Fig. 5.

Noradrenergic modulation of prefrontal firing. Each row illustrates a single trial of the delay period oculomotor task, but with different levels of noradrenaline. This has little effect on the performance of the task. Even in the lowest row, where a premature saccade takes place, this mistake is corrected at the next time-step. Note that this error is a consequence of the random sampling of actions from beliefs about policies and does not occur on the majority of trials. It has been retained here to illustrate the change in strategy that leads to the successful completion of the task. While there are no clear behavioural consequences of this manipulation, the physiological implications are much more striking. These are shown as raster plots of prefrontal cortical neurons representing the remembered target location. Each row represents the firing of a population of neurons representing the probability of one of the target locations at specific times throughout the trial. The lower rows within these plots indicate later times. This means that at the first time-step (first column), the last row represents beliefs about the future. By the final time-step (last column), the last row represents beliefs about the present. As the concentration of noradrenaline increases, more structure becomes apparent in the lower parts of the plots, indicating a more successful propagation of the inferences drawn from observing the initial cue to the later points in the trial. Note that the increase in persistent activity is accompanied by a decrease in the firing of other neurons, representing the probability of alternative states