Background: More consistent differences have been found between students educated with problem-based learning (PBL) and those taught in traditional methods in clerkship performance than on examinations. Purposes: To determine if change to a PBL method in a single course influences students' performance on specific types of examinations compared to courses taught to the same students in a lecture-based format. Methods: From 1994 to 1995, a large component of the 2nd-year program was changed from lecture based to PBL. The rest of the course material remained in standard lecture format. A standard, year-end exam and a research exam based on clerkship skills were administered for both subjects in 1994 and 1995. Results: The performance of the 1995 class on the standard exam was significantly lower than that of the 1994 class on all subjects, regardless of teaching method (p <. 05). On the segment the research exam designed to mimic clerkship performance, the 1995 students did equally well on the PBL-taught material as the 1994 students. On the non-PBL material, they did significantly worse (p <. 05). Conclusions: In comparison to courses taught in stable didactic format, we conclude that PBL may confer a selective advantage on exam questions that relate to clerkship performance. We suggest that the reason for the specific advantage on these kinds of exam questions relates to the concept of elaborated versus dispersed knowledge.