Paired and Group Oral Assessment

The assessment of second language (L2) speaking has long been an important yet challenging area of research in language testing. L2 testers are often concerned with designing authentic speaking tasks that resemble real-life speaking activities so that score interpretations are generalizable to non-test contexts. The conversational nature of speaking skills has promoted the widespread integration of more authentic and interactive assessment tasks, such as paired or group orals. Such direct test formats typically “involve candidates interacting together to perform a task while one or more examiners observe their performances and rate their language proficiency” (Van Moere, 2013, p. 1). The earliest attempt to incorporate paired interaction traces back to the Foreign Service Institute Tests (Fulcher, 2003), and to Folland and Robertson (1976) who were first to suggest using group discussion in oral assessment (Fulcher, 1996). In recent decades, the paired or group speaking format has been incorporated within a few large-scale high-stakes tests. For instance, group oral assessment has been integrated within the English A/S level Examination in Hong Kong since 1994 (Swain, 2001) and the College English Test-Spoken English Test (CET-SET) in Mainland China since 1999 (He & Dai, 2006). The most influential adoption of the paired speaking format comes from the University of Cambridge Local Examinations Syndicate (UCLES), who first introduced the paired speaking format in the First Certificate of English (FCE) examination in 1996 (Saville & Hargreaves, 1999; Taylor, 2000). Given the increasing popularity of paired and group orals, Foot (1999) lamented that little empirical evidence was available to support the use of such a test format. More specifically, he cast doubt on the quality of test takers’ performances when communication breaks down due to factors, such as anxiety, different accents, proficiency levels, and personality. Additionally, he criticized that the presence of an interlocutor and an assessor “threatens...the illusion of a natural conversation” (p. 39). His criticism is also concerned with the length of the test. Specifically, he contended that given the same length of test time, paired candidates appear to have far less time to allow adequate amount of linguistic output than in one-to-one interviews, thus challenging any inferences drawn from such limited linguistic samples. Although Foot’s (1999) concerns are not completely unwarranted, due at least in part to the scarcity of empirical research available back then, more researchers argued for the potential benefits of using paired and group interactions to assess L2 speaking. For example, it has been suggested that testing candidates in dyads or groups lowers their communicative anxiety and stress (Ikeda, 1998; Norton, 2005; Saville & Hargreaves, 1999), allowing them to demonstrate their best language proficiency and interactive skills. Furthermore, unlike the traditional oral