This chapter illustrates the representational nature of causal understanding of the world and examines its implications for causal learning. The vastness of the search space of causal relations, given the representational aspect of the problem, implies that powerful constraints are essential for arriving at adaptive causal relations. The chapter reviews (1) why causal invariance—the sameness of how a causal mechanism operates across contexts—is an essential constraint for causal learning in intuitive reasoning, (2) a psychological causal-learning theory that assumes causal invariance as a defeasible default, (3) some ways in which the computational role of causal invariance in causal learning can become obscured, and (4) the roles of causal invariance as a general aspiration, a default assumption, a criterion for hypothesis revision, and a domain-specific description. The chapter also reviews a puzzling discrepancy in the human and non-human causal and associative learning literatures and offers a potential explanation.