Kinship underlies costly cooperation in Mosuo villages.

The relative importance of social evolution theories such as kin selection, direct reciprocity and need-based transfers in explaining real-world cooperation is the source of much debate. Previous field studies of cooperation in human communities have revealed variability in the extent to which each of these theories explains human sociality in different contexts. We conducted multivariate social network analyses predicting costly cooperation-labouring on another household's farm-in 128 082 dyads of Mosuo farming households in southwest China. Through information-theoretic model selection, we tested the roles played by genealogical relatedness, affinal relationships (including reproductive partners), reciprocity, relative need, wealth, household size, spatial proximity and gift-giving in an economic game. The best-fitting model included all factors, along with interactions between relatedness and (i) reciprocity, (ii) need, (iii) the presence of own children in another household and (iv) proximity. Our results show how a real-world form of cooperation was driven by kinship. Households tended to help kin in need (but not needy non-kin) and travel further to help spatially distant relatives. Households were more likely to establish reciprocal relationships with distant relatives and non-kin but closer kin cooperated regardless of reciprocity. These patterns of kin-driven cooperation show the importance of inclusive fitness in understanding human social behaviour.