Age-sex registers were compiled and updated for the east London general practices participating in a screening study for hypertension. Of 1435 addresses in the registers of two practices that were checked, 228 (16%) were incorrect, according to the return by the post office of the screening invitations and checking the addresses of the non-responders using telephone directories and the medical records. The non-responders to the screening invitation for whom a new address was not found, were visited at the address as recorded on the age-sex register. This showed that the true address error rate from the original age-sex registers was 26% and thus substantially greater than that calculated from returned letters. It is concluded that the non-acceptance rate of screening in general practice might be exaggerated as a result of the lack of a correct address for a substantial proportion of the patients on a general practice list.