Figure 4.
A, B, Model ventral striatal weights (WiS) (A) and phasic dopamine cell responses (above baseline) to conditioned stimuli (B) after learning. Activations are normalized by the activation elicited by the conditioned stimulus predicting the largest expected value. Responses to conditioned stimuli elicit cell activation as a function of expected reward value, p(R*|CS)|R*|. Model neurons' activations discriminate among expected reward values. Black, gray, and white squares show the outcomes of training with reward magnitudes, |R*|, of 0.05, 0.15, and 0.50, respectively. B, Inset, The corresponding neurophysiological data [from Tobler et al. (2005), their Fig. 1C, reprinted with permission from AAAS], where the two curves (squares and circles) represent data from two different monkeys.