A small proportion of brain tumors are attributed to a genetic predisposition; however, the hereditary proportion is undetermined. This study evaluates the degree of familial aggregation of cancer in a large series of brain tumor patients. Our study included 5,088 relatives of 639 probands (3,810 first- and 1,278 second-degree), diagnosed with a glioma between June 1992 and June 1995 at The University of Texas M. D. Anderson Cancer Center, Houston, Texas, with diagnosis under age 65 years, and residents of the United States or Canada. We conducted an in-person or telephone interview with patients and/or their next-of-kin, and obtained family histories for the probands' first-degree (parents, siblings, offspring) and selected second-degree relatives (aunts, uncles, grandparents) using a sequential sampling strategy. Reported cancers were documented by medical records and/or death certificates (if the relative was deceased and medical records were unavailable). We conducted segregation analysis using the Pedigree Analysis Program (PAP). The analyses were divided into two categories: (1) all 639 families, and (2) a subset of families whose gliomas stained positive on p53 immunohistochemistry analysis. We demonstrated that a multifactorial Mendelian model was favored, while a model postulating a purely environmental cause of brain cancer was rejected. This study indicates that familial cancer in relatives of glioma patients are probably a result of multigenic action, and familial clustering of cancer among relatives of glioma patients may involve unknown environmental exposures.