Causal reasoning is a critical part of everyday cognition. We ask how people reason about causes when faced with inconsistent sources of knowledge. Causal models arise from multiple sources of information regarding their constituent parameters. Knowledge sources may be inconsistent both within parameters (when one source says a variable should appear often and another says it should appear rarely), and between parameters (when dependencies among parameters result in an internally inconsistent causal model). We provide a normative model for resolving both these sources of conflict. An experiment found that our model of belief integration predicted the qualitative pattern of adults causal inferences under uncertainty.