Exploring the influence of suprasegmental features of speech on rater judgements of intelligibility

Abstract

The importance of suprasegmental features of speech to pronunciation proficiency is well known, yet limited research has been undertaken to identify how raters attend to suprasegmental features in the English-language speaking test encounter. Currently, such features appear to be underrepresented in language learning frameworks and are not always satisfactorily incorporated into the analytical rating scales that are used by major language testing organisations. This thesis explores the influence of lexical stress, rhythm and intonation on rater decision making in order to provide insight into their proper place in rating scales and frameworks. Data were collected from 30 raters, half of whom were experienced professional raters and half of whom lacked rater training and a background in language learning or teaching. The raters were initially asked to score 12 test taker performances using a 9-point intelligibility scale. The performances were taken from the long turn of Cambridge English Main Suite exams and were selected on the basis of the inclusion of a range of notable suprasegmental features. Following scoring, the raters took part in a stimulated recall procedure to report the features that influenced their decisions. The resulting scores were quantitatively analysed using many-facet Rasch measurement analysis. Transcriptions of the verbal reports were analysed using qualitative methods. Finally, an integrated analysis of the quantitative and qualitative data was undertaken to develop a series of suprasegmental rating scale descriptors. The results showed that experienced raters do appear to attend to specific suprasegmental features in a reliable way, and that their decisions have a great deal in common with the way non-experienced raters regard such features. This indicates that stress, rhythm, and intonation may be somewhat underrepresented on current speaking proficiency scales and frameworks. The study concludes with the presentation of a series of suprasegmental rating scale descriptors.