Summary. A consistent finding in the literature on measures of clinical problem-solving scores is that there are very low correlations across different problems. This phenomenon is commonly labelled ‘content-specificity’, implying that the scores differ because the content knowledge necessary to solve the problems differs. The present study tests this hypothesis by presenting groups of residents and clinical clerks with a series of simulated patient problems in which content was systematically varied. Each subject also completed a multiple choice test with questions linked to each diagnosis presented in the clinical problems. Three of the four problem-solving scores showed low correlations, even to two presentations of the same problem, and no relationship to content differences. None of the scores were related to performance on the multiple choice test. The results suggest that variability in problem-solving scores is related to factors other than content knowledge, and several possibilities are discussed.