Widespread cooperation is a defining feature of human societies from hunter-gatherers to nation states, but explaining its evolution remains a challenge. While positive assortment –of cooperators with cooperators - is recognized as a basic requirement for the evolution of cooperation the mechanisms governing it are debated. Moreover, the social structure of modern hunter-gatherers, characterized by high mobility, residential mixing and low genetic relatedness, undermine assortment and add to the puzzle of how cooperation evolved.3 Here, we analyze four years of data (2010, 2013, 2014, 2016) tracking residence and levels of cooperation elicited from a public goods game (PG), in a population of current hunter-gatherers – the Hadza of Tanzania. Data were collected from 56 camps, comprising 383 unique individuals, 137 of whom we have data for two or more years. Despite significant residential mixing, we observe a robust pattern of assortment necessary for cooperation to evolve: In every year, Hadza camps exhibit high between-camp and low within-camp variation in cooperation. We further consider the role of homophily in generating this assortment. We find little evidence that cooperative behavior within individuals is stable over time or that similarity in cooperation between dyads predicts their future cohabitation. Both sets of findings are inconsistent with homophilic models that assume stable cooperative and selfish types (e.g. partner choice). Consistent with social norms, culture and reciprocity theories, the data suggest that the strongest predictor of an individual’s level of cooperation in any given year is the mean cooperation of their campmates in that year. These findings underscore the adaptive nature of human cooperation – particularly its responsiveness to social contexts – as features important in generating the assortment necessary for cooperation to evolve.