A criterion-referenced approach was used to examine the validity of a set of 15 Patient Management Problems (PMPs) representing a broad range of medical problems. Results of performance of 175 medical students indicated that a new problem-solving index that includes an open-ended differential diagnosis section had greater validity than the more traditional proficiency index, as the spread between the distributions of masters and nonmasters was greater for the problemsolving index. Implications for medical education and research are discussed.