It has been proposed that Hodgkin's disease (HD) may have an infectious origin and that delayed exposure to infection may increase the risk of HD in young adults. This hypothesis is addressed by studying the family structure among children and young adults. The Civil Registration System was used to establish a population-based cohort consisting of all persons whose mothers were born in Denmark since 1935. Persons who developed HD were identified by linkage with the Danish Cancer Registry. HD incidence rate ratios were estimated based on a log-linear Poisson regression model. The cohort of 2.1 million persons (aged 0-42 years) was followed for 31.1 million person years, during which period 378 cases of HD occurred. Among children (< 15 years, n = 72), the relative risk (RR) of HD tended to increase with increasing sibship size, the relative increase in risk per increase in sibship size (the trend) being 1.28 [95% confidence interval (CI) 1.00-1.63]. The trend for birth order was 1.26 (95% CI 0.92-1.73). Among young adults (> or = 15 years, n = 306) the risk of HD, on the contrary, tended to decrease with increasing sibship size [trend = 0.91 (95% CI, 0.81-1.03)] and birth order (trend = 0.85 (95% CI, 0.71-1.01). These trends among young adults were significantly different from the corresponding trends among children (p < 0.05). Siblings of cases were at increased risk of HD (RR = 18; 95% CI, 2.2-65, n = 2). Our findings are compatible with the hypothesis that delayed exposure to infection may be a risk factor for HD in young adults, and that early exposure perhaps to another infectious agent may increase the risk of HD in children.