• A partial validation of the contextual validity of the Centre listening test in Japan

      Yanagawa, Kozo (University of BedfordshireUniversity of Bedfordshire, 2012-07)
      The purpose of this study was to validate the listening comprehension component of the Centre Test in Japan (henceforth, JNCTL) in relation to contextual parameters and cognitive processing. For the purpose of this study, a comprehensive framework of contextual parameters and a L2 listening processing model was established. This provided a solid theoretical framework for this study, whereby empirical evidence was elicited in relation to contextual parameters and cognitive processing. The elicitation was made through document analysis, focus group interviews, and a large-scale questionnaire administered to stakeholders including 110 high school English teachers and 391 third year students of high schools. The elicited data was subjected to descriptive, quantitative and qualitative analysis. The results of Preliminary studies identified ten possible key parameters to help the JNCTL achieve greater validity. They included the number of opportunities to listen to the input, a lack of hesitations, a lack of overlapping turns, a lack of multi-participant discussions, a lack of variety in the English accents used, a lack of L2 speakers, a lack of inference questions, a lack of non-linear texts, a lack of sandhi-variations, and a lack of natural speech rate. The results of the questionnaire revealed that sandhi-variation was the key parameter to help the current JNCTL achieve greater validity in a direction that would be accepted by the stakeholders, and it was further explored in Main Study in attempt to investigate the effect of sandhi-variation on listening comprehension test performance and the level of cognitive load imposed on the test takers. A series of experiments was conducted involving the manipulation of sandhi-variation. The results revealed that although no statistical difference was found in item difficulty estimates between the sandhi-variation and non-sandhi-variation versions, sandhi-variation may involve double effects on listening comprehension for the test takers. The positive effects could involve providing more prominent phonological difference between accented and unaccented words in connected speech which are produced by sandhi-variation, and this difference may reduce the cognitive load imposed on the test takers. The negative effects may involve increasing the cognitive load imposed on the test takers by obscuring sounds through elision or unclear pronunciation, and disturbing speech perception or word recognition. Recommendations are provided for improving the validity of the current JNCTL and for the development of listening comprehension tests more generally. Implications are also suggested for the teaching of listening at secondary schools in Japan. Lastly, the limitations of the study are outlined and suggestions for further research are proposed.