Proceedings of the Annual Meeting of the Cognitive Science Society

Although many theories of causal cognition are based on causalgraphical models, a key property of such models—the inde-pendence relations stipulated by the Markov condition—isroutinely violated by human reasoners. Two accounts of whypeople violate independence are formalized and subjected toexperimental test. Subjects’ inferences were more consistentwith a dual prototype model in which people favor networkstates in which variables are all present or all absent than aleaky gate model in which information is transmitted throughnetwork nodes when it should normatively be blocked. Thearticle concludes with a call for theories of causal cognitionthat rest on foundations that are faithful to the kinds of causalinferences people actually draw.