The impact of predicted and unpredicted events on the spatial priority map: evidence from multivariate pattern analyses in EEG

Under the framework of predictive coding, perception is a constant interplay between the brain’s prediction of upcoming events and incoming sensory information. One way that predictions shape perception is through the allocation of attention to locations that will likely contain relevant information. Spatial attention is assumed to be allocated according to the weights of an assumed integrated spatial priority map, but little is known about how this priority map changes in response to predictive processes. We used the high temporal resolution of electroencephalography (EEG) in combination with time-resolved multivariate pattern analyses to investigate the influence of predictions on the spatial priority map. We showed participants sequences of Gabor patches that appeared in a clockwise or counter-clockwise pattern of six locations, with varying spatial frequencies and orientations. Crucially, at random moments some Gabor patches were omitted from the sequence, while others violated expectations regarding the upcoming location of the subsequent Gabor patch. In combination with an independent localizer, this design allowed us to establish how and when the present shapes future representations in the context of established spatial associations. The EEG decoding results revealed that omitted anticipated locations were internally represented at their expected moments in time. Furthermore, we observed strong modulations in decoding accuracy for location violations. Together these findings advance our understanding of how we utilize recurrent patterns from past experiences to forecast future occurrences.