ABSTRACT In spite of its checkered intellectual history, and in spite of the myriad proposals of alternative models that claim to account for the broad range of human behavior and to dispense with the need for selection above the organism level, a multilevel selection framework remains the only coherent means of accounting for the persistence and spread of behavioral inclinations which, at least upon first appearance at low frequency, would have been biologically altruistic. This argument is advanced on three tracks: through a review of experimental and observational evidence inconsistent with a narrow version of rational choice theory, through a critique of models or explanations purporting to account for prosocial behavior through other means, and via elaboration of the mechanisms, plausibility, and intellectual history of group selection.