Decision-making processes and their outcomes were investigated in six consensus development conferences at the National Institutes of Health in which panels of experts evaluated new medical technologies. One hundred seventy-seven self-administered questionnaires were obtained from participants in these conferences. Questionnaire data were analyzed along with data derived from content analyses of the six consensus statements (CS) produced by the conferences. Results of these analyses provide considerable support for the hypotheses that the quality of the outcome (i.e., the CS) is determined by the existence of an interaction process, a decision procedure, and a chairperson, which facilitate the exchange of relevant information. Strong disagreements among the panelists appear to inhibit such exchange and harm the quality of the CS. Personal satisfaction appears to be more strongly related to the quality of the process and of the information disseminated than to the quality of the outcome. A clear relation was found between the panelists' status and expertise, their participation in the process, and their contribution to the CS. The pattern of these findings is quite similar to that obtained in laboratory studies. The role of preconference organizational factors, such as the selection of conference questions, panel, and speakers, and the characteristics of the technology are discussed.