Direct Evidence for Two Different Neural Mechanisms for Reading Familiar and Unfamiliar Words: An Intra-Cerebral EEG Study

After intensive practice, unfamiliar letter strings become familiar words and reading speed increases strikingly from a slow processing to a fast and with more global recognition of words. While this effect has been well documented at the behavioral level, its neural underpinnings are still unclear. The question is how the brain modulates the activity of the reading network according to the novelty of the items. Several models have proposed that familiar and unfamiliar words are not processed by separate networks but rather by common regions operating differently according to familiarity. This hypothesis has proved difficult to test at the neural level because the effects of familiarity and length on reading occur (a) on a millisecond scale, shorter than the resolution of fMRI and (b) in regions which cannot be isolated with non-invasive EEG or MEG. We overcame these limitations by using invasive intra-cerebral EEG recording in epileptic patients. Neural activity (gamma-band responses, between 50 and 150 Hz) was measured in three major nodes of reading network - left inferior frontal, supramarginal, and inferior temporo-occipital cortices - while patients silently read familiar (words) and unfamiliar (pseudo-words) items of two lengths (short composed of one-syllable vs. long composed of three-syllables). While all items elicited strong neural responses in the three regions, we found that the duration of the neural response increases with length only for pseudo-words, in direct relation to orthographic-to-phonological conversion. Our results validate at the neural level the hypothesis that all words are processed by a common network operating more or less efficiently depending on words' novelty.