The purpose of this article is to evaluate the stance of the cognitive science of religion (CSR) with respect to the problem of the definition of religion. Firstly, I defend the necessity of an approximate definition of religion due to the fact that (a) definitions are microtheories and (b) there is considerable social demand for a comprehensive definition of religion because of the inclusion of the concept in the majority of contemporary legal systems. Secondly, I present a representative sample of statements about the nature of religion put forward by scholars working within the cognitive tradition, which reveals considerable convergence on what the CSR thinks religion is about and justifies the concept of a "cognitive definition of religion". Thirdly, in a brief historical sketch, I try to identify two opposite tendencies in historical attempts at defining religion and their respective philosophical backgrounds: Essentialist definitions perpetuate the venerable Western tradition harking back to Plato's Euthyphro, while recent non-essentialist definitions draw on the work of late Wittgenstein (in what I term "power-innocent" social constructionism) and Nietzsche, Foucault and Bourdieu (in what I term "power-based" social constructionism), respectively. Lastly, against the background of an essentialist vs. non-essentialist dialectic, I consider the definition of religion provided by the CSR, which, while prima facie almost indistinguishable from Tylor's doctrine of animism, is based philosophically on Kant and Chomsky (and therefore at odds with the prevalent practice of social constructionism) and capable of providing much more cogent justification for a universalistic approach to religion than any of its essentialist predecessors.