A Survey of the Issue of Generalizability in Task-based Language Assessment

Abstract

The use of authentic tasks in Task-Based Language Assessment seems to facilitate the extrapolation of assessment performance to the real-life situation. The question arises as to how effective Task-Based Language Assessment is in predicting test-takers’ performance in the real-life situation based on their task performance in the assessment settings and what proposals there are to improve the generalizability of Task-Based Language Assessment. The study draws upon Bachman and Palmer (1996) and Douglas’s (2000) frameworks in order to identify the fixed and varied characteristics of assessment tasks and conditions in the Target Language Use (TLU) situation. These frameworks function as a template by presenting an organized list of the items to compare and contrast the assessment tasks with the real-life target tasks. The present study proposes standardizing the fixed conditions present in both the assessment and the TLU situation, such as setting and equipment. However, since the standardization of the varied conditions could undermine the extrapolation inferences at the cost of improving the reliability of assessment, the study concludes that the assessment procedure should standardize a set of short tasks or observation conditions that are representative of the TLU domain rather than just one specific type of task or condition.