Does social isolation in adulthood predict survival because socially isolated individuals are already unhealthy due to adversity earlier in life ("health selection")? Or do adult social environments directly cause poor health and increased mortality risk ("social causation")? These alternative hypotheses are difficult to disentangle in humans because prospective data on survival and the environment for both early life and adulthood are rarely available. Using data from the baboon population of Amboseli, Kenya, a model for human behavior and aging, we show that early adversity and adult social isolation contribute independently to reduced adult survival, in support of both health selection and social causation. Further, strong social bonds and high social status can buffer some negative effects of early adversity on survival. These results support a growing change in perspective, away from "either-or" hypotheses and towards a multi-causal perspective that points to multiple opportunities to mitigate the effects of social adversity.