Carotid artery plaques with intraplaque haemorrhage or atheromatous debris have been found to be associated with an increased risk of embolic stroke. Other methods have failed to detect plaque morphology, and it is not clear whether MRI allows differentiation between prognostically and therapeutically relevant plaque types. We examined 17 carotid bifurcation plaques which had been removed in toto by MRI. For quantifying MR signal intensities (I) the contrast-to-noise ratio (CNR) was used: (ITissue-IRef)/SDRef, with normal saline (0.9%) as reference (Ref) and the standard deviation (SD) of the noise. Measurements were correlated with the histopathological appearance of "simple plaques", consisting of fibrous intimal thickening, lipid deposits and/or atheromatous tissue with cholesterol crystals, largely calcified plaques, and "complicated plaques", containing recent intramural haemorrhage or friable atheromatous debris. Significantly different mean CNR could be measured in the three plaque types on T1- and T2-weighted sequences (p < 0.00001) and using the FLASH pulse sequence with a flip angle of 15 degrees (p < 0.001). With the T1-weighted sequence simple plaques showed a CNR of 4.4 +/- 2.3, calcified plaques -4.8 +/- 2.6 and complicated plaques 15.1 +/- 4.3. Using this technique, each single plaque could be correctly classified, an unalterable prerequisite for a clinical application. To date, motion artefacts due to patient movement or insufficiently triggerable vessel pulsation in combination with relative long acquisition times (6-7 min) have limited in vivo investigations. If these problems could be overcome, MRI might become a valuable technique for studying carotid plaque morphology.