A matched case-control study of retinoblastoma was conducted by the Children's Cancer Study Group (CCSG) to investigate the hypotheses that postconception exposures affect the risk of the nonheritable (post-zygotic origin) form of this disease and that preconception exposures affect the risk of the sporadic heritable (prezygotic origin) form. Eligible cases were those patients with retinoblastoma diagnosed in 1982-1985 at any of the CCSG member hospitals. Cases were classified as familial heritable, sporadic heritable, or nonheritable based on family history, tumor laterality, and cytogenetic analysis. Telephone interviews of parents of 201 cases and their pair-matched controls selected by random digit dialing were completed. Analysis of possible risk factors for the 67 sporadic heritable cases and the 115 nonheritable cases was performed. (The 19 familial cases were excluded). For the nonheritable group, gestational exposure to X-ray [odds ratio (OR) = 2.3, P = 0.08] and morning sickness medication (OR = 2.8, P = 0.02) and low maternal educational level (OR = 5.5, P = 0.03) were associated with increased risk; anemia (OR = 0.3, P = 0.02) and multivitamin use (OR = 0.4, P = 0.03) during pregnancy and periconceptional use of barrier contraceptive (OR = 0.1, P = 0.02) or spermicide (OR = 0.2, P = 0.02) were associated with decreased risk. In the sporadic heritable group, observations included a negative association with multivitamins during pregnancy (OR = 0.2, P = 0.02) and nonsignificant positive associations with preconception gonadal X-ray (maternal, OR = 2.0, P = 0.30; paternal, OR = 1.8, P = 0.42) and older parental age (case-control difference 1.0-1.2 years, P = 0.24-0.27). Many of the associations support study hypotheses, although the possibility of recall bias and chance findings suggest cautious interpretation.