Because of the absence of data on smoking history for patients in cancer registries, large population-based studies have been unable to directly assess the role of smoking in the risk of multiple primary cancers. In a random sample of 618 adults diagnosed with invasive cancer in 2002 and reported to the population-based Connecticut (CT) cancer registry, however, search of hospital records identified smoking history (current, former, never- or nonsmoker) for 83.3%. Number of cigarettes per day was known for 71.3% of 289 current or former smokers, while duration of quitting was found for 92.3% of 194 former smokers. Smoking status was strongly associated with current heavy alcohol use. Hospital records appear to be a promising resource for population-based studies on smoking histories in relation to risk of multiple primaries, although completeness of information needs to be improved in CT and assessments are needed in other cancer registries.