Our aim was to test the reliability of interpreting MRI studies in neuro-Behçet's disease (NBD) and to determine the sensitivity and specificity of different MRI findings. We prospectively studied 50 patients: 24 had chronic NBD, 12 multiple sclerosis, 5 vasculitis other than Behçet's disease (BD) and 9 patients had BD without neurological involvement. MRI studies were performed according to a standard protocol with a 0.2 T imager. Two neuroradiologists, blinded to the diagnosis, age and sex of the subjects, reviewed the films independently, twice. Separate assessments were made for a set of items: dural sinus pathology, widening of ventricles and sulci, brain stem atrophy, lesions of the cerebral cortex, discrete lesions of deep white matter, basal ganglia, brain stem and cerebellum and the presence of smooth periventricular high-signal foci. Intraobserver agreement was substantial or better, and interobserver agreement moderate to substantial for most items. In these patients with chronic NBD we found low sensitivity on all assessed items. Dural sinus pathology or brain stem atrophy were highly specific, but parenchymal lesions in different sites had uniformly low specificity.