The construct of perceived L2 speaking proficiency in a paired testing format

Abstract

The main aim of the research was to investigate the construct validity of L2 speaking proficiency in a paired testing format. 100 secondary school students and university students were recruited for the study (m=24, f=76), with a mean age of M=18.67. The par-ticipants were tested in pairs (compare-and-contrast task) by two independent raters on four trait categories: accuracy (A), fluency (F), word choice (WC) and turn taking (TT), on a pre-specified 6-point scale. The inter-rater reliability was computed using two approaches (Pearson’s Product Moment correlation and Cronbach’s alpha), and they both indicated the ratings were reliable. The average scores on the scale and subscales also showed high values of Cronbach’s alpha, which confirmed the reliability of the measurement instruments as well. In order to examine the construct of the participants’ speaking performance, we performed exploratory and confirmatory factor analyses. The final model, with very good fit indices, confirmed the initial hypothesis: the construct of L2 oral proficiency is comprised of all four assumed trait factors; however, the strength of correlations rather suggests that there are two facets of the said construct (linguistic: A, F, WC and sociolinguistic: TT).