Both twins in a pair in which one had suffered breast cancer (115 pairs) were asked about the occurrence of cancer among first- and second-degree relatives. For all cancer sites together, there was agreement about 129 cases; another 48 cancers were mentioned only by the case twin and another 55 only by the co-twin. For breast cancer, there was agreement about 22 events; another 12 cases were provided by the cancer twin and four by the co-twin. The odds ratio for a family history of cancer was 1.14; for breast cancer the corresponding value was 1.48. The results are suggestive of some recall bias in subjective reports on familial cancer.