This study examined the hypothesis that item overlap, or measurement confounding, accounts for the correlation between temperament and behavior problem symptoms in children. First, a conceptual approach was taken in which 41 experts rated temperament (Children's Behavior Questionnaire, CBQ) and behavior problem symptom items (Preschool Behavior Questionnaire, PBQ) for their fit to both constructs. With this approach, 10% of temperament and 38% of symptom items were confounded. Second, an empirical approach was taken and CBQ and PBQ items were factor analyzed with data from a multi-informant longitudinal study of 451 children. Using this method, 9% of temperament and 23% of symptom items were confounded. Most importantly, removing the confounded items from the CBQ and PBQ scales did not affect the relation between temperament and symptoms, suggesting that the associations were not due to measurement confounding. In addition, the predictive power of earlier temperament for DSM-IV symptoms (Health and Behavior Questionnaire) remained high with the purified CBQ scale. The findings of this study contribute to the understanding of the relation between normal-range temperament and extreme behavior.