Volume 6, 2006, Paper 5


Paper 5: Exploring difficulty in Speaking tasks:
an intra-task perspective

Authors
Cyril Weir
University of Bedfordshire, UK

Barry O’Sullivan
Roehampton University, UK

Tomoko Horai
Roehampton University, UK

This study looks at how the difficulty of a speaking task is affected by changes to the time offered for planning, the length of response expected and the amount of scaffolding provided (eg suggestions for content).

Abstract

The oral presentation task has become an established format in high stakes oral testing as examining boards have come to routinely employ them in spoken language tests. This study explores how the difficulty of the Part 2 task (Individual Long Turn) in the IELTS Speaking Test can be manipulated using a framework based on the work of Skehan (1998), while working within the socio-cognitive perspective of test validation. The identification of a set of four equivalent tasks was undertaken in three phases. One of these tasks was left unaltered; the other three were manipulated along three variables: planning time, response time and scaffolded support. In the final phase of the study, 74 language students, at a range of ability levels, performed all four versions of the tasks and completed a brief cognitive processing questionnaire after each performance. The resulting audio files were then rated by two IELTS trained examiners working independently of each other using the current IELTS Speaking criteria. The questionnaire data were analysed in order to establish any differences in cognitive processing when performing the different task versions.
Results from the score data suggest that while the original un-manipulated version tends to result in the highest scores, there are significant differences to be found in the responses of three ability groups to the four tasks, indicating that task difficulty may well be affected differently for test candidates of different ability. These differences were reflected in the findings from the questionnaire analysis. The implications of these findings for teachers, test developers, test validators and researchers are discussed.