Figure 6.

The lag between rule learning and execution varies across animals. (a) Relation between the onset of rule prediction (first peak in Error Prediction index) and the onset of rule execution (estimated by combining the time points at which hit index, target distance and PR score exceed their respective critical thresholds) for five animals across original task (black data points), first reversal (dark grey) and second reversal (light grey). Note that data points are concentrated above the diagonal. (b) Lag between the onset of rule prediction and rule execution (see panel a) for all five animals in original task (black diamonds), first reversal (dark grey diamonds) and second reversal (light grey diamonds). Arrows: Mean lag across animals for original task (black) and two reversals (dark and light grey). Dashed line: Zero lag. (c) Left panel: Relationship between Error Prediction index (indicating rule acquisition) and hit index (indicating successful rule execution) for one animal across the course of the second rule reversal training. Trials later in the training are drawn in warmer colours. Right panel: Same for another animal in the first rule reversal training. In the example on the left, rule prediction markedly precedes rule execution, while on the right, rule prediction and execution improve largely synchronously.