The interobserver variability of a Papanicolaou smear diagnosis of atypical glandular cells of undetermined significance (AGUS) has not been measured. Four expert cytopathologists restrospectively reclassified 100 smears originally diagnosed as AGUS; on follow-up examination, 54 had a clinically significant lesion and 46 had a benign lesion. The mean sensitivity and specificity of reclassification were 86% and 21%, respectively. The κ statistic for pairwise cytopathologist comparison varied from 0.16 to 0.27. In 45% of cases the cytopathologists all used different Bethesda System diagnoses, and in no case did all the cytopathologists use the same diagnosis. There was little agreement on which cytologic criteria were important in separating clinically significant and benign lesions. We conclude the following: interobserver agreement in reclassifying AGUS lesions is very poor; the AGUS category is poorly understood, and there is no agreement on diagnostic cytologic criteria; and when reclassifying slides, cytopathologists make a number of false-negative diagnoses.