An EEG study of entropy and surprisal during artificial grammar processing

TABULLO, ÁNGEL; PALLIER, CHRISTOPHE; WAINSELBOIM, ALEJANDRO; BACHRACH, ASAF

Nueva Orleans

Congreso; SOCIETY FOR PSYCHOPHYSIOLOGICAL RESEARCH 52nd Annual Meeting; 2012

SOCIETY FOR PSYCHOPHYSIOLOGICAL RESEARCH

Prediction seems to be a central mechanism in multiple cognitive domains and in particular in language. A wide range of behavioral and electrophysiological paradigms have analyzed the correlates of prediction during parsing (for a review see Van Petten and Luka, 2012). The main correlate of prediction that has been studied is prediction error or surprisal (Hale, 2001), or the degree to which an actually occurring event fits previous predictions. Violation of predictions at both lexical and syntactic level have been shown to modulate the N400 (Van Berkum et al., 2005; DeLong et al., 2005) and P600 (Kaan, 2000) components respectively. Another correlate of prediction, albeit   far less studied, is the level of uncertainty (or entropy) regarding an upcoming event. Entropy effects independent of surprisal have been previously studied outside the language domain (Bestmann, 2008), and in the context of sentence comprehension (Roark et al., 2009). However, the electrophysiological correlates of the separate effects of entropy and surprisal has not been attempted yet. In the present study, an artificial grammar was designed in order to manipulate entropy and surprisal at both rule (abstract) and item (surface) levels. Participants were exposed to the statistical properties of the stimuli, and their subsequent EEG responses were compared with entropy and surprisal measures by means of general lineal modelling. The objective of the study was to study the time-course and topography of entropy and surprisal modulations of EEG activity.