20 525 patients from general practitioners' lists were randomly allocated into test and control groups. The 10 253 test subjects were invited to perform haemoccult faecal occult blood testing over 3 days. 3613 (36 . 8%) of the 9807 who received their invitations completed the test. Compliance was improved by direct invitation from the general practitioner and by prior health education by letter or interview. 77 people (2 . 1%) had a positive test result, and 50% of these on investigation had neoplastic disease--12 had invasive carcinomas (9 Dukes' stage A, 2 stage B, 1 stage C) and 27 had 40 adenomas (12 over 2 cm, 2 of which contained areas of severe dysplasia). In the year following the screening test 1 carcinoma (stage C) has presented in the group which accepted the test, and 10 carcinomas (4 stage B, 4 stage C, 2 stage D) have presented in the control group. This respresents a 3 . 6 times greater detection rate per 1000 persons in the test group than in the control group. Only 8 adenomas have presented in the control and non-responding groups. Fibreoptic sigmoidoscopy identified the 10 carcinomas within its range and 39 of the 40 adenomas. Double-contrast barium enema identified only 9 of the 12 carcinomas and 24 (62%) of the 40 adenomas. All 3 carcinomas not identified by barium enema were polypoid Dukes' stage-A lesions.