Because of their increased risk for second cancers, childhood cancer survivors are people who really should not smoke, but available evidence suggests that they do. We studied the smoking habits of long-term childhood cancer survivors in data collected from 1289 adult survivors of childhood cancer and 1930 of their sibling controls. Survivors were diagnosed with cancer between 1945 and 1974 when they were less than 20 years old. Using matched analyses that controlled for the influence of family, survivors were 8% less likely than controls to be current smokers, 13% less likely to be ever-smokers, but 12% less likely to have quit smoking; these differences were not statistically significant. In a logistic regression analysis there was a significant difference by year of diagnosis for current smoking rate ratios (RR); survivors were less likely to be current smokers if diagnosed in recent years (RR = 0.76; 95% confidence intervals = 0.58-0.98, between 1965-74) and quite similar to controls if diagnosed in earlier years (RR = 1.05 between 1945 and 1954). In our group of long-term cancer survivors, the reduction in current smoking came about because survivors were more inclined never to start smoking than controls. Once addicted to tobacco, they were less likely to quit. While the fact that survivors are less likely to start smoking is encouraging, the persistence of smoking habits strongly suggests the need for continuing efforts to prevent smoking in this most vulnerable group.