Fig. 4.

Dopaminergic modulation of saccadic choices. This figure illustrates the effect of dopaminergic modulation of decision making during the delay period task. Each row shows a different level of prior precision over policies (highest for the first row and lowest for the last). Simulated dopaminergic firing rates are shown on the left (note the differences in axis ranges). When the precision is very low, the selected saccadic target is random, as all possible saccadic policies become (nearly) equally probable. As the prior precision is increased, the first notable change occurs in the dopamine plots (third row). Here, there is a decrease in dopamine firing during the first saccade, as uncertainty about the policy pursued increases. This is because this saccade is inconsistent with the policy consistent with reaching the target. Having committed to the incorrect policy, there is a dopamine spike coinciding with a confident inference that this policy is being pursued. As dopamine levels increase further, this spike moves earlier, as the saccade performed (although still premature) is still consistent with reaching the target. When sufficiently high, dopamine levels show little change throughout the trial, with the correct policy inferred quickly and confidently from the first time-step. The key message to take away from this figure is that, in the absence of dopamine, oculomotor decisions become increasingly random. This is because the distribution over action sequences becomes less precise. As the precision tends towards zero, all plans of action become equally probable