John E. Kiat¹ (), Steven J. Luck¹; ¹University of California-Davis

Prior research has shown that covert attention enhances the perceptual processing of attended locations and stimuli, boosting behaviorally measured discriminability and neural response amplitudes. However, the time point during perceptual processing at which attended information becomes more discriminable than unattended information is unclear. Magnitude-based neurophysiological measures with high temporal precision (e.g., the N2pc or Alpha Power Suppression) provide useful information about attention but have important limitations. For instance, observing an N2pc in response to a visual search target is certainly evidence for when covert attention becomes focused on the target, but it does not provide direct evidence for when information about the attended object becomes more discriminable than information about unattended objects. To address this issue, we applied multivariate pattern analysis to ERPs to analyze the timecourse of the separability of the neural representations of covertly attended versus unattended naturalistic scene images. Participants (N = 28) were presented with a series of trials, with each trial consisting of two different scene pairs presented simultaneously at fixed locations. Each pair was preceded by a cue indicating which of the two scenes should be attended. We found greater representational similarity between the ERPs and a computational model of early visual processing for the attended scenes beginning at 80 ms poststimulus. At approximately 250 ms, we found greater representational similarity between the ERPs and a computational model of more abstract visual representations as well as greater categorical decoding accuracy for the attended scenes. These findings show that covert attention can boost the neural representation of simple visual features during early visual processing, with more categorical-based separability likely emerging later.

Acknowledgements: Grant support: This work was made possible by grant R01EY033329to SJL.