In the investigation of the relationship between childhood adversity and adult psychopathology, scales that require retrospective reports of parental style are often employed. Controversy exists as to whether these scales reflect the actual behavior of the parents concerned or reflect the perceptions of the child. One such scale, the Parental Bonding Instrument, was completed by a volunteer sample of 672 twins. For females, agreement between twins was generally high and a shared environment model fitted all scales except maternal care. This is consistent with the view that the ratings reflect the behavior of the parents. Agreement between male twins was substantially lower and well-fitting models did not include a shared environment component. It is argued that the scales may be affected by competition between twins or comparisons made by the respondent with his or her co-twin.