Research on nuclear families has reported support for an hypothesized inverse relationship between power and role-taking accuracy. Family members who are in subordinate (dependent) power positions are better able to predict attitudes and behavior of other family members than are those in relatively more powerful and independent positions. This hypothesis was tested within the context of the extended family and was seen as a significant shift in focus, given the alleged role reversal between adult children and their parents during the latter part of the family life cycle. Six persons representing three generations in each of 47 extended family units responded to a mailed questionnaire. Findings provided only equivocal support for the hypothesis. Support was found when the independent variable was operationalized in terms of perceived power or perceived physical dependency and in family dyads where the role-taker was a female sibling, a father, a mother, or a grandmother. For dyads where the role-taker was a male sibling or a grandfather, the relationship between power and role-taking accuracy was opposite that hypothesized. (Although data on which this study is based was obtained in 1972, the topic is of high current interest). (Author/HLM)