Prefrontal neural activity predicts and mitigates spatial uncertainty in a gaze task

To predict the future, the brain must integrate past with current sensory information. Research has suggested that prefrontal cortex predicts the timing of events (Fu et al., 2023). Here, we investigated if it also predicts spatial uncertainty. To do this, we recorded neural activity in the frontal (FEF) and supplementary (SEF) eye fields of two rhesus macaques, trained to saccade toward remembered visual targets (T) in presence of a landmark (L) that was surreptitiously shifted to a new position (L’) by a fixed amplitude in one of eight randomized directions arranged circularly around L. Previously, we showed that this results in retrospective shifts in FEF/SEF memory and gaze signals (Bharmauria et al., 2020, 2021). Here, we examined the period from the initial visual response to 300 ms after the landmark shift in 147/68 spatially tuned FEF/SEF neurons for prospective coding of this shift. We used a model-fitting technique to test memory delay coding along a T-T’ continuum. Remarkably, just before the landmark shift, SEF coded a shift toward T’. Since this direction was randomized, we hypothesized that SEF might be ‘guessing’ the direction of the shift. We tested this using a 2D analysis with the real shift (T’) rotated to the right and seven other imaginary shifts circularly arranged. Shortly after the visual response, response fields developed a donut-like prediction in all directions. This did not occur in shuffled controls and could not be accounted for by attraction toward landmark position (TL) or gaze error (TG). Eventually, after the real shift, this predictive ‘donut’ code shifted toward the actual L’. These data suggest that after thousands of training trials, the monkey brain, specifically SEF, created a guessing strategy based on learned probabilities and anticipation. This might allow the brain to optimize behavior and mitigate spatial uncertainty in the surrounding world.