Prediction and memory: A predictive coding account.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The hippocampus is vital for episodic memory and online predictive processing.
• Existing frameworks suggest distinct, opposing hippocampal-neocortical interactions for memory recall versus prediction.
• Episodic recall involves hippocampal excitation of the neocortex, while prediction relies on cortical inhibition to suppress prediction errors.

Purpose of the Study:

• To reconcile the seemingly opposing roles of the hippocampus in memory and prediction.
• To propose a unified hypothesis for how the hippocampus facilitates both episodic memory and predictive processing.
• To explore the role of neuromodulatory gain in distinct hippocampal processing modes.

Main Methods:

• This is a hypothesis piece, integrating existing empirical and theoretical work.
• It proposes a novel framework for hippocampal function within predictive coding.
• The framework considers how neuromodulatory gain influences hippocampal-neocortical interactions.

Main Results:

• The hippocampus may inhibit neocortical prediction errors during prediction and increase their gain during memory recall.
• These distinct modes are proposed to depend on the neuromodulatory gain assigned to prediction error units.
• Memory recall is reframed as the use of 'fictive' prediction errors to train generative models without sensory input.

Conclusions:

• A unified model can explain the hippocampus's dual role in memory and prediction through modulation of prediction error signaling.
• Neuromodulatory gain is a key factor determining whether the hippocampus facilitates prediction or memory.
• This framework offers new perspectives on hippocampal function and its contribution to cognitive processes.