There has been much debate about whether the inclusion of possible punishments is beneficial for better performance and learning. To understand the neuronal mechanisms of the impact of aversive stimuli on decision making, we trained two monkeys (Macaca fascicularis) a choice saccade task. In the task, three fractal images were separately associated with a rewarding juice (R), a neutral tone with a small reward (T), or an aversive airpuff (A). After foveating a central fixation point (FP), a pair of two images; R-T, R-A, or T-A, appeared in the left and right of FP, followed by an eye movement to one of the images to obtain a reward and/or to avoid a punishment.
We found that both animals often showed deviation from the optimal choice in T-A trials even after extensive training. To further examine the effect of the prospect of possible outcomes, one monkey was trained for the same task in the block design; the same image pair was repeated 8~10 trials and the pair was changed pseudo randomly. We found that the inclusion of 'A' significantly affected the behavior even the optimum choice was identical; the rate of optimum choice for R-A was lower than that for R-T. Reaction times were the shortest in R-A while the variability was highest for T-A.
The basal ganglia (BG) are thought to be involved reward-based motor control, learning, and decision making. Then, are there the neuronal mechanisms of the BG for the decision making in the situation where punishments may occur, and if so, are they independent or integrated with the reward system? We found that 52 (43%) out of 122 task-related neurons in the primate caudate nucleus (CD) showed differential modulation in the pre-target fixation-period activity depending on the expected image pair. Among them, 28 neurons showed differential activity depending on the suboptimum image even the optimum image was the same (i.e. R-T vs. R-A), indicating the impact of the inclusion of an aversive image. Further, nearly half of them (13/28) did not discriminate R-A and T-A indicating that they are independent of reward circuits. These results indicate a group of CD neurons are affected by aversive information, sometimes independent of reward process.