A statistical model designed to estimate the contributions of stable and changing symptomatology to levels of minor psychiatric symptoms is developed. This model is fitted to data obtained from three longitudinal studies. These studies involved subjects from Canberra (Australia), Christchurch (New Zealand) and Groningen (Holland). Data from all three data sets were shown to fit the proposed model adequately. However, there were systematic differences in the findings of the study. The findings from the Canberra and Groningen studies suggested that a large amount (50-75%) of the variance in symptom levels could be attributed to between subject difference in stable levels of symptomatology. In contrast the Christchurch study suggested a smaller contribution of stable symptomatology. These differences may be explained by the nature of the samples studied. All three studies showed evidence of strong correlations (0.79-0.94) between stable levels of symptomatology and the measure of trait neuroticism. It is concluded on the basis of this evidence that the neuroticism may be little more than a way of measuring the subject's characteristic level of minor psychiatric symptoms. The model also made it possible to secure estimates of the extent to which measures of neuroticism were contaminated by short-term mental state. Estimates of contamination effects varied between studies. For the Canberra data contamination was negligible, for the Groningen data mild contamination effects were present but for the Christchurch data contamination was larger. These differences may be explained by differences in the nature of the samples studied.