The relationship between exposure to family change, exposure to parental discord during the period from birth to 10 years and risks of offending by the age of 13 years was studied in a birth cohort of New Zealand children. This analysis showed that while exposure to parental discord during middle and early childhood led to increased risks of early offending, exposure to family change in the absence of parental discord did not lead to increased risks of offending. The results also suggested that children with a history of early conduct problems were particularly susceptible to parental discord but that the effects of discord did not vary with the child's gender. These results persisted when errors of measurement in the reporting of offending were taken into account using latent class methods.