Dynamical latent state computation in the male macaque posterior parietal cortex

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Real-world task success relies on dynamically tracking hidden environmental states.
• Neural populations may estimate these states via recurrent interactions, reflecting an internal world model.

Purpose of the Study:

• To investigate how neural populations in the posterior parietal cortex (PPC) encode and track hidden states during navigation.
• To determine if recurrent neural interactions in PPC contribute to an internal world model for state estimation.

Main Methods:

• Recorded neural activity in the PPC of monkeys navigating a virtual environment using optic flow.
• Analyzed neural population dynamics, interneuronal interactions, and decoding of hidden states (displacement from goal).
• Introduced task manipulations to perturb the world model and observed effects on neural representations.

Main Results:

• Identified sequential neural dynamics and strong interneuronal interactions in PPC.
• Found that the hidden state (displacement from goal) was encoded in single neurons and decodable from population activity.
• Observed that task manipulations altering the world model significantly changed neural interactions and hidden state representation, while sensory/motor representations remained stable.

Conclusions:

• Neural interactions in the PPC embody a world model that consolidates information and tracks task-relevant hidden states.
• Task demands shape neural interactions, influencing the formation and function of internal world models.
• Recurrent processing in PPC is critical for dynamic state estimation during navigation.