We examined the hypothesis that low plasma cholesterol concentration was associated with minor psychiatric disturbance in a cross-sectional study of 410 male and 138 female civil servants. Psychiatric disturbance was measured using the 30-item General Health Questionnaire (GHQ). Mean GHQ score did not vary significantly with quartile of total cholesterol concentration. After adjustment for confounding variables a significant trend of increasing mean GHQ with increasing cholesterol concentration emerged. In a logistic regression analysis subjects in the highest quartile of cholesterol concentration had an adjusted odds ratio for being a "psychiatric case" of 2.0 (95% confidence interval, 1.1-2.5) compared with those in the lowest quartile. This relationship reversed when using a higher cutoff point to define more severe cases, although the trend was not statistically significant.