Validity coefficients of the 30-item, 28-item and 12-item versions of the General Health Questionnaire (GHQ) were determined by comparison with the Present State Examination (PSE) in a sample of 200 17-year-olds. The PSE classified 7 people (3.5%) as cases, although only 47% were identified as free of symptoms. Misclassification rates, sensitivity and specificity values are presented for different cutting scores for the three versions of the GHQ. The GHQ-28 had superior values, especially with a cutting score of 5/6; the GHQ-12 with a 2/3 cutting score also had acceptable values. All versions of the GHQ correlated highly with the PSE Index of Definition and total scores, providing support for the treatment of GHQ scores as a continuous variable in this kind of population. Correlations between sub-scales of the GHQ-28 give further evidence for a general factor and the relative independence of the social dysfunction sub-scale.