Effective physician-patient interaction is an important part of surgical practice. This study had three goals: (1) to measure the interpersonal skills (IS) of surgical students and residents in structured clinical settings and to determine the reliability of such measurements; (2) to determine the relationship of IS to clinical performance; and (3) to determine the impact of level of training on IS. Twenty third-year medical students and 30 junior-level interns (23 PGY-1s, 7 PGY-2s) interacted with nine actual or simulated patients as part of an objective structured clinical examination. Using a global rating scale, faculty graded both the IS and the organizational skills of the candidates. A two-way analysis of variance indicated significant differences among the three groups of trainees (P = 0.0002) and among the IS scores for each of the nine patient problems (P < 0.0001). Both the PGY-2s and the medical students exhibited significantly better interpersonal skills than did the PGY-1s. The IS scores correlated significantly with the data gathering scores, the data interpretation scores, and the organizational scores. We conclude that faculty measurement of IS is moderately reliable even when a simple global rating scale is used. Overall IS scores were rather poor, particularly in the PGY-1 group. IS were highly correlated with overall objective clinical performance.