Figure 6.

Adaptive Prediction Error Coding Is Consistent across Positive and Negative Prediction Errors

(A) Average (± SEM) RPE slope magnitude was significantly higher for negative compared to positive RPEs (results collapsed over the SN/VTA complex and ventral striatum).

(B) Average (± SEM) estimated learning rates were significantly higher for negative compared to positive RPEs.

(C) The average (± SEM) degree of neural adaptation did no differ significantly between positive and negative RPEs in the SN/VTA complex and ventral striatum.

(D) Significant positive correlation between the effect of RPE error sign on learning rates and its effect on RPE coding slopes.

n-linear, non-linear; SN/VTA, substantia nigra/ventral tegmental area; Vstr, ventral striatum; LR, learning rate; Ψ, neural adaptation.