A theory of causal processing is proposed. The theory is based on the notion that people understand the causal connection as a generative relation representing the operation of a causal power of some thing under some appropriate releasing condition. Much causal processing is automatic and involves the perception of apparent generative relations with the aid of this basic concept and more particular beliefs about the causal powers of the things involved. Controlled causal processing occurs when automatic causal processing cannot and when its occurrence is in the interest of some practical concern. This also involves identifying some causal power or releasing condition that is already believed to be a possible cause of the effect in question. The origins and development of this way of understanding causation are briefly discussed. The preferred method of testing ideas about possible causes involves gathering more information about the occasion in question, not sampling other occasions, as in covariation-based models. Evidence supporting this contention is reported. Application of this basic theory in particular domains of knowledge is briefly discussed.