Magnetic resonance imaging (MRI) has rapidly attained a major position among the examinations used in the diagnostic approach of multiple sclerosis because it is highly sensitive in demonstrating lesions. However, these lesional images may have several meanings, and there is the problem of distinguishing between oedema, which is said to reflect recent lesions, and gliosis which is thought to betray old lesions. The intrinsic MRI parameters studied (i.e. relaxation times) are unable to make this distinction, whereas it is provided by paramagnetic contrast media such as gadolinium. There is no correlation between the changes observed at MRI and the severity of the disease. Another problem is the accuracy of lesion localization, since visualization is predominantly macroscopic. This raises several questions about the demonstration of correlations between clinical signs and site of the lesion(s). At the moment, several teams of neuroradiologists are trying to find the most reliable method to determine the size of the lesion. The possible prognostic value of this size and its changes as time goes by are other parameters to be considered once the basic diagnosis has been made.