PURPOSE. To determine the magnitude of students' knowledge gain in a problem-based surgery clerkship and to identify the relationship of this gain to measures of clinical performance. METHOD. Third-year students in a problem-based surgery clerkship at the University of Kentucky College of Medicine in 1990–91 were evaluated by a comprehensive set of objective measures: surgery subject examinations of the National Board of Medical Examiners (NBME) given as pretest and posttest to assess knowledge gain, two multiple-choice quizzes, a modified-essay examination, a standardized-patient examination, and an objective structured clinical examination (OSCE). The students were also evaluated by faculty tutors and preceptors and by their peers. NBME data were available for 66 students, and data were available on the other measures for 42 students. Statistical analysis involved two-way analysis of variance, single group t-test, Pearson correlations, and partial correlations. RESULTS. The students' knowledge gain was statistically significant. The posttest mean score did not differ significantly from the national candidate mean of 500. All but two of the other knowledge and performance measures (the preceptor and tutor evaluations) correlated significantly with the knowledge gain score. The highest correlations were for peer evaluations, the standardized-patient examination, and the modified-essay examination. The overall reliability of the eight measures of student knowledge and performance was .81; deleting any measure, except the preceptor evaluation, lowered the reliability below the benchmark of .80. CONCLUSION. The results suggest that a highly significant knowledge gain occurred during the problem-based clerkship and that this gain in knowledge was closely related to improved clinical performance.