A Study On The Difficulty Of Writing Test Tasks--From A SLA Perspective

This MA thesis, based on a cognitive perspective developed in SLA research, reports an empirical study into the difficulty of writing test tasks, which attempts to clarify: 1) whether the hypothesized task difficulty set by the framework of Skehan 1998 can predict the performance of the examinees and 2) whether the students' own perception of task difficulty can be relied on and used to predict the actual test performance.The thesis starts to review the issue of interface between SLA and LT, and relates the findings in SLA to the rationale of the present study of task difficulty in the domain of LT. It cites the researches done after Bachmann 1989 to illustrate how such interface can bring perspectives as well as methods to the study of the wide range of issues relevant to both fields. The emphasis is laid on LT to show how SLA insights can be borrowed in the study of LT issues and lead to changes in LT.Introducing TBLA (Task-based Language Assessment) in LT is one of the major changes, which are closely related to SLA research insights. On the findings of psycholinguistics and cognitive science, SLA researcher hold that in task-based language instruction, if the attention of learners can be effectively directed in the accomplishing of the 'tasks', the goals of acquiring communicative skills and language forms can be both achieved. Their fruitful researches on task features and task difficulty make task-based instruction an effective and practical approach in L2 teaching. To LT, these research findings have great implication on the issue of developing TBLA. LT researchers, who are not content with relying on language tests with multiple-choice and other discrete items to infer the learners' language performance, turn to do researches on TBLA and employed SLA insights on taskfeatures and task difficulty to study such properties of test task in test contexts. Following this trend of LT, the present study incorporates Skehan's (1998) framework of cognitive demand of task with different features to study the difficulty of semi-direct writing test tasks in the context of classroom test.Skehan 1998 proposed that the more cognitively demanding tasks generate greater level of task difficulty, and more complex tasks direct learners' attention to context and divert attention away from form so that simple tasks generate more fluent and more accurate language output, as opposed to more complex tasks which generate more complex output at the expense of accuracy and fluency. Skehan 1998 also comments on the present ability model in LT and suggests ways in how to operationalize ability model in testing practice.The results of the present study show that more cognitively demanding tasks do generate greater level of task difficulty and thus elicit poorer examinee performance. However, it does not find enough evidence that there is some systematic competition between accuracy, fluency and complexity in the performance. The relationships between the examinees' own perceptions of task difficulty and their actual test performance turn out to be more complex and multidimensional. It is held that the examinees' own perception of difficulty is variable and can not be relied on, though it is found the examinees seem to have some reasonable consciousness of the criteria about the accuracy and fluency of the language in their performance.