Parental smoking data have been abstracted from the interview records of the case-control study that first indicated that pregnancy radiographs are a cause of childhood cancer (Oxford Survey of Childhood Cancers, deaths from 1953 to 1955). Reported smoking habits for the parents of 1549 children who died from cancer were compared with similar information for the parents of 1549 healthy controls (matched pairs analysis). There was a statistically significant positive trend between paternal daily consumption of tobacco and the risk of childhood cancer (P< 0.001). This association could not be explained by maternal smoking, social class, paternal or maternal age at the birth of the survey child, sibship position or obstetric radiography. About 15% of all childhood cancers in this series could be attributable to paternal smoking.