A technique in which people are asked to think how they could behave in a certain role without acting it out. For example, they could be asked to make forecasts “assuming that you were the President of the U.S.” Taking roles can affect forecasts. Cyert, March and Starbuck (1961) divided subjects into two groups of 16 each. Each group was assigned a different role. Subjects were given the role as the chief cost analyst for a manufacturing concern and were asked to produce cost forecasts on the basis of preliminary estimates provided by two assistants in whom they had equal confidence. Other subjects were given the role as the chief market analyst and were also asked to provide sales forecasts. The data were identical for both roles. Seldom did the analysts simply average the estimates from their two assistants – the expected behavior if no role had been assigned. The cost analysts forecasted on the high side, and the market analysts forecasted on the low side. Roles can also affect the acceptability of forecasts as shown by Wagenaar and Keren (1986). The subjects’ acceptance of information depended upon the roles they were assigned. In this study of 388 subjects, the two roles were either individual decision maker (“a parent”) or a societal decision maker (“minister of traffic”). Half of the subjects were given each role and were provided with either anecdotal or statistical evidence on the need for safety belts in the back seats of automobiles. The anecdotal evidence was a three-sentence description of a traffic accident in which a seven-year old girl died because she did not wear a seat belt. The statistical evidence was two sentences stating that 150 children die each year in motor vehicle accidents and that this could be reduced to 50 if seat belts were used in the back seats. The societal decision makers were more influenced by the statistical evidence (62% favoring the use of seat belts) than were the individual decision-makers (47% favoring the use of seat belts).