Neurobehavioural Correlates of Surprisal in Online Comprehension
Project A1
This project has used psycholinguistic experimentation and explicit computational modelling to establish that comprehension-centric surprisal—expectations driven by both knowledge about the world and linguistic experience—is reflected in the P600 component of the Event-Related Potential (ERP) signal. An overarching goal of the third phase is to generalize our previous findings to the processing of more naturalistic texts, and further to identify factors that may dynamically determine surprisal, such as depth of processing and domain knowledge. To address these questions, we will first establish the degree to which the P600 covaries with reading times—which we take as a behavioural index of comprehension-centric surprisal—during the continuous reading of naturalistic texts (WP1). The result of WP1 will essentially be a corpus of short texts, which includes the dependent neurophysiological and reading time measures for each word, as well as subject-specific comprehension and recall performance. This dataset will then be used as the basis for deriving a comprehensive taxonomy of predictors (WP2) for the processing indices recorded for each word—crucially allowing us to assess how to best determine comprehension-centric surprisal in order to estimate P600 and RT measures on a word-by-word basis (e.g., in terms of a combination of bottom-up n-gram surprisal and top-down ‘world knowledge’-driven expectations; Venhuizen et al., 2019a)—while also identifying and quantifying the influence of other predictors, such as association with the context, as well as behavioural performance. Based on the outcomes of WP1 and WP2, we will more directly examine how task and expertise modulate not only surprisal effects (WP3), but differentially influence the depth of comprehension and subsequent recall. Finally, we seek to refine the neurocomputational model developed in the previous phases (WP4) with a notion of depth of processing, and the consequence for surprisal, in order to account for findings from the experimental work packages.
Keywords: comprehension, scripts, schemas, events, discourse, surprisal,event-related potentials, eye-tracking
Publications A1
Other Area-A Projects
Successfully