We examined the contribution of genetic and environmental influences on the stability of withdrawn behavior (WB) in childhood using a longitudinal multiple rater twin design. Maternal and paternal ratings on the withdrawn subscale of the Child Behavior Checklist (CBCL) were obtained from 14,889 families when the twins were 3, 7, 10 and 12 years old. A longitudinal psychometric model was fitted to the data and the fit of transmission and common factor models were evaluated for each variance component. WB showed considerable stability throughout childhood, with correlation coefficients ranging from about .30 for the 9-year time interval to .65 for shorter time intervals. Individual differences in WB as observed by the mother and the father were found to be largely influenced by genetic effects at all four time points, in both boys (50-66%) and girls (38-64%). Shared environmental influences explained a small to modest proportion (0-24%) of the variance at all ages and were slightly more pronounced in girls. Non-shared environmental influences were of moderate importance to the variance and slightly increased with age, from 22-28% at age 3 to 35-41% at age 12 years. The stability of WB was largely explained by genetic effects, accounting for 74% of stability in boys and 65% in girls. Shared environmental effects explained 7% (boys) and 17% (girls) of the behavioral stability. Most shared environmental effects were common to both raters, suggesting little influence of rater bias in the assessment of WB. The shared environmental effects common to both raters were best described by a common factor model, indicating that these effects are stable and persistent throughout childhood. Non-shared environmental effects accounted for the remaining covariance over time.