Parental smoking data have been reabstracted from the interview records of the Oxford Survey of Childhood Cancers (deaths from 1971 to 1976). Reported smoking habits for the parents of 2587 children who died with cancer were compared with similar information for the parents of 2587 healthy controls (matched pairs analysis). Maternal daily consumption of cigarettes and paternal use of pipes or cigars were unimportant, but there was a statistically significant positive trend between paternal daily consumption of cigarettes and the risk of childhood cancer (P < 0.001). This association could not be explained by maternal smoking, social class, parental ages at the birth of the survey child, sibship position or obstetric radiography. Relations between maternal consumption of cigarettes and birth weights suggested that (maternal) smoking data were equally reliable for case and control subjects. About 14% of all childhood cancers in this series could be attributable to paternal smoking. These data were combined with smoking data from two previously published reports from the Oxford Survey (deaths from 1953 to 1955, deaths from 1977 to 1981) to obtain further information on risks for different types of cancer and different ages at onset of disease. Paternal cigarette smoking emerged as a potential risk factor both for the generality of childhood cancer and for all ages at onset.