Aims. To examine the symptom profile and factorial structure of DSM-IV alcohol dependence. Design. Subjects were interviewed in program facilities by trained interviewers using a standardized questionnaire. The response rate was 95% in Mexico and 90% in the United States. Setting. Two publicly funded inpatient facilities in Mexico and five in the United States. In Mexico the programs were located in Mexico City. In the United States the programs were located in Santa Clara County, California. Participants. Three hundred and ninety-one Mexican and 212 Mexican American men in treatment for alcohol problems in Mexico and the United States. Measurement. Dependence criteria were assessed with the Composite Diagnostic Interview-Substance Abuse Module (CIDI-SAM). Findings. The unidimensional structure of alcohol dependence fits the Mexican American data but not the Mexican data. However, when Mexican clients were divided according to place of interview (treatment program), the unidimensional structure fits one of the groups but not the other. Conclusions. The test of unidimensionality must be seen as inconclusive with regard to the Mexican data. These results highlight the potential influence that client selection methods may have on study results, especially in cross-cultural projects.