In much of the world, the issue of accountability and measurement of educational outcomes is highly controversial. Exit testing is part of the movement to ascertain what students have learned and hold institutions and teachers to account. However, compared to the large number of teacher washback studies, learner washback research is lacking (Cheng, 2008). By exploring standardized exit test washback in Taiwan, this study aims to test Alderson and Wall's hypotheses (1993) regarding learner washback variability, namely that a test will influence 1) degree/depth of learning, 2) attitudes toward methods of learning, and 3) some learners but not others. This study is one of very few washback studies that compare learner behaviors before and after the introduction of exit tests, thereby obtaining a clearer picture of washback effects on learners. Questionnaires and scores on two exit tests (GEPT and TOEIC) were collected from 589 Taiwanese university students before and after the introduction of English exit tests in Taiwan. This study highlights the differential effects of exit requirements for different groups of learners depending on their years of study, proficiency levels, and viewpoints of tests.