In 39 hospitalized patients with suspected gastrointestinal bleeding and given intravenous 51Cr-labeled red cells, reactions of three chemical spot tests for fecal occult blood were compared with the "true" blood loss as determined by stool radioassay. Guaiac reagent and orthotolidine (Hematest) tablets were extremely sensitive, but yielded false-positive reaction rates of 72% and 76%, respectively on the 240 stool specimens compared. A modified guaiac test (Hemoccult) exhibited a false positive rate of 12%. Of the 27 patients entering the study due to positive guaiac or Hematest screening tests, 17 (63%) were not bleeding. Hemoccult, approximately 1/4 as sensitive as guaiac and Hematest, could miss lesions with low rates of bleeding unless multiple stools were tested. While barium had no effect, iron therapy or laxatives tended to lower both false-positive and false-negative reactions for all reagents. A positive Hemoccult test usually indicated significant gastrointestinal bleeding and would appear to be the test of choice provided at least 3 stools are tested to minimize false-negative results.