抄録
This research describes a part of the development process of a task-based writing performance assessment for Japanese learners of English. In the previous study, two assessment tasks to elicit the underlying constructs and rating scales were developed. The pre-testing was conducted on the participants comprised of 15 university students and five experienced teacher raters, who provided scores for two sets of 15 texts. The correlational analysis showed that the teacher raters could produce relatively high scores with each rating scale The results of FACETS analysis, however, suggested that rating scales for task 1 (accuracy) and 2 (communicability) should be recommended to be optimized. In this research, we explored how rating scales can be developed and improved based on the, results of the pre-testing. First, we optimized the rating scales on the basis of the criteria that Linacre (2002) proposed. Then, the scoring guide was revised in each section. As a result, the new rating scales are comprised of clearer descriptions of each construct and of 5-point Likert scales. The descriptors of each category are also provided with written samples as an explanatory part of the scale in order to give raters a shared understanding of the constructs and to make them more self-consistent in scoring.
