Perlormance-based examinations that use standardized-patient (SP) cases have been used to assess the clinical competence of senior medical students in seven classes (1986- 1992) at Southern Illinois University School of Medicine. Because of the length of the examinations, two or more (multiple) SPs have been used to simulate about half of the cases in each examination. An earlier study found little or no effect of using multiple SPs on the intercase reliability of checklist scores, written scores, and total scores. The present study was conducted to assess the effect of multiple SPs on the intercase reliability of SP ratings of interpersonal and communication skills. Although the ratings were not used to determine total scores or to make pass - fail decisions, it seems reasonable to expeJt that multiple SPs might have a greater impact on ratings on non-cognitive dimensions than on ostensibly more objective checklist and written scores. In general, the results showed that generalizability coefficients free of measurement error due to multiple SPs were larger than those reflecting measurement error due to multiple SPs by an average of about 5 percentage points, suggesting that the use of multiple SP raters on the same case reduced reliability by about 5 percentage points. Compared to earlier findings, the effect of multiple SPs on the reliability of ratings was about 4 or 5 times greater than their effect on checklist, written, or total scores. Nevertheless, the magnitude of the effect on ratings was not excessive.