Researchers find that rare rewards amplify dopamine responses during learning

Dopamine responses are activated by rare rewards. A) Schematic drawings of the uniform (left) and normal (right) reward probability distributions. The expected values (EV) and positive and negative prediction errors (+PE and -PE, respectively) are all identical in both distributions. The only difference between the reward distributions is how frequently +PE and -PE occur. B) The responses from one dopamine neuron to identical rewards drawn from uniform (green) and normal (magenta) reward distributions. The top section shows Peri-Stimulus Time Histograms (PSTHs) aligned onto reward delivery (vertical dashed lines). The bottom section shows raster plots where every tick mark represents the timing of an action potential. The raster plots and PSTHs show that the larger rewards evoke positive prediction error response – the activations above zero – whereas smaller rewards evoke negative prediction error responses – the responses that dip below zero. Note that, despite + and – PE being identical in uniform and normal distributions, the responses to rewards drawn from the normal distribution are amplified, relative to the responses to rewards drawn from the uniform distribution. Thus, rare rewards amplify dopamine responses. Credit: Rothenhoefer et al.