Abstract: Few outstanding questions in the human behavioral sciences are timelier or more urgently debated than the evolutionary source of religious behaviors and beliefs. Byproduct theorists locate the origins of religion in evolved cognitive defaults and transmission biases. Others have argued that cultural evolutionary processes integrated non-adaptive cognitive byproducts into coherent networks of supernatural beliefs and ritual that encouraged in-group cooperativeness, while adaptationist models assert that the cognitive and behavioral foundations of religion have been selected for at more basic levels. Here, we survey these differing approaches, noting their respective strengths and weaknesses. We then advance a novel model that centers on the ability of language to generate alternative worlds independent of immediate empirical facts, and thus highlight the similarities between religious belief and the modes of cognition that underlie institutions in general. The institutional cognition model of religion accounts for some of the shortcomings of extant approaches and draws attention to the human ability to create non-empirical worlds; that is, worlds that are imaginary. Both religious beliefs and institutional facts—such as jurisdictional borders—are non-empirical assertions, yet they are socially accepted as truths and reified through ritual and behavior. One type of non-empirical, linguistically generated belief—supernatural agent belief—is particularly effective for stabilizing systems of arbitrary norms by rooting them in deontic rather than utilitarian reasoning. The evolutionary roots and continued persistence of religion are thus functions of the capacity for humans to generate cognitive alternatives to empirical reality, and the need to stably coordinate those alternative conceptions. Read Full Article →