Prospective data on 202 consecutive patients who had a total of 123 total hip and ninety-four total knee arthroplasties were collected from two university medical centers. The findings of routine surveillance for deep venous thrombosis performed with ascending contrast venography were compared with those of surveillance with duplex ultrasonography complemented with color-flow Doppler imaging. All of the studies were performed between the third and seventh postoperative days. Of the 202 patients (342 extremities) who were examined, fifty-five (27 per cent) were found to have deep venous thrombosis; fifty-two (95 per cent) of the thrombi were in the calf and three (5 per cent) were in the proximal veins. All of the thrombi were clinically asymptomatic and all were nonocclusive, allowing passage of contrast medium around an intraluminal filling defect. Duplex ultrasonography with color-flow Doppler imaging correctly identified two of the three proximal thrombi and five of the fifty-two thrombi in the calf (sensitivity, 10 per cent). The sensitivity for the detection of thrombi in the calf was zero of sixteen at one of the institutions involved in the study and 14 per cent (five of thirty-six) at the other. There were two false-positive findings on ultrasonographic examination; one involved a proximal thrombus and one, a distal thrombus. We believe that the interinstitutional variability and insensitivity of duplex ultrasonography with color-flow Doppler imaging for the detection of asymptomatic deep venous thrombi in the calf after total joint replacement make it unreliable as a routine surveillance tool after total hip or knee arthroplasty.