The purpose of this study was to provide data concerning the relationship between features of residency training and a test of cognitive achievement gathered at the end of residency. To accomplish this, data collected in the late 1970s by three national organizations were joined and analyzed with the aid of experts in internal medicine. Although graduate medical education has evolved since this information was gathered, it does provide a baseline for assessing the impact of changes on the cognitive skills of residents. The findings suggest that better program performance on the examination is associated with attracting more knowledgeable residents to begin with and that programs are able to maintain the advantage of their residents throughout training. Moreover, program characteristics have an impact on the cognitive skills of residents over and above what would be predicted by test scores at the end of medical school. Programs with better examination performance tend to provide residents an extensive, well-supervised educational experience stressing ambulatory care.