Multiple systems in macaques for tracking prediction errors and other types of surprise

Area of Science:

• Neuroscience
• Animal Behavior
• Cognitive Science

Background:

• Learning involves adjusting predictions based on prediction errors or surprising events.
• Current models often focus on average reward prediction errors, potentially overlooking other surprise detection mechanisms.

Purpose of the Study:

• To investigate distinct neural mechanisms for detecting different types of surprising events in non-human primates.
• To differentiate between scalar reward prediction errors and rare reward events.

Main Methods:

• Six macaques were trained to associate visual stimuli with varying juice rewards.
• Functional magnetic resonance imaging (fMRI) was used to observe neural activity during exposure to three types of surprises: reward amount, rare reward, and visuospatial.
• Analysis focused on dissociating neural responses to scalar reward prediction errors versus rare reward events.

Main Results:

• Dopaminergic midbrain activity specifically reflected surprise related to the reward amount.
• Lateral prefrontal cortex showed a general role in detecting various surprising events.
• Posterior lateral orbitofrontal cortex uniquely identified rare reward events in learnable contexts, independent of average reward expectations.

Conclusions:

• The brain employs distinct neural pathways to process different categories of surprising events.
• Specialized regions, like the posterior lateral orbitofrontal cortex, are crucial for detecting specific types of learning-relevant surprises, such as rare reward events.