Macroregenerative nodules (MRNs), probably representing a pathway for human hepatocarcinogenesis, are generally classified into type I MRNs (or ordinary adenomatous hyperplasia) and type II MRNs (or atypical adenomatous hyperplasia), on the basis of imprecise definitions of cytological and architectural atypia. It is currently believed that type II MRNs are probably true precursors of hepatocellular carcinoma (HCC), whereas type I lesions may simply represent large regenerative nodules. A series of 155 consecutive adult cirrhotic liver explants were examined for evidence of MRNs, HCC, and liver cell dysplasia (LCD) of large and small cell types, and their appearance, in terms of proposed classification schemes, was reviewed. There was evidence indicating that the presence of either type of MRN was associated with an increased incidence of HCC (all MRNs, P < .00019; type I MRNs, P < .067; type II MRNs, P < .012) compared with cirrhotic livers without MRNs. A subset of younger patients with a large (uncountable) number of MRNs in their livers, who did not show any increased incidence of carcinoma, was identified. Excluding these cases from statistical analysis, all associations were strengthened, implying either that malignant progression had not had time to occur in this younger population or that these nodules were simply large regenerative nodules without malignant potential. MRNs from these livers were histologically indistinguishable from MRNs occurring in more limited numbers, although atypical changes other than large cell type LCD were less frequent. No independent association between LCD of large cell type and HCC was found in the entire series. Deleting this feature from the criteria for cytological atypia resulted in a stronger association of both types of MRNs with HCC (redefined type II MRNs/HCC, P < .0001; redefined type I MRNs/HCC, P < .0306). Some of the type II MRNs remaining after exclusion of large cell type LCD showed “borderline” changes insufficient for a diagnosis of HCC, but most type II MRNs (82%) contained expansile “nodule-in-nodule” growth patterns. The conclusions of this report are that (1) histological examination of type I MRNs is insufficient in many cases to distinguish large regenerative nodules from neoplastic ones; (2) LCD of large cell type should not be used as a criterion for terming an MRN atypical; and (3) expansile “nodule-in-nodule” formation in MRNs should be considered to represent evidence of architectural atypia.