The purpose of this study was to investigate the washback effects of the General English Proficiency Test (GEPT) on English teaching in two applied foreign language departments in Taiwan. One had prescribed its GEPT requirement to its day-division students whereas the other had not. Overall, the GEPT did not induce a high level of washback on teaching in either department. Only courses which were linked to the departmental GEPT policy and whose objectives were to prepare students for the test were significantly affected. Results suggest that micro-level contextual factors (for example, the objectives of the course) and teacher factors had a greater impact on teachers' instruction. Finally, on the basis of current understandings of washback, I propose a new, tentative model to portray the washback of tests on teaching. (Contains 1 figure, 1 table and 15 footnotes.)