Nvivo in qualitative language testing research: Testing spoken interaction

This paper reports on a mapping technique to find evidence in paired candidate discourse for what is said by candidates and raters in verbal protocols being used to validate components of the spoken interaction construct underpinning a rating scale for speaking. Using N vivo to map the results from three previous studies onto each other, the aim is to investigate the possibility of demonstrating the correspondence between observable features in the candidates' speech sample and the features of 'engagement' that language experts claim to attend to while observing paired interaction which candidates also report on after performing a test. The data in this preliminary study includes the speech sample of 2 beginner dyads in a Spanish foreign language context, 2 Spanish teacher raters Verbal Protocols and Stimulated Verbal Recall from 2 candidates. The analysis of the speech sample is guided by features that have been already identified in previous studies which lead to the findings of the study that show what the speech sample and the protocols have in common. Studies on paired tasks in oral proficiency tests have largely focused on what affects candidate output and /or test outcomes; this study, however, validates the crucial mediating position between the output and outcomes. It is argued that at the point of intersection of what raters attend to and what candidates are monitoring lies an empirical basis for qualitative construct validation of the actual output which can be accessed with Nvivo.