People often act out of a desire to be responsible for good and not for bad events. Similarly, people frequently reward and punish other people if they perceive them to be responsible for the implementation of events that they like or dislike. When the implementation of an event depends on the interaction of multiple persons and, potentially, moves of nature, the determinants of such responsibility perceptions are not well understood. In this paper, I propose a notion of causal responsibility which attempts to objectively capture the causal importance of a person’s action for the implementation of an event in such situations. A laboratory experiment shows that the notion successfully predicts people’s responsibility perceptions. Furthermore, I incorporate the notion in a framework of responsibility preferences and study its implications for worker motivation and the design of voting rules. Finally, I show that the notion can explain experimentally elicited behavior and punishment and reward patterns in multi-agent situations that are not well-explained by existing social preference theories.