Prospective contingency explains behavior and dopamine signals during associative learning.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Behavioral Science

Background:

• Associative learning is fundamental to behavior.
• The neural basis of how stimulus-outcome contingency influences behavior is not fully understood.
• Dopamine signaling in the ventral striatum is linked to associative learning.

Purpose of the Study:

• To investigate the neural mechanisms of contingency in associative learning.
• To examine dopamine activity in the ventral striatum during a Pavlovian contingency degradation task in mice.
• To evaluate computational models explaining the observed dopamine responses.

Main Methods:

• Mice were subjected to a Pavlovian contingency degradation task.
• Anticipatory licking behavior was recorded.
• Dopamine activity in the ventral striatum was measured.
• Computational models, including temporal difference (TD) learning, were used to interpret results.

Main Results:

• Anticipatory licking and dopamine responses to a conditioned stimulus decreased when rewards were delivered uncued.
• These responses remained unchanged if additional rewards were cued.
• Results contradicted traditional contingency accounts and a causal learning model (ANCCR).
• Temporal difference (TD) learning models with intertrial interval state representation explained the findings.

Conclusions:

• Dopamine activity in the ventral striatum reflects temporal difference errors in associative learning.
• TD learning models provide a framework for understanding how contingency influences dopamine signals and behavior.
• Recurrent neural networks within a TD framework can develop relevant state representations.