One explanation for the weak relationship between neuropsychological deficits and conventional measures of disease burden in multiple sclerosis is that brain 'plasticity' allows adaptive reorganization of cognitive functions to limit impairment, despite injury. We have tested this hypothesis. Ten patients with multiple sclerosis and 11 healthy controls were studied using a functional MRI (fMRI) counting Stroop task. The two subject groups had comparable performances, but a predominantly left medial prefrontal region [Brodmann area (BA) 8/9/10] was more active during the task in patients than in controls (corrected P < 0.001), while a right frontal region (including BA 45 and the basal ganglia) was more active in controls than in patients (corrected P = 0.004). The magnitude of the differences correlated with the normalized brain parenchymal volume, a measure of disease burden (r = -0.72, P = 0.02). We then tested the effects of acute administration of rivastigmine, a central cholinesterase inhibitor, on patterns of brain activation. In five out of five multiple sclerosis patients there was a relative normalization of the abnormal Stroop-associated brain activation, although no change in the patterns of brain activation was found in any of four healthy controls given the drug and tested in the same way. We suggest that recruitment of medial prefrontal cortex is a form of adaptive brain plasticity that compensates, in part, for relative deficits in processing related to the reduced right prefrontal cortex activity with multiple sclerosis. This functional plasticity is modulated by cholinergic agonism and must arise from potentially highly dynamic mechanisms such as the 'unmasking' of latent pathways.