OBJECTIVE To define the strengths and weaknesses of existing research on the natural history of cervical squamous intraepithelial lesions (SIL) and to estimate rates of progression and regression without treatment. METHODS Studies of women whose cervical smears showed squamous atypia or worse and who were observed for a minimum of 6 months were identified by a search of MEDLINE from 1966 to 1996, Current Contents, the Federal Research in Progress database, and references of review articles and identified studies, and by experts in the field. METHODS Fifteen of 81 studies were eligible for data extraction. To be eligible, studies had to report a minimum of 6 months' follow-up without treatment; relate entry cytologic findings to outcomes; and report entry cytologic findings so that the study population could be stratified into categories of atypical cells of undetermined significance (ASCUS), low-grade SIL, or high-grade SIL. Studies published before 1970 were excluded. RESULTS Eligible studies, representing 27,929 patients, were stratified according to entry cytologic findings. The following rates of progression to high-grade SIL at 24 months were found: ASCUS, 7.13% (95% confidence interval [CI] 0.8%, 13.5%); low-grade SIL, 20.81% (6.08%, 35.55%); and high-grade SIL, 23.37% (12.82%, 32.92%). The following rates of invasive cancer at 24 months were found: ASCUS, 0.25% (0%, 2.25%); low-grade SIL, 0.15% (0%, 0.71%); and high-grade SIL, 1.44% (0%, 3.95%). The following rates of regression to normal were found: ASCUS, 68.19% (57.51%, 78.86%); low-grade SIL, 47.39% (35.92%, 58.86%); and high-grade SIL, 35.03% (16.57%, 53.49%). Study heterogeneity was not explained by regression analysis of study level variables. CONCLUSIONS Our findings for borderline and low-grade abnormal cervical cytologic results suggest a relatively low risk of invasive cervical cancer with observation up to 24 months and support the clinical policy of early colposcopy for high-grade lesions.