Although there has been substantial research interest in adolescent children of alcoholics, the generalizability and consistency of previous findings have been limited by the specialized nature of the research samples (e.g., treatment samples, school-based samples). However, attempting to address this problem by recruiting subjects from community sources raises a variety of sampling issues, including the comparability and accessibility of subjects from different sources. This article compares the recruitment rates and characteristics of families enrolled from three sources--DUI records, HMO records and telephone surveys. Implications for research on children of alcoholics are discussed.