We assessed the performance of students with a self-reported conflict between their religious belief and the theory of evolution in two sections of a large introductory biology course (N = 373 students). Student performance was measured through pretest and posttest evolution essays and multiple-choice (MC) questions (evolution-related and non-evolution-related questions) on the final exam and posttest. The two class sections differed only in exam format: MC with or without constructed-response (CR) questions. Although students with a reported conflict scored significantly lower on the final exam in the MC-only section, they scored equally well in the MC+CR section, and all students in the MC+CR section performed significantly better overall. As a result, (1) a religious conflict with evolution can be negatively associated with student achievement in introductory biology, but (2) assessment with constructed response was associated with a closed performance gap between students with and without a conflict. We suggest that differences in exam format and focus on student acceptance of evolution (either evidence-based or opinion), rather than reported conflict, may contribute to the inconsistencies in student learning of evolution across research studies, and that CR questions may help students overcome other obstacles to learning evolution.