Figure 3.

Bidirectional responses of model phasic dopamine neurons reflect deviations from the outcome expected after training with reward magnitudes |R*| = 0.05 (triangles) and |R*| = 0.15 (circles) with a deterministic schedule p(R*|CS) = 1.0 (black lines), and after training with probabilistic schedules p(R*|CS) = 0.25 (squares), p(R*|CS) = 0.50 (diamonds), and p(R*|CS) = 0.75 (crosses) with a reward magnitude |R*| = 0.15 (gray lines). Responses are normalized to the response after the unpredicted delivery of a reward of magnitude 0.15. Inset, The neurophysiological data [from Tobler et al. (2005), their Fig. 3B, reprinted with permission from AAAS] after training the monkey with a 0.15 ml reward with p(R*|CS) = 1.0, equivalent to the model response shown with black circles.