We sought to determine whether the variability in dysplasia rates in cases of atypical squamous cells of undetermined significance (ASCUS) reflects variability in interpretation of cervical biopsy specimens. In phase 1, 124 biopsy specimens obtained because of a cytologic diagnosis of ASCUS were reviewed independently by 5 experienced pathologists. Diagnostic choices were normal, squamous metaplasia, reactive, indeterminate, low-grade squamous intraepithelial lesion (LSIL), and high-grade squamous intraepithelial lesion (HSIL). The rate of dysplasia ranged from 23% to 51%. All pathologists agreed in 28% of cases. In 52% of cases, the diagnoses ranged from benign to dysplasia. The overall interobserver agreement was poor. In phase 2, 60 cervical biopsy specimens (21 obtained for ASCUS, 22 for LSIL, and 17 for HSIL) were evaluated using the same diagnostic choices. Agreement was better in biopsies performed for HSIL and LSIL compared to those for ASCUS. Intraobserver reproducibility in the interpretation of biopsies performed for ASCUS ranged from poor to excellent. We conclude that variability in the interpretation of biopsy specimens plays an important role in the differences in rates of dysplasia reported for the follow-up of ASCUS.