The study evaluated the factorial structure of the 12-item General Health Questionnaire (GHQ-12) in a population of Kenyan adults and adolescents. A sample of 1216 people aged 12–60 years completed the English version of the GHQ-12. To evaluate the factor structure of the GHQ-12, a confirmatory analysis using MPLUS was carried out. A unidimensional model of GHQ-12 as originally conceptualized did not provide a good fit for the data (Comparative Fit Index = 0.843, Tuckler Lewis Index = 0.841, Root Mean Square of Error Approximation = 0.078, Standardized Root Mean Square Residual = 0.063). Unidimensional models, which partial out the effects of negative wording showed an adequate fit to the data. Additionally, multidimensional models showed an excellent fit to the data. Based on the principles of parsimony, pattern of item loading and the correlation between identified factors in the multidimensional models, it was concluded that the GHQ−12 is a unidimensional measure whose structure is significantly influenced by wording effects. The GHQ-12, when applied to a literate population in Africa, showed similar factor structure as in other regions of the world; hence, it can be recommended for use as a screening tool for general symptoms of psychological distress. Copyright © 2011 John Wiley & Sons, Ltd.