Objectives: To evaluate the accuracy of routinely available non-invasive tests (spiral computed tomographic angiography (CTA), time of flight magnetic resonance angiography (MRA), and colour Doppler ultrasound (DUS)), individually and together, compared with intro-arterial digital subtraction angiogrophy (DSA) in patients with symptomatic tight carotid stenosis; and to assess the effect of substituting non-invasive tests for DSA on outcome, interobserver variability, and patient preference. Methods: Patients referred from a neurovascular clinic were subjected prospectively to DUS imaging. The operator was blind to symptoms. Patients with a tight carotid stenosis on the symptomatic side were admitted for DSA. CTA and MRA were performed during the admission. The CTA, MRA, and DSA films were each read independently by two of six experienced radiologists, blind to all other data. Results: 67 patients were included (3,4 had all four imaging procedures). DUS, CTA, and MRA all agreed with DSA in the diagnosis of operable v non-operable disease in about 80% of patients. CTA tended to underestimate (sensitivity 0.65, specificity 1.0), MRA to overestimate (sensitivity 1.0, specificity 0.57), and DUS to agree most closely with (sensitivity 0.85, specificity 0.71) the degree of stenosis as shown by DSA. When using any two of the three non-invasive tests in combination, adding the third if the first two disagreed would result in very few misdiagnoses (about 6%). MRA had similar interobserver variability to CTA (both worse than DSA). Patients preferred CTA over MRA and DSA. Conclusions: DUS, CTA, and MRA all show similar accuracy in the diagnosis of symptomatic carotid stenosis. No technique on its own is accurate enough to replace DSA. Two non-invasive techniques in combination, and adding a third if the first two disagree, appears more accurate, but may still result in diagnostic errors.