Expectation suppression depends on werking memory but not perceptual load
No Thumbnail Available
To cope with the amount of information in the sensory world, predictive coding suggests that efficient observers suppress irrelevant hut predictable information. However, such expectation suppression (ES) occurs inconsistently across studies. Previous research raises the possibility that this might be explained due to the tasks towards which attention is diverted. To test this hypothesis, we manipulated subjects' attention and perceptual expectation during functional magnetic resonance imaging (fMRI). We exposed subjects to predictive and non-predictive audiovisual stimulus pairings while varying their available processing resources. Participants would either attend the (un-) predicted stimuli or their attention would be diverted to a task that increases perceptual or working memory load. We further aimed to explore the wholebrain correlates of perceptual expectations. Several brain regions, among which hippocampus, superior temporal gyrus, and cerebellum were more responsive to predictive than non-predictive pairings. Converging evidence suggests that superior temporal gyrus (STG) in particular responds to the informational value contained in predictive cues. Importantly, we found ES occurs for unattended stimuli, yet only under perceptual but not working memory load. This may explain why previous research failed to replicate ES. It furthermore dissociates expectation and repetition suppression, suggesting that the two are implemented through different physiological mechanisms. Predictive sequences furthermore induced increases in functional connectivity between STG, visual cortex and a thalamic region projecting to prefrontal cortex. We tentatively suggest that top-down modulations ofvisual perception and working memory share a common neural substrate, possibly in prefrontal cortex.
Faculteit der Sociale Wetenschappen