A normalization circuit of attention in primate lateral prefrontal cortex

The way in which visual neurons encode information pertaining to a cluttered scene with multiple stimuli, and subsequently filter behaviorally relevant information using attention remains poorly understood. Neurons of area 8a in the macaque lateral prefrontal cortex have been shown to encode visual and attentional signals. We trained two macaque monkeys in a visuospatial attention task and performed neurophysiological recordings to test how neurons in this area encode multiply presented stimuli and attentionally filter target stimuli from distractors. We found area 8a neuronal responses to several concurrently presented stimuli to resemble the average of individual responses to those stimuli when presented alone; this nonlinear response is characteristic of divisive normalization, a canonical brain computation seen to operate in various neural systems. Interestingly, the strength of normalization was dependent on visuospatial tuning, with neurons tuned for the ipsilateral visual hemifield displaying stronger normalized responses than those tuned for the contralateral hemifield. Furthermore, when presented with multiple stimuli and attending toward a target stimulus lying in the receptive field, contralateral-tuned neural activity increased and resembled that of when the target was presented alone (i.e. Winner-take-all response), whereas ipsilateral-tuned neurons were less modulated by attention and remained best-described by an average response. Taken together, our findings suggest a normalization circuit underlying attention in the primate lateral prefrontal cortex