Computational fluid dynamics (CFD) provides a means for the quantitative analysis of haemodynamic disturbances in vivo, but most work has used phantoms or idealised geometry. Our purpose was to use CFD to analyse flow in carotid atherosclerosis using patient-specific geometry and flow data. Eight atherosclerotic carotid arteries and one healthy control artery were imaged with magnetic resonance angiography (MRA) and duplex ultrasound, and the data used to construct patient-specific computational models used for CFD and wall shear stress (WSS) analysis. There is a progressive change in three-dimensional (3-D) velocity profile and WSS profile with increasing severity of stenosis, characterised by increasing restriction of areas of low WSS, change in oscillation patterns, and progressive rise in WSS within stenoses and downstream jets. Areas of turbulent, retrograde flow and of low WSS are demonstrated in the lee of the stenoses. This study presents the largest CFD analysis of abnormal haemodynamics at the atheromatous carotid bifurcation using patient-specific data and provides the basis for further investigation of causal links between haemodynamic variables and atherogenesis and formation of unstable plaque. We propose that this provides a means for the prospective assessment of relative stroke risk in patients with carotid atherosclerosis.