Although Hemoccult (Smith Kline Diagnostics, Inc., Sunnyvale, Calif.) screening for colorectal cancer is based on detection of the peroxidase activity of blood, little is known about the relationship between patterns of bleeding from cancers or adenomas and sensitivity of the test slides. 51Cr-labeled blood loss was measured for a mean of 9.1 days in 46 patients with colorectal cancer and 28 with adenomas. Duplicate sets of Hemoccult II slides were prepared by the patients and mailed for testing with and without preliminary rehydration. Mean blood loss was significantly related to the site of the cancers, but was unrelated to Dukes' staging. Geometric mean levels of blood loss from the cancers were 9.3 ml/day for cecum and ascending colon, 1.5 ml/day for transverse and descending colon, 1.9 ml/day for sigmoid colon, and 1.8 ml/day for rectum. Rehydration increased Hemoccult sensitivity for blood independently of storage delay, almost doubling sensitivity at the levels of bleeding commonly encountered with cancers of the left side of the colon. Relating results to testing over 3 days, the Hemoccult false-negative rate for cancer was 31% without rehydration, falling to 9% with rehydration. With adenomas, blood loss and Hemoccult positivity were closely related to lesion size. The findings support the validity of 3-day occult blood testing as a method of screening for colorectal cancers and large adenomas, provided the test employed has a high sensitivity for blood. Further information is needed about the specificity of rehydrated Hemoccult slides in subjects following diets of differing peroxidase content. The degree to which rehydration of slides lowers the false-negative rate for cancer indicates that this modification of slide development should be evaluated in large population screening programs.