BACKGROUND There is conflicting evidence concerning the magnitude of genetic and shared environmental influences on juvenile antisocial behaviour (AB). The use of more than one assessment of AB may yield more accurate estimates of these influences. METHODS Retrospective reports of antisocial behaviour prior to age 18 were obtained on two occasions from a population-based sample of 3522 adult males from male-male twin pairs: phone interviews (wave 1) and self-report questionnaires obtained 19 months later (wave 2). Structural equation modelling estimated the genetic and environmental influences on reliably-measured AB. Factors related to participation of co-twin at wave 1, attrition between waves 1 and 2, and reliability of wave 1 and wave 2 assessments were also investigated. RESULTS Twin analyses revealed that genetic, shared environmental, and non-shared environmental influences accounted for approximately 33% (95% CI = 9-57%), 31% (95% CI = 10-51%) and 36% (95% CI = 29-44%) of the variance of reliably measured AB, respectively. We also found significant occasion-specific genetic influences on wave 1 AB. Wave 1 AB did not predict wave 1 participation of co-twin or attrition, but was related to reliability. Co-twins of MZ twins and younger twins were more likely to participate at wave 1; attrition was predicted by being a DZ twin, lack of initial participation of co-twin, fewer years of education, and fewer children. Being older, being unmarried, and having less psychopathology were associated with greater reliability. CONCLUSIONS When measurement error is taken into account, both genetic and shared environmental factors are significant influences on juvenile AB, accounting for approximately one-third of variation. The origin of the specific genetic influences on wave 1 AB is unclear, but may be due to factors related to measurement.