The social learning strategies tournament was an open computer-based tournament investigating the best way to learn in a changing environment. Here we present an analysis of the impact of memory on the ability of strategies entered into the social learning strategies tournament (Rendell, Boyd, et al., 2010) to modify their own behavior to suit a changing environment. The tournament showed that a strategy's ability to remember the past and to predict the future were both key to its success. The possibility that a strategy needs to engage in an approximation of 'mental time travel' to succeed in the tournament strongly implies that investment in randomly timed social learning is not enough to guarantee success. A strategy must use social learning strategically with reference to both predicted future environmental states and past environmental states. We examine the two most successful strategies (DiscountMachine and Intergeneration) in terms of their use of memory and discuss the impact of their complex memory use on their ability to time learning moves strategically and track environmental change. The tournament suggests that the human capacity for mental time travel may have improved the efficiency of social learning and allowed humans to invest in more sophisticated social learning than is seen elsewhere in the animal kingdom. (Contains 2 figures and 1 table.)