The Thought–Problem scale (TP) of the CBCL assesses symptoms such as hallucinations and strange thoughts/behaviors and has been associated with other behavioral disorders. This study uses parental reports to examine the etiology of variation in TP, about which relatively little is known, in 7-year-old twins. Parental ratings on TP were collected in 8,962 7-year-old twin pairs. Because the distribution of TP scores was highly skewed scores were categorized into 3 classes. The data were analyzed under a threshold liability model with genetic structural equation modeling. Ratings from both parents were simultaneously analyzed to determine the rater agreement phenotype (or common phenotype [TPc]) and the rater specific phenotype [TPs] that represents rater disagreement caused by rater bias, measurement error and/or a unique view of the parents on the child's behavior. Scores on the TP-scale varied as a function of rater (fathers rated fewer problems), sex (boys scored higher) and zygosity (DZ twins scored higher). The TPc explained 67% of the total variance in the parental ratings. Variation in TPc was influenced mainly by the children's genotype (76%). Variance in TPs also showed a contribution of genetic factors (maternal reports: 61%, paternal reports: 65%), indicating that TPs does not only represent rater bias. Shared environmental influences were only found in the TPs. No sex differences in genetic architecture were observed. These results indicate an important contribution of genetic factors to thought problems in children as young as 7 years.