Self-efficacy is an important measure in science education as it is predictive of persistence and success in science, technology, engineering, and mathematics (STEM) courses and is an influential factor in students' decisions to major in STEM fields. It is unclear what effect active teaching strategies have on students' self-efficacy, which is typically measured with a pretest at the beginning of the semester and a post-test at the end of the semester. To better understand what happens to self-efficacy over the course of an actively taught physics class, in addition to the typical pretest and post-test, we used a reflective pretest. At the end of the semester, we asked students to reflect on their abilities at the beginning of the semester and we compared this "reflective" self-efficacy to both their presemester and postsemester self-efficacy. We found that students' reflective self-efficacy was systematically lower than their self-efficacy at the beginning of the semester. Interviews reveal that discrepancies between presemester self-efficacy and reflective self-efficacy are the result of response-shift bias. Because of students' limited experience with active learning environments, response-shift bias makes it difficult to accurately measure students' change in self-efficacy over the semester of an actively taught physics course. We conclude that reflective pretests in combination with interviews can help educators and researchers understand if changes in self-efficacy are being masked by response-shift bias.