The reliability and the factor structure of the Yoruba translation of the General Health Questionnaire (GHQ) were investigated using data collected during a study of psychiatric morbidity among a sample of 787 primary care attendees. The instrument was shown to have a high degree of internal consistency. A two-factor solution provided the best description of the GHQ in this setting. Using subsamples selected on the basis of sex and age, these two factors were shown to be stable. Also, using data collected on a second-stage sample of 214 patients who were given a structured clinical interview, both factors discriminated "cases" from "non-cases" and one of the factors discriminated between depressed from non-depressed "cases".