Many childhood cancer survivors have psychosocial late effects. We studied the risks for cohabitation and subsequent separation. Through the Danish Cancer Register, we identified a nationwide, population-based cohort of all 1877 childhood cancer survivors born from 1965 to 1980, and in whom cancer was diagnosed between 1965 and 1996 before they were 20 years of age. A sex-matched and age-matched population-based control cohort was used for comparison (n=45,449). Demographic and socioeconomic data were obtained from national registers and explored by discrete-time Cox regression analyses. Childhood cancer survivors had a reduced rate of cohabitation [rate ratio (RR) 0.78; 95% confidence interval (CI): 0.73-0.83], owing to lower rates among survivors of both noncentral nervous system (CNS) tumors (RR 0.88; 95% CI: 0.83-0.95) and CNS tumors (RR 0.52; 95% CI: 0.45-0.59). Male CNS tumor survivors had a nonsignificantly lower rate (RR 0.47; 95% CI: 0.38-0.58) than females (RR 0.56; 95% CI: 0.47-0.68). The rates of separation were almost identical to those of controls. In conclusion, the rate of cohabitation was lower for all childhood cancer survivors than for the population-based controls, with the most pronounced reduction among survivors of CNS tumors. Mental deficits after cranial irradiation are likely to be the major risk factor.