Psychophysiological data support the concept that migraine is the result of cortical hypersensitivity, hyperactivity, and a lack of habituation. There is evidence that this is a brain-stem related information processing dysfunction. This cortical activity reflects a periodicity between 2 migraine attacks and it may be due to endogenous or exogenous factors. In the few days preceding the next attack slow cortical potentials are highest and habituation delay experimentally recorded during contingent negative variation is at a maximum. These striking features of slow cortical potentials are predictors of the next attack. The pronounced negativity can be fed back to the patient. The data support the hypothesis that a change in amplitudes of slow cortical potentials is caused by altered habituation during the recording session. This kind of neurofeedback can be characterized as "empirically based" because it improves habituation and it proves to be clinically efficient.