The accuracy of standardized patient clinical problem presentation was evaluated by videotape rating of a random sample of 839 student-patient encounters, representing 88 patients, 27 cases and two university test sites. Patient-student encounters were sampled from a collaborative inter-university final-year clinical examination of fourth-year medical students which was conducted at the University of Manitoba and Southern Illinois University in 1987 and 1988. The accuracy, replicability and portability of standardized patient cases were evaluated. The average accuracy of patient presentation was 90.2% in 1987 and 93.4% in 1988. Perfect accuracy scores were obtained by 15 patients; however, 11 patients had average scores below 80% with the accuracy of presentation in some encounters being as low as 30%. There were significant differences in the accuracy score achieved by patients trained together for the same case in 6 of 35 possible comparisons. There was also a systematic trend for patients trained at Southern Illinois to be more accurate in their presentation than patients trained at the University of Manitoba. These differences were significant in 5 of the 15 cases used in the examination.