The goal of this review is to demonstrate the effective interaction of epidemiologic methods and molecular genetics in the identification of familial cancer predisposition. The example involves a hospital-based population of childhood soft tissue sarcoma patients who were less than age 16 years at diagnosis at the University of Texas M. D. Anderson Cancer Center, Houston, Texas, in 1944 through 1976, had survived at least 3 years from diagnosis, and were diagnosed at least 5 years before the start of our study. Familial data were collected on the patients' offspring, full siblings, parents, aunts, uncles, and grandparents. The initial analysis revealed a small but significant cancer excess in first-degree relatives. Genetic analysis demonstrated that the cancer distribution in families could best be explained by a rare autosomal dominant gene with penetrance such that the risk of cancer by age 35 years was nearly 50%. Most of the evidence for a dominant gene came from nine kindreds. Laboratory investigation of fibroblasts from those kindreds provided an in vitro model of cellular immortalization and carcinogenesis. Germline mutations in the tumor suppressor gene p53 were found in two of the families, and studies are ongoing in the other kindreds. This review demonstrates the power of genetic epidemiologic methods to characterize statistically a cancer-predisposing gene and the application of molecular genetics to define the genetic defect.