TOEFL is widely used as a certificating device, and is strongly claimed accountable by the people who utilize it. However, there are indications of the vulnerability of TOEFL to the teaching of test taking strategies. The purpose of the present study is to provide empirical evidence that coaching toward TOEFL may invalidate its results as indicators of test-takers' proficiency levels. To test the hypothesis, the scores of 75 subjects on a TOEFL and on a task-based test of language proficiency were compared. The subjects were selected from among those who attended coaching classes for TOEFL. The findings of this study revealed that coaching leads to spurious scores and consequently to unaccountability of TOEFL results. Theoretically, it is claimed that teaching toward task-based tests would not be harmful to pedagogy and learning, because performance on such tests are manifestation of the performance of test takers on real life tasks that they will be expected to perform. This contends that testing practitioners should move toward the use of more' authentic' and performance-based assessments to cause the least possible damage to teaching, learning, and decision-making processes. Also the implications of this study concerning the safety of coaching toward a particular task, domain, or subdomain in order to enhance students' achievement on the one hand, and test validity on the other, are discussed.
)