We studied 121 consecutive families referred to a cancer genetics clinic. The proband provided a detailed written history of cancers occurring in first, second and third degree relatives and we attempted to obtain medical records to validate each reported case of cancer. For 216 cases of cancer in 180 first degree relatives, the primary cancer site was correctly identified in 83%. For second and third degree relatives the written histories of primary cancer were accurate for 67 and 60% respectively. The effect of the corresponding levels of error should be estimated in epidemiologic studies and should motivate clinicians to seek medical records in situations where a family history of cancer influences clinical management.