Primary hepatocellular carcinoma (PHC), epidemiologically associated with chronic hepatitis B virus (HBV) infection, has historically been felt to be caused by the activation or introduction of an oncogene. However, transforming sequences from human PHC have not been reproducibly isolated. In this paper, evidence is presented that suggests PHC may result instead from the loss of an anti-oncogene. Seven of 12 human primary liver tumors tested against a panel of restriction fragment length polymorphisms (RFLPs) demonstrated loss of constitutional heterozygosity for markers on chromosome 4. Tumor and nontumor liver tissue were typed for 11 chromosome 4 RFLPs. In addition, at least one RFLP on nine other chromosomes (1, 2, 6, 7, 9, 11, 13, 14, and 17) was tested for allelic loss. Seven of nine tumors constitutionally heterozygous for chromosome 4q markers showed allele loss in tumor tissue. Six of the seven samples were jointly informative for both 4p and 4q markers. Five of the six demonstrated loss for only 4q RFLPs. In one individual, in which two samples were taken from distant locations within the same tumor, both samples showed loss of the same alleles. Among the other chromosomes informative for allele loss, one tumor showed changes on 13q. No other changes were observed in RFLPs located on the eight other chromosomes tested. These results indicate that an anti-oncogene may be located on 4q and suggest a mechanism for PHC and other cancers seroepidemiologically related to virus infection. Liver cancer caused by chronic HBV infection or other environmental agents may be linked through genetic events responsible for the loss of a tumor suppressor locus (anti-oncogene) located on chromosome 4.