Brain activity during proverb reading

SHALOM DE; BIANCHI B; KAMIENKOWSKI JE

Congreso; XVI SAEL; 2018

As in almost every daily visual task, the brain generates a prediction on the forthcoming stimuli. In reading, this prediction is usually operationalized as the Predictability, i.e. the probability of knowing a future word before reading it. These predictions could be built on different factors depending on the stimuli, such as syntactic, semantic, phonological relations with the context or even memory retrieval of known sentences.In the present study, we aimed to separate the memory encoded contribution to the predictability using proverbs and common sentences on a Serial Visual Presentation EEG experiment. The data was first analyzed on a classical ERP analysis, finding the well described N400 effect for predictability, but without a robust effect of the sentence type. We attribute this lack of effect to the average, where the proverbs? words are taken all together. In order to overcome this issue (and also the lack of control over the other covariables), we implemented a Linear Mixed Models (LMM) for each sample (electrode and time-point). Although, this type of analysis comes with a cost: it implies to run a lot of models. Thus, we corrected for multiple comparisons and extracted global statistical measures by embedding the LMMs results for each sample in a cluster-based permutation approach. Since the current literature on cluster-based permutation procedures [Maris and Oostenveld, 2007] is based on univariate statistical tests we explored different possibilities for the implementation of permutations for our multivariate data.Using this procedure we observed the classical predictability effect (N400), a conspicuous late effect of the word position in sentence and an interaction effect of the sentence type (proverb vs non-proverb) with the predictability, which clearly supported a classical N400 effect for the common sentences, and no effect for the proverbs. Moreover, we observed differences in lower frequency bands (theta and alpha) distinguishing between the sentence types. Altogether, these results support different prediction mechanisms.