This paper describes a method for increasing the spatial detail of the EEG and for integrating physiological data with anatomical models based on magnetic resonance images (MRIs). This method includes techniques to efficiently record EEG data from up to 124 channels, to measure 3-D electrode positions for alignment with MRI-derived head models, and to estimate potentials near the outer convexity of the cortex using a spatial deblurring technique which uses a realistic model of the structure of the head and which makes no assumptions about the number or type of generator sources. The validity of this approach has been initially tested by comparing estimated cortical potentials with those measured with subdural grid recordings from two neurosurgical patients. The method is illustrated with somatosensory steady-state evoked potential data recorded from 5 healthy subjects. Results suggest that deblurred 124-channel topographic maps, registered with a subject's MRI and rendered in 3 dimensions, provide better spatial detail than has heretofore been obtained with scalp EEG recordings. The results also suggest that the potential for EEG as a functional neuroimaging modality has yet to be fully realized.