We evaluated the influence of the expertise of the technician on the accuracy of compression ultrasonography as a screening test to detect postoperative deep venous thrombosis in patients who had had primary or revision arthroplasty of the hip or the knee. The study was performed in two phases on two groups of patients who were evaluated with both compression ultrasonography and conventional venography. All of the patients received prophylaxis for deep venous thrombosis. In the first phase of the study, adequate venograms and ultrasonograms were made for 126 extremities (121 patients). Venography demonstrated seven thrombi in the proximal veins. Only three ultrasonograms were considered positive. Ultrasonography had a sensitivity of 0 per cent because none of the seven proximal thrombi identified with venography were detected with ultrasonography. Three ultrasonograms were false-positive (that is, the ultrasonogram was positive and the venogram was negative), so the specificity was 97 per cent. The accuracy of compression ultrasonography was 92 per cent. Both tests were performed on a second group of eight-seven extremities (eighty-four patients). Proximal clots were detected with venography in five extremities. The ultrasonogram was positive for seven extremities, including the five in which a clot had been identified with venography. There were no false-negative and two false-positive results. In this phase of the study, ultrasonography had a sensitivity of 100 per cent, a specificity of 98 per cent, and an accuracy of 98 per cent. The only difference that was identified between the two groups of patients was the experience of the ultrasonography technician. It appears that the reliability of compression ultrasonography is directly dependent on the experience of the technician.