Intraoperative autotransfusions for selected patients can ease demands on blood banks. In 33 cases (trauma, elective vascular and cardiac procedures), 80.7 l of blood were saved with a cell washer blood recovery system. In 8 trauma cases, including 3 with intestinal perforation, cell washing salvaged 31.5 l of shed blood as a lifesaving emergency procedure. This stimulated interest in defining the limits of the instrumentation in cleansing blood of bacteria. In 18 in vitro experiments, discarded banked blood was inoculated with Escherichia coli to simulate light, moderate or heavy bacterial contamination. Volumes of blood (500 ml) containing a total of 6.3 .times. 105, 4.8 .times. 107 and 3.2 .times. 1010 bacteria were processed. The original blood, filtered blood, plasma supernate, effluent wash solution and final washed packed red cells were cultured quantitatively in each experiment. The mean total number of E. coli retained in the final washed packed red cells was 1.4 .times. 105, 6.3 .times. 106 or 1.6 .times. 109 bacteria, or 23, 13 and 5% of each original inoculum. Additional washing with 10 l of saline did not remove significantly more bacteria (P > 0.20). Although emergency autotransfusion of blood contaminated with intestinal contents was lifesaving, caution is recommended since these results show that cell washing does not remove all bacteria. Further laboratory and clinical studies are needed to determine the levels of bacterial contamination of autotransfused blood that can be tolerated and to determine adjunctive means of rendering contaminated or potentially contaminated blood safe for autotransfusion.