This study used the theory of power proposed by Mintzberg (1983) to pursue two goals: a) to identify the best types of power configurations that represent the power relations in a research organization; b) to investigate whichever relations between some variables and the power perception that employees have of their organization. The research was carried out in the Empresa Brasileira de Pesquisa Agropecuária - Embrapa, Brazil, using a stratified random sample of 1013 employees. It were used the Organizational Power Configuration Scale (Paz, 1996) to measure five power configurations (dependent variables) and a short version of the Gordon et al. (1980) instrument to measure union commitment (one of the independent variables of the research). There were also another ten itens related to biographical and organizational employee data (independent variables). Descriptive and inferential statistics were used in the data analysis. The results of this research showed that missionary and meritocracy seemed to be the best configurations to explain the power relations in the surveyed organization. Two organizational variables - time of employment and held position - were found to be the best predictors of these two power configurations. Despite the exploratory nature of this study, it was observed that the typology proposed by Mintzberg (1983) proved to be useful for describing and explaining the power phenomena in the organizations.