Objectives: This cross sectional study aimed to investigate the relative effectiveness of a problem based learning curriculum (PBL) and a traditional curriculum (TC) in the development of medical students' diagnostic reasoning skills. Methods: Junior and senior clinical students (n = 431) at a single clinical school of the University of Melbourne self-administered Bordage et al.'s Diagnostic Thinking Inventory (DTI) to assess diagnostic reasoning skills during transition from a TC to a PBL curriculum. Mean scores for the flexibility in thinking and memory structure sub-scales were compared for students at junior and senior clinical levels under the two curricula using multivariate analysis of variance (MANOVA). Results: An effect of curriculum, Wilks' Lambda = .95, F (2, 426) = 11.67, p < .001, p2 =.05 and level of trainee, Wilks' Lambda = .86, F (2, 426) = 35.00, p <.001, p2 = .14 was evident on DTI scores overall. Senior students on average had higher flexibility in thinking (86.7 vs. 80.9) and memory structure scores (84.8 vs. 77.9) than junior students. Students in the TC had higher average flexibility in thinking (84.4 vs. 83.0) and memory structure scores (82.6 vs. 79.9) than students in the PBL curriculum. Conclusions: These results suggest both a developmental effect and a curriculum effect in the development of diagnostic reasoning skills. Contrary to expectations, a TC appears to favour the development of diagnostic reasoning skills compared with a PBL curriculum. This unexpected finding highlights to educators the importance of monitoring student attributes during curriculum change.