Three hundred and eighty one men admitted to hospital for the first time with any psychiatric diagnosis were individually matched for age and year to patients admitted to general hospitals. An occupational history was obtained from 90% of this study group by telephone interview or mail. Exposure to solvents was assessed by three methods, individual rating of each job recorded, application of an exposure matrix based on job title, and assessment of lifetime job histories of selected case-referent pairs. A sample of individual ratings used in the analysis was compared with ratings made by five experts. The panel values tended to be lower but all six sets of ratings correlated well. There was no increased risk of psychiatric illness among subjects exposed to moderate or greater solvent concentrations for at least 10 years (odds ratio (OR) 1.0, 90% confidence interval (90% CI) 0.7-1.4, individual rating; OR 1.1, 90% CI 0.6-2.0, job title matrix; OR 0.9, 90% CI 0.5-1.7, lifetime assessments). At higher exposures the risk was increased--although not to a statistically significant degree--especially for cases with non-psychotic diagnoses (ICD-9 codes 300-316). This negative result, by all three methods of assessment of exposure, contrasted with that from a parallel investigation of cases of organic psychoses and cerebral degeneration.