Abstract
Capacity limitations are ubiquitous in cognition; only so much information can be processed simultaneously. Primates compensate for capacity limits by prioritizing information processing based on their goals. For example, visuospatial attention compensates for limits in sensory processing by selecting a subset of stimuli for enhanced representation. Similarly, working memory has a limited capacity, and a ‘selection’ process akin to visual attention can selectively enhance a subset of items in working memory at the expense of others. However, neither the neural mechanism underlying selection nor its relation to attention is well understood. To explore this question, we trained two macaque monkeys to switch between an ‘attention’ and a ‘selection’ task. In both tasks, animals were presented with one or two colored squares (the samples) at different locations on a screen. After a memory delay, the animal reported one of the two items on a continuous scale. Critically, a spatial cue indicating which item was task-relevant appeared either before (prospective) or after (retrospective) the samples. Behavioral analyses indicated that prospective cues encouraged attention and retrospective cues encouraged selection. Report precision decreased with load on retrospective trials but not prospective trials, suggesting that the animals attended only the cued sample on prospective trials. Report accuracy increased when retrospective cues appeared earlier during the delay on two-item trials, suggesting that the animals prevented decay due to interference by selecting the cued sample in memory. To understand the neural mechanisms supporting attention and selection, we recorded neural activity simultaneously in prefrontal and parietal regions associated with attentional control. We identify neurons that encoded the locus of the task-relevant sample on prospective and retrospective trials and examine they extent to which these signals overlap across the brain.
Acknowledgement: MFP was supported by a National Defense Science and Engineering Graduate Fellowship and TJB was supported by a grant from the Office of Naval Research (ONR N00014-16-1-2085).