This paper examines the conceptual and empirical adequacy of the Eurogang Network's survey measurement of gang membership. Using data from a nationally representative survey of young people in England and Wales, we employed a latent class analysis to model variation in the characteristics of peer groups. We found that while Eurogang survey items identified a distinct group of young people involved in more frequent and serious offending, this definition also extended to a separate group whose only 'vice' was recreational drug use. We discuss the conceptual validity of extending the 'gang' label to this latter group, together with the pressing need for more developmentally sensitive measures of peer networks in adolescence. © The Author(s) 2013.