Abstract
A theoretical model designed to estimate the associations of variables of interest with the trait component of maternal and teacher ratings of conduct disorder is outlined. This model was applied to data collected in a birth cohort of New Zeland children. The analysis showed marked differences between the correlations for various measures and observed conduct disorder scores and the corresponding correlations with the estimated trait components of these scores. In some cases correlations increased as a result of correction for measurement error whereas in other cases correlations declined as a result of correction for the effects of method factors.