Evidence for a constant occipital spotlight of attention using MVPA on EEG data.

While traditional behavioural and electroencephalographic studies claim that visuospatial attention stays fixed at one location at a time, recent research has rather shown that attention rhythmically fluctuates between locations at different rates. However, little is known about the temporal dynamics of this fluctuation and whether it changes over time. We addressed this question by investigating how the neural pattern of visuospatial attention behaves over space and time by employing classification and conventional analysis of occipito-parietal EEG activity. Furthermore, we simulated data with the attentional electrophysiological correlates to control for the ground truth that would give rise to certain classification patterns. We analysed two visuospatial cueing tasks, with a peripheral and a central cue to control for sensory-driven processes, where attention was covertly oriented to the left or right hemifield. First, to decode the spatial locus of attention from neural activity, we trained and tested a classifier on every timepoint from the attentional cue to the stimulus onset. This resulted in one temporal generalization matrix per participant, which was time-frequency decomposed to identify the sampling rhythm. Independently, we calculated a lateralization index based on ERPs and alpha-band power and correlated these indices with classifier performance. Eventually, we simulated two dataset, with ERPs and alpha-band attentional modulations, and employed the same decoding approach. Our results show that attention settled on the cued hemifield in a late time window, but an early and rhythmic sampling of both hemifields exclusively after the peripheral cue. Only the ERP lateralization index correlated with classifier performance in the periperhal cue dataset, suggesting that the early rhythmic state did not reflect attentional sampling but instead was driven by the cue location, idea also supported by our simulations. Together, our results characterise the occipital attentional sampling as a constant process slightly delayed after the cue.