PURPOSE. To compare results from 1981 and 1992 national surveys of the writers of medical school dean's letters, and to rate the 1992 letters based on the guidelines recommended by the Association of American Medical Colleges (AAMC). METHOD. In early 1992 a survey was administered to the dean's letter writers at all 125 U.S. medical schools with fully accredited four-year programs; the survey incorporated many items from a 1981 survey. In addition, 550 dean's letters from all U.S. medical schools for the graduating class of 1992 were collected and rated based on guidelines published by the AAMC in 1989. RESULTS. The response rate of the 1992 survey was comparable to that of the 1981 survey (85% and 87%, respectively). In both surveys, slightly more than half the schools used more than one letter writer; however, the 1992 letters were longer and there were more of them, in spite of the fact that there were approximately 700 fewer graduates. In 1992 the estimated total cost per school was $25,000 (comparable data were not collected in 1981). The ratings of the letters revealed that only 38% of the schools introduced their letters as letters of “evaluation,” as recommended by the AAMC, and that 15% of the schools failed to use the AAMC guidelines for format. When the schools were rated for overall quality (i.e., format combined with information about the students' performance in comparison with that of peers), 55% of the schools passed and 45% failed. CONCLUSION. Several recommendations for improving dean's letters are discussed, including the following: (1) all dean's letters should be formatted according to the AAMC guidelines; (2) each school should have one person responsible for central overview of the school's letters; and (3) for comparative performance information, schools should at least give the percentages of grades given in the required clerkships, and it would be preferable for them to employ systems that group students into four to six groups and to indicate the percentage of students in each group.