This population-based study (the largest on this issue conducted in Southern Europe) has examined mortality among the parents of 2622 children diagnosed with cancer in Piedmont during 1967-1994. Parents were followed up from the date of the index child's birth until the end of 2000, yielding a total of 118 090.7 person-years of observation. Standardized mortality ratios (SMRs) were estimated using mortality rates for the whole population of Piedmont as the reference. Among mothers, total mortality was similar to that expected [SMR 1.02, 95% confidence interval (CI) 0.85-1.23, 117 cases]. A reduced risk of mortality was seen in fathers (SMR 0.91, 95% CI 0.81-1.02, 293 cases); this was largely due to causes other than cancer and the reduction in risk disappeared after the index child's death (SMR 0.98, 95% CI 0.84-1.15, 168 cases). Deaths from cancers of the lymphohaematopoietic system were in excess among mothers (SMR=2.13, 95% CI 1.02-3.92, 10 cases) and breast cancer deaths were in excess specifically among mothers of leukaemic children (SMR 2.32, 95% CI 1.16-4.14, 11 cases). Three mothers dying with breast cancer had index children who had been diagnosed with a bone sarcoma. Parental cancer of the respiratory tract was significantly associated with both tumours of the central nervous system and Hodgkin's lymphoma in the index child. The excess risks identified here may be due to genetic factors or due to parental psychological stress consequent to cancer in a child that may lead to increased mortality either through the direct effects of stress or through consequent changes in lifestyle.