Analysis of liver from rats exposed to chemical hepatocarcinogens has led to a model that postulates sequential premalignant changes, culminating in hepatoma formation from neoplastic nodules. Several experimental protocols devised during the last quarter century have focused upon this lineage model. But proof that neoplastic nodules are the definitive premalignant lesions has not been achieved. Recent work, using α-fetoprotein as a marker for liver cell alterations induced by different carcinogen-feeding regimens, suggests that chemically induced hepatomas may also arise from nonnodular cell populations. Therefore, the extensive biochemical and biological studies of presumed “premalignant” cells may have utilized the wrong cells. Unequivocal identification of the cell population at risk for malignancy is needed to delineate mechanisms by which chemicals cause hepatocellular carcinoma.