Generalizability of scores across different test tasks in a high stakes language proficiency test

The current study uses Bachman’s (1990) test method facet framework as an organizing framework to study the effect of task type on test takers’ performance on a high stake language proficiency test. Analysis of the performance of a subsample of examinees who sat for Iranian university entrance exam (UEE) in 2014, showed no significant effect for single effect of task type. While it was shown that the greatest variance was due to two way interaction of person and task type. Moreover a perfect reliability estimate also was observed. The study provides evidence that test is free from test task bias and enjoys from a high degree of reliability.