​​​​​​​Talk from Ingmar Devries – Radboud University​​​

Short abstract:​​

For adaptive behavior, our brain needs to continuously predict unfolding external events. While predictive processing theories propose such dynamic prediction, empirical evidence is often limited to static snapshots and indirect consequences of predictions. I will present a series of studies in which I investigated the continuous predictions themselves. I did so by applying a dynamic extension to representational similarity analysis (dRSA) to capture neural representations of unfolding events across hierarchical levels of processing (from perceptual to conceptual), by investigating the match between a temporally variable stimulus model at a given time-point and the neural representation across time. I will present evidence for neural (MEG) predictions of observed action sequences across hierarchical timescales, with high-level features predicted earlier in time and low-level features predicted closer to real-time. Second, disrupting high-level configural priors by up-down inversion of the action sequences, redirects predictions from high- to mid-level features. Third, I extend these findings to naturalistic audio-visual free-view movie watching by showing earlier perceptual predictions compared to simple action sequences, and by showing that both neural predictions and anticipatory eye-movements emerge earlier with increased stimulus familiarity (i.e., second vs. first movie viewing). To conclude, I will show you how naturalistic continuous input is predicted across hierarchical timescales, and how high-level priors, stimulus complexity and familiarity advance both perceptual predictions and anticipatory eye-movements.