This study validates the GHQ-28 on a sample of paralysed, spinally injured out-patients under the care of the West Midlands Spinal Injuries Unit. The validity of the GHQ-28 was 0.83 (95% confidence interval: 0.70 to 0.93) using the Clinical Interview Schedule (CIS) as the validating criterion. The effectiveness of the GHQ-28 as a screening instrument using CIS 11/12 as case criterion, and thus a prevalence rate of 0.18, was judged by relative operating characteristic (ROC) analysis. The area under the ROC curve was 0.91 +/- 0.03. 'Optimum' discrimination occurred near GHQ 3/4 (0011 scoring scheme) giving a sensitivity of 0.81, a specificity of 0.82 and a misclassification rate of 18%.