Objectives To compare the performance and acceptability of unsupervised self-sampling with clinician sampling for high-risk human papillomavirus (HPV) types for the first time in a UK screening setting. Setting Nine hundred and twenty women, from two demographically different centres, attending for routine cervical smear testing Methods Women performed an unsupervised HPV self-test. Immediately afterwards, a doctor or nurse took an HPV test and cervical smear. Women with an abnormality on any test were offered colposcopy.Results Twenty-one high-grade and 39 low-grade cervical intraepithelial neoplasias (CINs) were detected. The sensitivity for high-grade disease (CIN2 þ ) for the self HPV test was 81% (95% confidence interval [CI] 60-92), clinician HPV test 100% (95% CI 85-100), cytology 81% (95% CI 60-92). The sensitivity of both HPV tests to detect high-and low-grade cervical neoplasia was much higher than that of cytology , clinician test 80% [95% CI 68-88], cytology 48% [95% CI 36-61]). For both high-grade alone, and high and low grades together, the specificity was significantly higher for cytology (greater than 95%) than either HPV test (between 82% and 87%). The self-test proved highly acceptable to women and they reported that the instructions were easy to understand irrespective of educational level. Conclusions Our results suggest that it would be reasonable to offer HPV self-testing to women who are reluctant to attend for cervical smears. This approach should now be directly evaluated among women who have been non-attenders in a cervical screening programme.