Abstract

MEG/EEG are the only non-invasive methods to instantaneously and directly measure the currents underlying cerebral information processing, but their ability to localize those currents is limited. Source localization from MEG is always uncertain, unless the signal is already known to be coming exclusively from a single focal source, or a few highly separated focal sources. Furthermore, since many cerebral currents produce little or no MEG signal, even accurate localization of the MEG sources may provide a very incomplete map of brain activation. Intracranial EEG (iEEG) can unambiguously localize sources, using steep voltage gradients, doubly-inverting gradients, traverses of the source structure from multiple directions, and/or spatial arrays of microelectrodes. These recordings show that except for a few milliseconds after the first cortical sensory response, multiple overlapping sources are active. For the commonly-studied cognitive potential components, N400 and P3b, iEEG demonstrates distributed sources in multiple lobes with similar time-courses. These data, as well as basic cortical neurophysiology from animal studies, do not support the common assumptions that the MEG generating sources are focal, and/or independent. Although focal hemodynamic activation is often described, this may be an artefact of the usual data analysis schemes. In summary, MEG source localization depends on prior assumptions of unknown accuracy, and MEG is insensitive to much cerebral activity. MEG publications should explicitly acknowledge these limitations. If possible, reference should be made to more certain knowledge, which in some cases includes iEEG.