Many disciplines find it puzzling that costly cooperation exists within groups of non-kin. Cooperation can be sustained when non-cooperators are punished or when cooperators are rewarded, but these sanctions are themselves costly to provide. As such, we must ask: what forces maintain the existence of sanctioning? Why do people possess a psychology that includes punitive sentiment and a willingness to reward? Many theoretical models rely on "second-order punishment", which means that people will punish those who do not punish non-cooperators. However, our review of the evidence suggests that people do not readily do this, and do not particularly like punishers. This calls into question any theories that rely on this second-order punishment. By contrast, people will readily reward those who reward cooperators, which suggests that rewards may function as part of a system of indirect reciprocity where cooperators (and rewarders) are seen as "good" thus worthy of help. So what does sustain punitive sentiment? Punishment may function to signal qualities of the punisher that are otherwise difficult to observe, such as the punisher's trustworthiness or willingness to retaliate against personal affronts. Alternately, punishment may simply be a "Volunteer's Dilemma" where it becomes rational to "volunteer" to punish non-cooperators if no one else in the group will. Finally, we discuss how positive and negative sanctions may function differently to maintain large-scale human cooperation, depending on the proportion of cooperators in a population and the necessity for unanimous cooperation.