Experiments have become a popular method to study altruism and cooperation in laboratory and, more recently, in field settings. However, few studies have examined whether behavior in experiments tells us anything about behavior in the "real world." To investigate the external validity of several common experimental economics games, we compare game behavior with prosocial behavior among Tsimane forager-horticulturalists of lowland Bolivia. We find that food-sharing patterns, social visitation, beer production and consumption, labor participation, and contributions to a feast are not robustly correlated with levels of giving in the economics games. Payoff structure and socioecological context may be more important in predicting prosocial behavior in a wide variety of domains than stable personality traits. We argue that future experimental methods should be tailored to specific research questions, show reduced anonymity, and incorporate repeat measures under a variety of conditions to inform and redirect ethnographic study and build scientific theory.