Abstract: The application of network perspectives to communities requires some appreciation of the variety of ways people are now writing about communities. Some scholars and practitioners have drifted toward the view that a community is composed very largely of the personal networks of the individuals who are members of the community. But the whole community is more than the sum of those related parts, and the structure of a community must include not only those direct interpersonal relations but also the relations among the clusters and groups and corporate entities that interact in and about this whole. If scientific knowledge about these matters is to accumulate, we must be able to compare findings among various studies. From the 1940s well into the 1960s the local community was the recognized social unit that sociologists and anthropologists studied. Linton wrote of the necessity of the local group. Many sociologists and anthropologists gave their full attention to this local level of social integration through a field called "community studies." The work of Conrad Arensberg, Sol Kimball, Carl Taylor, Robert Redfield, and others had views of communities that had a network cast to them. The category "community" includes a wide range of social formations, generally local systems of fairly densely connected persons in households and organizations, systems on a scale somewhere between those domestic households and the wider society -- state or nation. Recently scholarly focus has shifted to individuals and their personal networks, with less attention to the social structures in which they are embedded. The concept "community" really needs to be defined because it is used in many situations where what it means has real consequences. A network perspective suggests a whole complex social system organized in levels, from a household/family level, upward through a hierarchy of levels, to the national (nation-state) and even beyond that to a supranational (above-state) level. Within that complete social system are embedded several levels of networks that could be called community. The networks that form such a community must be sorted out from the entire complex system. A network perspective on communities - or on structures relating to communities - includes seeing groups both as networks of the individuals composing them and as nodes related to each other through their common members. Such "affiliation networks" are more complex than just the sum of the personal networks. Improved techniques of data collection and data analysis are bringing us closer to sorting subsystems in ways that permit their proper comparison. Gaps or seams between segments or clusters, structural equivalence, structural holes, regular equivalence, role analysis, and block modeling are improving our understanding of systems at all levels. We are approaching an understanding of the concept of community in network perspective.