Screening has had a major impact on cervical cancer in many countries. Although there can be no doubt about its effectiveness in preventing squamous-cell carcinoma, there is little evidence of any benefit on adenocarcinoma and adenosquamous carcinoma of the cervix, and many authors have concluded that it is ineffective. A population-based case-control design was used in women aged 20-69 in the United Kingdom, with information on screening obtained from routine databases. Among 3,305 cases with known histology, 641 had adenocarcinoma and 133 adenosquamous carcinoma. The risk reduction associated with 3-yearly screening was greater for squamous carcinoma (75%, 95%CI 71-79%) and adenosquamous carcinoma (83%, 95%CI 68-91%) than for adenocarcinoma (43%, 95%CI 24-58%). Among stage 1B+ cases, 83% (335/406) of women with adenocarcinoma had been screened within 10 years of diagnosis. This is very similar to controls (82%, 3,292/3,965), but much higher than in women with squamous carcinoma (57%, 852/1,493). Incidence of adenocarcinoma was low within 2.5 years of a negative smear (OR 2.3, 95%CI 0.15-0.34), but was no different from the background rates 4.5-5.5 years after a negative smear. We conclude that screening has reduced the incidence of adenocarcinoma of the cervix, but the prognostic value of cytology is less (in both magnitude and duration) for adenocarcinoma than for squamous carcinoma. The impact of screening on adenosquamous carcinoma is similar to its impact on squamous carcinoma.