A computer-assisted learning (CAL) package was developed on non-verbal communication. Its effectiveness was evaluated by comparing learning based on use of the package with that based on a didactic lecture covering the same topic. A class of 151 first-year medical students was divided into two groups, balanced for gender and home/overseas students. One group was asked to use the CAL package, the other group attended the lecture. Knowledge was assessed one week later by a written test, and reactions to using the CAL package were obtained via a questionnaire. Each group was then allowed and encouraged to use the other resource and then asked about their preferences for type of resource at the end of term. Mean score on the knowledge test was reliably better in the CAL group. In addition, scores increased as the time spent using the CAL package increased: this relationship was highly significant. Use of the CAL package varied from 15 to 120 minutes (median 45). Users reported that it was easy to operate, was an adequate or good resource for learning about the subject, and was a good or reasonable use of their time. After using both types of learning resource half the students judged the CAL package more useful for learning about the subject, and half preferred it to the lecture (the other half had the opposite judgement and preference). This study provides evidence that a CAL package can effectively substitute for traditional didactic teaching in a medical school. Good quality CAL, however, requires substantial resources and high calibre staff to develop and maintain it.