Purpose: The Knowledge Based Inference Tool (KBIT DDx), a computerized tutor for diagnosing chest pain, has improved student performance, but the relative contributions of its components - (a) descriptions of the prototypical signs or symptoms of each disease that can cause the presenting complaint, (b) displays contrasting the prototypes of each disease with its closest competitors, and (c) example cases - are not known. Methods: Eighty-six students, participating individually, in small groups, or in a classroom setting, were randomized to study one of five booklets of information about six chest pain diagnoses. Three of these booklets contained information from the KBIT DDx tutor, with case examples: (1) prototypes of each diagnosis, (2) contrasts between confusable pairs of diagnoses, or (3) both prototypes and contrasts. Control booklets were: (4) textbook descriptions of the diagnoses and (5) the same text with case examples. Results: There were no differences among the different KBIT DDx booklets in the increased proportion of cases that participants correctly diagnosed (prototype: 25%; contrast: 25%; combination: 27%). For the textbook control groups, those who saw case examples improved more (20%) than those who only read the text (5%). Overall, the KBIT DDx groups learned more than the control groups, although they did not learn significantly more than the "text plus case examples" group. Conclusions: The KBIT DDx tutor's prototype and contrast presentations pro-mote learning of chest pain diagnosis equally and may be substitutable. Case examples promote learning, whether supplementing KBIT DDx materials or extracts from conventional textbooks.