Covert attention increases the gain of stimulus-evoked population codes

AbstractCovert spatial attention has a variety of effects on the responses of individual neurons. However, relatively little is known about the net effect of these changes on sensory population codes, even though perception ultimately depends on population activity. Here, we measured the electroencephalogram (EEG) in human observers (male and female), and isolated stimulus-evoked activity that was phase-locked to the onset of attended and ignored visual stimuli. Using an encoding model, we reconstructed spatially selective population tuning functions from the pattern of stimulus-evoked activity across the scalp. Our EEG-based approach allowed us to measure very early visually evoked responses occurring ~100 ms after stimulus onset. In Experiment 1, we found that covert attention increased the amplitude of spatially tuned population responses at this early stage of sensory processing. In Experiment 2, we parametrically varied stimulus contrast to test how this effect scaled with stimulus contrast. We found that the effect of attention on the amplitude of spatially tuned responses increased with stimulus contrast, and was well-described by an increase in response gain (i.e., a multiplicative scaling of the population response). Together, our results show that attention increases the gain of spatial population codes during the first wave of visual processing.Significance StatementWe know relatively little about how attention improves population codes, even though perception is thought to critically depend on population activity. In this study, we used an encoding-model approach to test how attention modulates the spatial tuning of stimulus-evoked population responses measured with EEG. We found that attention multiplicatively scales the amplitude of spatially tuned population responses. Furthermore, this effect was present within 100 ms of stimulus onset. Thus, our results show that attention improves spatial population codes by increasing their gain at this early stage of processing.