Male homosexual preference (MHP) has long been of interest to scholars studying the evolution of human sexuality. Indeed, MHP is partially heritable, induces a reproductive cost and is common. MHP has thus been considered a Darwinian paradox. Several questions arise when MHP is considered in an evolutionary context. At what point did MHP appear in the human evolutionary history? Is MHP present in all human groups? How has MHP evolved, given that MHP is a reproductively costly trait? These questions were addressed here, using data from the anthropological and archaeological literature. Our detailed analysis of the available data challenges the common view of MHP being a “virtually universal” trait present in humans since prehistory. The conditions under which it is possible to affirm that MHP was present in past societies are discussed. Furthermore, using anthropological reports, the presence or absence of MHP was documented for 107 societies, allowing us to conclude that evidence of the absence of MHP is available for some societies. A recent evolutionary hypothesis has argued that social stratification together with hypergyny (the hypergyny hypothesis) are necessary conditions for the evolution of MHP. Here, the link between the level of stratification and the probability of observing MHP was tested using an unprecedented large dataset. Furthermore, the test was performed for the first time by controlling for the phylogenetic non-independence between societies. A positive relationship was observed between the level of social stratification and the probability of observing MHP, supporting the hypergyny hypothesis.