Twelve four-person female groups o f subjects displaying the typical underestimation o f their peers' (relative to their own) riskThe weight of relevant research indicates that the group risky shift is caused by the operation, through discussion, of a pre-existing bias toward risk acceptance. The characteristics of this 'value', and how it operates, remain controversial questions, and the present study is intended to provide some clarification.Various types of evidence have been submitted to show that risk acceptance is * Conducted a t the Sonderforschungsbethank Wiltraud Paule for assisting in ninreich fur Sozial-und Wirtschaftspsycholoning subjects and Jorg Graff for his help in gische Entscheidungsforschung, Universitat the computer analysis of the data. Dr. Paul Mannheim, with financial aid from the Gebhard and Jiirg Graff authored the comDeutsche Forschungsgemeinschaft. We puter programs used.

A review of evidence from a number of experiments suggests that as a result of discussion the average of the group members' positions (attitudes, judgments) becomes more extreme. These shifts seem to occur primarily when group members' initial positions are distributed across the two sides of the scale in such a way as to create a majority-minority constellation. Accordingly, it is proposed that underlying the extremity-shifts reported in the literature is the movement of a minority toward the majority's side. That such minority-change is nor the only source of extremization is suggested by one study finding that group discussion enhances the extremity of individual positions. An analysis of the special distribution of positions existing on the items selected in that experiment permits the following conclusion. Discussion-induced extremization is (also) caused by the impact of arguments in the discussion favoring the side which most or all members preferred to begin with. Other explanations cannot be discarded at the present time. At the least they point to important aspects of small-group functioning which it would seem fruitful to investigate.