This article focuses on two sets of literature that have developed out of a shared concern with networks: the network governance school, which has been engaged in a set of macro-level questions about the extent to which networks are changing the nature of state–society relations; and the policy network analysis school, which has focused on the relationship between processes of interest intermediation and their impact on policy-making outcomes. We examine how each school is underpinned by important epistemological differences between positivist, interpretivist and critical realist approaches. We argue that these differences complicate and make contestable what would otherwise seem to be an intuitively attractive argument in favour of combining these two schools. In seeking to understand better how these two schools might be combined, we adopt a critical realist approach and make a distinction between vertical coordination on the state–society axis and horizontal coordination on the interest integration axis. This produces a typology of governance arrangements, which are evaluated according to the level of input and output legitimacy that they are likely to generate, two criteria that are taken as overarching measures of how governance outcomes vary between different governance arrangements. This provides the basis for a broader discussion of how these outcomes are conditioned by both a network's structural characteristics and the way in which it is managed.