Purchase this article with an account.
Amirsaman Sajad, Jeffrey D Schall; Cortical microcircuitry encoding expected utility and reward prediction error for visually guided saccades. Journal of Vision 2020;20(11):325. doi: https://doi.org/10.1167/jov.20.11.325.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Cognitive control requires evaluating the outcome of actions relative to expectations, then updating a model of the world when expectations are violated (i.e., prediction error). Medial frontal cortex (mFC) plays a key role in these functions and is a likely source for associated event-related scalp potentials. However, the microcircuitry is unknown. To address this, we sampled neurons across all layers of the supplementary eye field, an area in mFC involved in monitoring visually-guided eye movements. Neural discharges were recorded from two monkeys while performing the saccade stop-signal task. On most trials, monkeys were rewarded for looking at a peripheral visual stimulus, but occasionally a stop-signal instructed them to inhibit the gaze shift. Each direction was associated with low or high reward amounts and this association reversed unpredictably after ~20 rewarded trials. Thus, the first rewarded trial in each block was associated with the highest and the following trials with progressively lower reward prediction error (RPE). As expected, the response time (RT) of both monkeys showed sensitivity to reward amounts. Saccade RT to high-reward targets was significantly faster than that for low-reward targets. Upon block reversal, RT adaptation was observed with progressive reduction of RT to the high-reward target and elevation to the low-reward target. Negative RPE (i.e., lower than expected reward) was associated with a relative facilitation of neurons in layers 2 and 3 (L2/3) and suppression of neurons in L5/6. Neurons signaling higher than expected reward (positive RPE) exhibited a complementary laminar pattern with higher proportion of facilitated neurons in L5/6 and suppressed neurons in L2/3. These results reveal the laminar microcircuitry underlying value and prediction error encoding and complement our previous report of differential contributions of upper and lower layers to encoding negative and positive outcomes.
This PDF is available to Subscribers Only