Focusing on the social environment as well as the individual should both enhance our understanding of HIV transmission and assist in the development of more effective prevention programs. Networks are an important aspect of drug injectors' social environment. We distinguish between (1) risk networks (the people among whom HIV risk behaviors occur) as vectors of disease transmission, and (2) social networks (the people among whom there are social interactions with a mutual orientation to one another) as generators and disseminators of social influence. These concepts are applied to analyses of data from interviews with drug injectors in two studies. In the first study drug injectors' risk networks converge with their social networks: 70% inject or share syringes with a spouse or sex partner, a running partner, or with friends or others whom they know. Qualitative data from interviews with injectors in the second study also show that the social relationships between drug injectors and members of their risk network are often based on long-standing and multiplex relationships, such as those based on kinship, friendship, marital and sexual ties, and economic activity. In the first study the vast majority of injectors, over 90%, have social ties with non-injectors. Injectors with more frequent social contacts with non-injectors engage in lower levels of injecting risk behavior. Risk settings may function as risk networks: injectors in this study who inject at shooting galleries are more likely than those who do not to rent used syringes, borrow used syringes and inject with strangers. Since the adoption of a network approach is relatively new, a number of issues require further attention. These include: how to utilize social networks among drug injectors to reduce risk through peer pressure; how to promote risk reduction by encouraging ties between injectors and non-injectors; and how to integrate biographical and historical change into understanding network processes. Appropriate methodologies to study drug injectors' networks should be developed, including techniques to reach hidden populations, computer software for managing and analyzing network data bases, and statistical methods for drawing inferences from data gathered through dependent sampling designs.