Differential brain activity on mnemonic predictions during natural reading: an EEG and eye tracking co-registration study

BIANCHI B; LOREDO R; CARDEN JR; JAICHENCO V; VON DER MALSBURG T; SHALOM DE; KAMIENKOWSKI JE

Congreso; Society for Neuroscience; 2021

During reading our brain is continuously generating predictions on upcoming words. These predictions, if correct, could speed up the processing of the word when it is finally read. Also, they are a reflection of the ongoing integration of the text. It has been amply shown that Predictability (i.e. the estimated probability of guessing the next word) have an impact on how we move our eyes across the text: eye tracking (ET) experiments showed that more predictable words are fixated for shorter periods of time than less predictable words. Additionally, EEG experiments showed that the amplitude of the N400 potential, a hallmark of semantic processing, is modulated by the Predictability: the more predictable words are, the less amplitude the N400 has. Typically, EEG experiments were performed avoiding eye-movements, because the muscular contraction interferes with the brain electrical signal. But, during the last years, co-registration of EEG and ET have become a powerful technique for understanding our brain in more natural contexts, allowing eye movements during the task.Previous ET studies have shown that mnemonic predictions increase the skip probability and decrease the gaze duration on words. Meanwhile, previous EEG studies during Serial Visual Presentations of words have shown that the N400 potential is only present for predictions performed based only on linguistic context. In the present work, we analyzed brain potentials and eye movements during natural reading of Common and Memory-Encoded sentences (like proverbs), with the goal of understanding how different sources of predictions are used during natural reading. Firstly, we replicated the results from both experiments under the co-registration paradigm. Secondly, our results showed that during memory-encoded sentence reading there is also a decrease in regression and refixation probabilities. And thirdly, fixation-related potentials analysis using a novel Linear Mixed Models (LMM-CBP) approach showed that the classical N400 component is only present for common-sentence reading and for words previous from the point in which the Memory-Encoded sentences are recognized. These results indicate that the N400 potential could be associated only with predictions performed based only on linguistic context.