The development of DSM-III through DSM-5 has relied heavily on expert consensus. In this essay, we provide an historical and critical perspective on this process. Over the last 40 years, medicine has struggled to find appropriate methods for summarizing research results and making clinical recommendations. When such recommendations are issued by authorized organizations, they can have widespread influence (i.e. DSM-III and its successors). In the 1970s, expert consensus conferences, led by the NIH, reviewed research about controversial medical issues and successfully disseminated results. However, these consensus conferences struggled with aggregating the complex available evidence. In the 1990s, the rise of evidence-based medicine cast doubt on the reliability of expert consensus. Since then, medicine has increasingly relied on systematic reviews, as developed by the evidence-based medicine movement, and advocated for their early incorporation in expert consensus efforts. With the partial exception of DSM-IV, such systematic evidence-based reviews have not been consistently integrated into the development of the DSMs, leaving their development out of step with the larger medical field. Like the recommendations made for the NIH consensus conferences, we argue that the DSM process should be modified to require systematic evidence-based reviews before Work Groups make their assessments. Our suggestions – which would require leadership and additional resources to set standards for appropriate evidence hierarchies, carry out systematic reviews, and upgrade the group process – should improve the objectivity of the DSM, increase the validity of its results, and improve the reception of any changes in nosology.