paired speaking assessments. The current study compared the oral performance of 24 university students in two different pairing conditions: once paired with a partner of the same proficiency level and once with a partner of a different proficiency level. Results of this replication study partially confirmed Davis’s (2009) results. There were only minimal differences in test-takers’ scores between both conditions. A multi-facet Rasch analysis confirmed these results indicating that the pairing conditions were equivalent in difficulty. There were, however, observable differences in the quantity of talk depending on the interlocutor’s proficiency. Unlike Davis (2009), this study found that low-proficiency test-takers produced fewer words when paired with high-proficiency partners. Even though the number of words produced by test takers was affected by their partner’s proficiency, their performance scores remained constant.