The surveys were administered to groups of inmates during researchers' visits to the correctional institutions from January to March 2001. The final sample included 388 participants from 11 prisons in Kentucky, 300 inmates from 8 correctional facilities in Tennessee, and 366 inmates from 11 institutions in Ohio. The dependent variable was measured through four questions that assessed drug abuse in prison. The independent variables were measured at the inmate and prison levels. Variables relevant to the importation theory--which attributes in-prison behavior to the importation of preprison experiences, behaviors, and attitudes--included prior use and sale of illegal drugs on the street. Variables derived from differential association, which attributes drug-related behavior inside prison to interaction with other inmates, included the number of rule-violating friends inside prison. A number of contextual variables were measured, with prison crowding being the most relevant to drug use. The findings supported a multilevel model of drug use in prison that includes an element from an inmate's past (prior street drug use) that interacts with an institutional feature (prison crowding) to predict drug abuse inside prison. Variables from differential association theory also influenced drug-related activities in prison. Drug-related behavior was positively predicted by association with friends who were convicted or could have been convicted of a rule infraction during the past year. Also, negative evaluations of the prison rules were predictive of drug-related behavior in prison. Implications are drawn for research and policy. 4 tables and 43 references