We report a prospective, blinded comparison of compression ultrasound (US) and Tc-99m erythrocyte-labelled venous blood pool scintigraphy (BPS) in patients presenting with symptoms of deep venous thrombosis (DVT). Contrast venography (CV) was used as the gold standard. Ninety-eight lower limbs of 76 patients were examined. DVT was present at CV in 38 of 98 limbs and was isolated to the calf veins in eight. Sensitivity and specificity of ultrasound for femoropopliteal thrombus were 81.5% and 96% and of venous blood pool scintigraphy were 55% and 96%. For deep venous thrombosis in the whole limb sensitivity and specificity of ultrasound were 74% and 90% and of venous blood pool scintigraphy were 61% and 88%. In the calf sensitivity and specificity of US were 61% and 94% and of venous blood pool scintigraphy were 61% and 89%. Excluding equivocal venous blood pool scintigraphy results, the predictive values of a positive and negative venous blood pool scintigraphy study for the whole limb were 84% and 86%. The predictive values of a positive and negative ultrasound where the examination was adequate were 82% and 86%. US is a more sensitive alternative to CV than BPS for femoropopliteal DVT. When neither US nor CV can be performed, BPS remains a useful initial test for DVT, provided it is unequivocally positive or negative.