The article first establishes the conceptual link between power and causality. It then attempts to review the literature on power from the causal modelling perspective. The implications of seeing power as a slope and a constant relationship are developed: the advantages of Harsanyi's measure of power are discussed as well as weaknesses in the pluralists' argumentation. The latter is shown to assume a block-recursive model. The second part of the article documents the extent to which the study of power raises the same methodological problems as causal analysis. It is argued that pluralists, by ruling out misperceptions, by neglecting the impact of the political structure and by taking a “neutral” stand, do in fact make specific assumptions about feedback effects and error terms. It is suggested that the pluralist-elitist debate revolves around the validity of these assumptions.