Sociologists have shown little interest in friendship. The rather fragmentary empirical work which does exist on friendship in modern society paradoxically tends to confirm the impression that the subject is of marginal importance; for friendship emerges from much research as a relationship which provides emotional support and small services but little else. Admittedly, there is also evidence indicating links with more classical sociological themes—such as access to jobs. But this evidence is often seen as representing ‘exceptional’ circumstances. Thus, among an immigrant group, friendship ties may be seen as particular to the culture from which immigrants come; among an elite group, the importance of friendship ties may be seen as peculiar to that elite circle. We can talk of a genuine ‘paradigm’ of modern society as being based on relatively impersonal relations. The paper maintains that if friendship is approached differently, it emerges as of more central and ‘structural’ interest. What is required is a shift on both theoretical and methodological levels. Sociology needs to shift its theoretical perspective and recognize that people articulate their values, enact their strategies, develop their language in face-to-face groups, not as isolated individuals. In terms of empirical research, what is needed are methods capable of tracing these groupings. This is difficult at present because most sociological methods rely heavily on data collected on individuals and their attributes.