Science is an example of highly successful institutionalized social learning, leading to rapid cumulative cultural evolution. In this talk, I outline the striking convergence between this view of science and the one developed in science studies. Using data from millions of scientific papers, I illustrate how scientists use social cues to select research problems and how these heuristics lead to more (and less) efficient discovery. I then argue that formal theories of learning and cultural evolution offer new insight on old puzzles in the history and sociology of science.