The literature on the evolution of religion has been divided by a fundamental debate between adaptation theories, which explain religious traits as products of selection for religion, and byproduct theories, which explain religious traits as products of selection for other, non-religious functions. Recently, however, a new position has emerged in this debate, as an influential new theory based on cultural selection claims to integrate adaptation theories with byproduct theories, yielding a single, unified account. I argue that the proponents of this (...) view do not say enough about how integration is actually supposed to work, from a logical point of view. Basic questions arise from the assumptions required for unifying these apparently conflicting approaches, which the authors of the account do not address. In response to these questions, I provide a model of the religious phenotype, the Goldberg Exaptation Model, which shows that adaptation and byproduct theories are consistent, and explains how they are positively related, over and above mere consistency. On this view, the religious phenotype is best understood on analogy with a Rube Goldberg device: it is assembled by selection for religion, but using parts designed by selection for other, non-religious functions. (shrink)