Philosophers of science have recently argued that the structure of scientific networks matters to scientific progress. But where do these networks come from? In this talk, I discuss how social norms of collaboration and resource division might influence scientists' choices about who to collaborate with. In particular, drawing on joint work with Hannah Rubin, I show how discriminatory norms might lead to homophilic networks - those where individuals tend to collaborate with those like themselves. I discuss the fit of these models to related empirical work, possible interventions, and implications for science.