To determine whether the Quick Medical Reference (QMR) program can improve diagnosis or enhance learning among internal medicine residents, we compared the diagnostic accuracy of the program with that of residents at various training levels. The cases were from a prospective convenience sample of 40 patients admitted by 10 first-year residents (interns) and two chief medical residents. Four sets of differential diagnoses were created for each case--the first set by an intern, the second set by a chief resident, and the third and fourth sets by QMR, using the findings of the interns and chief residents, respectively. The diagnostic accuracy of the interns and chief residents was significantly greater than that of QMR. However, the chief residents indicated that QMR did increase their understanding of disease processes and offered educational value.