Doppler ultrasound was used to study 120 legs of 106 patients with suspected deep vein thrombosis (DVT) or pulmonary embolism. Venography was subsequently performed in all. DVT was confirmed by venography in 44 legs and was confined to the calf in 10 of these. Ultrasound detected three calf thromboses and 29 out of 34 more extensive thromboses. Of five undetected thrombi that were proximal to the calf one was associated with partial occlusion and four with extensive collateral circulation. Of the 76 limbs without venographic evidence of thrombosis 21 were thought to have DVT by ultrasound; 18 of these false-positive results could be attributed to external compression of veins, two to excessive tenderness precluding adequate examination; and in one no explanation was found. This test gives more accurate results than judging by clinical signs alone, but users must be aware of its limitations and, particularly, the causes of false-positive and false-negative results.