Epidemiologically, screening is justified by the importance of the disease and the lack of prospects for primary prevention, but evidence from natural history is unhelpful since men are more likely to die with, rather than from, prostate cancer. The available screening tests do not always detect men whose lesions could result in future morbidity or mortality. Evidence is limited for the benefits of treatment for localised cancers detected through screening, whereas the evidence for harm is clear. Observational evidence for the effect of population screening programmes is mixed, with no clear association between intensity of screening and reduced prostate cancer mortality. Screening for prostate cancer cannot be justified in low-risk populations, but the balance of benefit and harm will be more favourable after risk stratification. Prostate cancer screening can be justified only in research programmes designed to assess its effectiveness and help identify the groups who may benefit.