Imaging was done in 80 patients with gastrointestinal hemorrhage using red blood cells labeled with 99mTc by a modified in vivo technique. Bleeding was detected in 65% of 40 patients with bright red blood per rectum, 71% of 35 patients with melena, and in none of the five patients with occult bleeding and chronic anemia. No clinical evidence of further gastrointestinal hemorrhage occurred in 26 of the 29 patients who were negative on imaging. The imaging study could be carried out for over 24 hr which increased the sensitivity of the test, since only 16% of the studies were abnormal on the initial images. Neither the need for blood replacement nor the presence of bright red blood per rectum correlated strongly with early image positivity, supporting the contention that patients with gastrointestinal hemorrhage do not necessarily bleed continuously. Labeled red cell imaging was more sensitive in the detection of bleeding than angiography. It corresponded to the overall angiographic findings in 23 of 31 cases and in 17 of the 18 patients who had demonstrable extravasation. 99mTc red blood cells may be used effectively in patients with melena as well as with bright red blood per rectum, and thus would seem to be the preferred radiopharmaceutical for imaging.