<p>Network models are pervading ecology, genomics, epidemiology, as well as social science and neurosciences. In this paper, I survey the uses of network analysis in biological and social sciences, distinguishing between various jobs they can fulfill. I explore the various methods based on network analysis. I argue that in many cases they are indeed explanations, such explanations being “topological explanations”, which in turn is a subkind of structural explanations. I defend this view against the idea that network analysis is rather a description than an explanation, and against the claim that it is a kind of mechanistic explanation based on a focus on organization and hence very abstract. In the reasoning, I emphasize the difference between interaction and correlation networks, argue that both of them play roles in explanations, highlight their equivalence as graphs, and acknowledge that many of the debates raised by network analysis stem from the use of both families of networks, notwithstanding their differences. While ‘network analysis’ is not a natural kind, the explanatory potential it holds should be recognized, together with its potential for discovery in the study of complex systems.</p>