In 1983, we reported results from the Oxford Family Planning Association contraceptive study regarding the association between oral contraceptives (OCs) and cervical neoplasia, after a 10 year follow-up of a cohort of 17,000 women. Further findings from this study are reported here after an additional 12 years of follow-up. A nested case--control design was used in which cases were all women diagnosed under 45 years of age with invasive carcinoma (n = 33), carcinoma in situ (n = 121) or dysplasia (n = 159). Controls were randomly selected from among cohort members and matched to cases on exact year of birth and clinic attended at recruitment to study. Conditional logistic regression analysis was used to determine odds ratios (ORs) and 95% confidence intervals (CIs) associated with various aspects of OC use relative to never users adjusted for social class, smoking, age at first birth and ever use of diaphragm or condom. Ever users of OCs had a slightly elevated OR for all types of cervical neoplasia combined (OR = 1.40, 95% CI 1.00-1.96). Odds ratios were highest for invasive carcinoma (OR = 4.44, 95% CI 1.04-31.6), intermediate for carcinoma in situ (OR = 1.73, 95% CI 1.00-3.00) and lowest for dysplasia (OR = 1.07, 95% CI 0.69-1.66). The elevated risk associated with OC use appeared to be largely confined to current or recent (last use in the past 2 years) long-term users of OCs. Among current or recent users, ORs for all types of cervical neoplasia combined were 3.34 (95% CI 1.96-5.67) for 49-72 months of use, 1.69 (95% CI 0.97-2.95) for 73-96 months and 2.04 (95% CI 1.34-3.11) for 97 or more months. These results suggest a possible effect of OC use on later stages of cervical carcinogenesis, although residual confounding due to sexual factors or human papillomavirus (HPV) infection cannot be ruled out.