Acetylcholine: a candidate substrate for hippocampal predictive learning?

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Acetylcholine in the hippocampus is linked to various functions like learning, memory, and arousal.
• A unified framework for understanding cholinergic function is currently lacking.
• Predictive models offer a new perspective on neural functions, viewing the brain as a prediction engine.

Purpose of the Study:

• To propose a unifying framework for hippocampal acetylcholine function within predictive coding.
• To frame acetylcholine as a signal for state transition prediction errors.
• To elucidate the role of theta sequences and the septo-hippocampal circuit in this process.

Main Methods:

• Theoretical modeling and conceptual framework development.
• Integration of existing knowledge on acetylcholine, theta oscillations, and predictive coding.
• Hypothesizing a mechanism involving theta sequences and the septo-hippocampal circuit.

Main Results:

• Hippocampal acetylcholine is proposed to encode state transition prediction errors.
• This mechanism parallels dopamine's role in reward prediction error signaling.
• Theta sequences are suggested as a substrate for calculating these errors and updating internal models.

Conclusions:

• Cholinergic signaling in the hippocampus guides structural learning by signaling state transition prediction errors.
• This provides a unified mechanistic account for diverse cholinergic roles, including novelty detection and hidden state inference.
• The framework emphasizes learning predictive internal world models through precisely-timed neural sequences.