A previously developed morphometric classification rule has been shown to be successful in identifying approximately 30% of patients with atypical hyperplasia of the endometrium in whom the finding does not imply a progression to malignancy. The reproducibility of the nuclear classification rule was tested in blind, duplicate morphometric assessments by different technicians. The results showed a satisfactorily high degree of consistency and reproducibility, with only one of ten cases classified differently. In a second series of experiments, the nuclear classification rule was applied to samples from 101 nonmalignant cases (39 proliferative endometriums, 7 secretory endometriums, 55 mildly atypical hyperplasias), 8 markedly atypical hyperplasias and 43 malignant cases (20 well-differentiated adenocarcinomas and 23 moderately to poorly differentiated adenocarcinomas of the endometrium). Ideally, the rule should classify all nonhyperplastic and mildly hyperplastic cases as nonprogressive and all carcinomas as progressive; there were, however, a considerable number of false positives and false negatives based on application of the classification rule to these cases. Therefore, the sensitivity and specificity of nuclear morphometry using the classification rule developed for atypical hyperplasias is too low to allow its random application. This emphasizes the selective nature of diagnostic morphometry, in which the full diagnostic capacity of the pathologist must be used in selection of the proper cases to be studied.