Dopaminergic action prediction errors serve as a value-free teaching signal

Area of Science:

• Neuroscience
• Animal Behavior
• Computational Neuroscience

Background:

• Animal choice behavior involves seeking rewards and repeating actions.
• Dopamine signals, including reward prediction error and action prediction error, are theorized to reinforce these distinct behavioral tendencies.

Purpose of the Study:

• To investigate the role of movement-related dopamine activity in the striatum.
• To determine if action prediction errors act as a value-free teaching signal for learning.

Main Methods:

• Auditory discrimination task in mice.
• Measurement of movement-related dopamine activity in the tail of the striatum.
• Causal manipulation of neural circuits.
• Computational modeling.

Main Results:

• Movement-related dopamine activity in the tail of the striatum encodes action prediction errors.
• Action prediction errors function as a value-free teaching signal, reinforcing repeated associations.
• Action prediction errors alone do not support reward-guided learning but consolidate associations when paired with reward prediction error circuitry.

Conclusions:

• Two distinct dopaminergic prediction errors (reward and action) operate synergistically to support learning.
• These signals reinforce different types of associations in distinct striatal regions, contributing to flexible and stable behavioral choices.