A Probability Distribution over Latent Causes, in the Orbitofrontal Cortex.

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Computational Psychiatry

Background:

• The orbitofrontal cortex (OFC) is involved in representing task states for decision-making and schemas for episodic memory.
• Both functions require inferring hidden causes that generate observed data, optimally solved using Bayesian inference.
• Previous research implicates the OFC in both state and schema representation, but the underlying computational mechanism remains unclear.

Purpose of the Study:

• To investigate whether the orbitofrontal cortex (OFC) represents a posterior probability distribution over latent causes.
• To test if this representation supports inference in tasks relevant to decision-making and memory.
• To compare the OFC's representation against alternative computational hypotheses.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to measure brain activity in human participants.
• Participants performed a task requiring them to infer a probability distribution over four possible latent causes based on observations.
• Pattern similarity analyses were conducted on blood-oxygen-level-dependent (BOLD) activity in the OFC.

Main Results:

• BOLD activity patterns in the OFC were best explained by a log-transformed posterior probability distribution over latent causes.
• This representation in the OFC was more accurate than alternative models, including the most probable cause or uncertainty.
• The findings suggest the OFC encodes probabilistic beliefs about unobserved environmental factors.

Conclusions:

• The orbitofrontal cortex (OFC) encodes probabilistic representations of latent causes, supporting Bayesian inference.
• This mechanism likely underlies the OFC's role in both reinforcement learning/decision-making and episodic memory.
• The study provides neural evidence for a unified computational principle across diverse cognitive functions in the OFC.