An examination of the diagnostic hypotheses generated by physicians in 119 primary care encounters showed that "correct" diagnostic hypotheses were produced simply on the basis of the patient's chief complaint in 78% of the cases and on the basis of the history in 16%; the physical examination and "stat" laboratory tests contributed little to generating "correct" hypotheses but more to eliminating alternatives. The early generation of frequently accurate diagnostic hypotheses on the basis of very limited information suggests that considerable clinical reasoning takes place in the earliest stages of a problem.