Background : Cervical dysplasia, also referred to as squamous intraepithelial lesion (SIL) in cytology or cervical intraepithelial neoplasia in histopathology, is thought to have the potential to advance in progressive stages to cervical cancer. However, not all cases of SIL progress, and most of the mild lesions spontaneously regress. Factors that govern regression, persistence, and progression of SIL are poorly understood. Purpose: Our analysis sought to identify factors that determined persistence or regression of SIL. Methods: Seventy subjects with histopathologically confirmed cervical dysplasia were followed at 3-month intervals for 15 months. At each visit, the cervix was evaluated by Pap smear and colposcopy, and exfoliated cervicovaginal cells were analyzed for human papillomavirus (HPV) DNA. For each subject, data from every two consecutive visits were grouped as a pair. Persistent SIL was considered present if a lesion was detected at a visit (t) as well as at the next visit (t + 1) and absent if a lesion was detected at visit t but not at visit t + 1. A statistical model for time-dependent data correlated persistent SIL with various risk factors. Results: Age, ethnicity, education, sexual behavior, smoking, and the use of oral contraceptives did not correlate with persistent SIL. The risk of persistent SIL was associated with continual HPV infection in visits / and t + 1 (HPV positive by Southern blot analysis: odds ratio [OR] = 3.91, and 95% confidence interval [CI] = 1.58–9.65; HPV positive by polymerase chain reaction [PCR]: OR = 2.42, and 95% CI = 1.03–5.67) and a persistent high viral load (OR = 4.07, and 95% CI = 1.35–12.30). When typed by PCR, individuals with type-specific persistent infection in visits t and t + 1, and particularly those with a continual high viral load (OR = 4.97; 95% CI = 1.45–17.02), had the highest risk for persistent SIL compared with those with a low level of type-specific persistent infection or non-type-specific persistent infection. The presence of persistent HPV infection in visits t − 1 and t (the preceding time interval) was also predictive of persistent SIL in visits t and t + 1, although the strength of association was weaker, suggesting that persistent HPV and SIL occur synchronously. Conclusion: HPV infection and its associated cervical lesions tend to occur concurrently, and type-specific persistent HPV infection, particularly with a high viral load, produces chronic cervical dysplasia. Implications: The natural history of genital HPV infection directly influences the prognosis of cervical dysplasia as measured by persistence of the lesion. Testing for HPV infection may be valuable in the clinical management of women with cervical dysplasia. [J Natl Cancer Inst 1995;87: 1365–71]