To help determine the role that test instrument formats play in evaluation, two parallel examinations were given to 227 second-year medical students. The tests were based on information presented in a medical case history. One required students to generate their own problem lists (the generate group); the other required the students to select problem lists from a list of alternatives (the select group). All the students had difficulty formulating problem lists as indicated by average overall scores of 42% and 57% correct for the generate and the select groups, respectively. Significant quantitative and qualitative differences were noted between the two groups in that the select group usually picked properly integrated problems while the generate group constructed partially correct answers composed of unintegrated cues. As predicted, the select group scored significantly higher than the group generating their own lists. (The relative utility of generate or select response formats for diagnostic and certifying examinations is discussed). (Author/GDC)