The article argues that there are two traditions of sociological interest in networks, with very different empirical foci, theoretical ideals and methods. Mainstream "structuralist" network analysis is mainly concerned with recasting descriptions of sociological objects in relational terms. There is also, however, a minority tradition of network analysis concerned mainly with the social ties individuals have in different social contexts. The article argues for the importance of these networks formed in the course of individual trajectories and claims that they provide a basis for descriptions structured not only by traditional sociological categories or individual logic. This contains the potential for radical revision of sociological categories (class, ethnicity, etc.) which has been promised but not really fulfilled in structural network analysis. The article shows how the methods of mainstream network analysis exclude from view the network patterns the alternative tradition is interested in.