Partenaires et organisations internationales
(Anglais)
|
Marseille, Villigen, l'Aquila, Cambridge
|
Résumé des résultats (Abstract)
(Anglais)
|
Midbrain dopamine (DA) neurons are phasically activated by unpredicted primary rewards (appetitive US) and by conditioned, reward-predicting stimuli (appetitive CS). However, they are not activated by predicted rewards. During learning, a transfer of the dopamine response occurs from the US to the CS. According to associative learning theories, only unpredicted reinforcers contribute to learning. This is based on the blocking paradigm in which a new stimulus is added to a fully conditioned stimulus. The new stimulus will not acquire associative strength because the reward is already predicted by the conditioned stimulus. In order to formally assess the relationships of midbrain dopamine neurons to the unpredictability of rewards, we tested their activity in a blocking experiment.
In a first step, we tested the discriminative capacities of dopamine neurons. In a classically conditioned discrimination task, monkeys were presented with two randomly alternating pictures which were either followed by liquid reward (A+) or not (B-). Animals discriminated well between the two stimuli, as documented by their licking behavior. DA responses were stronger for rewarded stimuli as compared to non-rewarded stimuli, both in terms of percentage of neurons responding and magnitude of responses.
As many dopamine neurons discriminated between rewarding and non-rewarding stimuli, we tested blocking. A new stimulus was simultaneously presented together with each previously experienced stimulus, thus producing AX+ and BY+, which were both followed by a reward. We then tested whether X and Y were conditioned. Licking behavior showed that prior pairing of stimulus A+ with a reward blocked the conditioning of stimulus X. In contrast, the previously unconditioned stimulus B did not block conditioning of Y. The electrophysiological responses of DA neurons showed a comparable blocking effect. More DA neurons were activated with higher magnitudes by the non-blocked stimulus Y as compared to the blocked stimulus X.
In conclusion, blocked stimuli produce weaker response in DA neurons, probably because they have acquired less associative strength and do not serve as a reward predictors for the animal. These results are consistent with the hypothesis that DA neurons are sensitive to the predictability of primary rewards. This would suggest that DA neurons code an error in the prediction of reward. Their response to unpredicted rewards would constitute an effective appetitive teaching signal.
The experiments are finished, and we are in the process of evaluating the data and preparing at least one publication. This work will serve as thesis (Dr. phil. II) for Pascale Waelti.
|
