An Investigation Into Fuzzy Scoring Methods For Speaking Tests

This paper, following a review of related studies on speaking tests, finds that raters, rating scales and candidates'performance (speaking tasks) are the three essential elements that constitute the rating process in the speaking tests. In addition, scoring methods for speaking tests can be defined as the way in which raters apply rating scales to the candidates'performance and subsequently make evaluations.According to the three stages ("pre-evaluation","evaluation"and"post-evaluation") of the speaking tests, scoring methods for speaking tests are classified as (1) holistic marking and analytic marking, (2) subjective marking and objective marking, and (3) accurate marking and fuzzy marking. Also, based on an analysis of various scoring methods for speaking tests, and on an integration of language testing theories with the fuzzy control theory, the paper comes up with three fuzzy scoring methods, all of which are subjective. They are: (1) holistic marking, (2) inference-based analytic marking, and (3) weighted analytic marking.The three methods are tried out on 34 overseas students who major in Chinese and detailed discussion and analysis are presented in this paper. The results show that firstly, the three methods are all feasible; secondly, the three methods yield scores that exhibit a normal distribution with no statistically significant difference between them; thirdly, a high correlation is found between the scores from the two raters, and the three methods all have a high inter-rater reliability; fourthly, these scores correlate highly with concurrent criteria (speaking test scores), moderately with concurrent criteria (written test scores) and moderately with predictive criteria, and the three methods all have a good criteria-oriented validity.