Reward-evoked dopamine transients are well-established as prediction errors. However the central tenet of temporal difference accounts-that similar transients evoked by reward-predictive cues also function as errors-remains untested. Here we addressed this by showing that optogenetically-shunting dopamine activity at the start of a reward-predicting cue prevents secondorder conditioning without affecting blocking. These results indicate that cue-evoked transients function as temporal-difference prediction errors rather than reward predictions.

The discovery that DA transients can be mapped onto the reward prediction errors in temporal difference models is a pinnacle achievement of neuroscience. Yet, there is abundant evidence that DA activity reinforces actions, suggesting it serves as an intrinsically rewarding event. These two possibilities are so conceptually intertwined that it is not surprising that they have been so far experimentally conflated. Here, using computational modeling, behavioural blocking and optogenetics, we show that stimulating VTA DA neurons promotes learning even when a natural reward and DA stimulation are held constant across the learning phases of blocking. These findings provide strong evidence in favour of the prediction error hypothesis rather than encoding the rewarding value of appetitive events.