Hippocampal prediction errors and claustral control at the perception–memory crossroads

Prediction errors (PEs)-the mismatches between what the brain expects and what it encounters-drive the continual alignment of perception with memory. Here, we review cross-species evidence showing that this cycle depends on hippocampal predictions that bridge sensory perception and declarative memory. The hippocampus generates probabilistic forecasts via pattern completion and separation, while the claustrum-a multisensory hub-constrains those forecasts with integrated perceptual evidence, jointly shaping the magnitude of the resulting PE signal. We combine two explanatory lenses: Bayesian inference, which prescribes how strongly a PE should update prior beliefs, and connectionist models, which reveal how local, neuromodulator-gated plasticity approximates that normative update in real circuits. Dopaminergic and cholinergic signals tune PE precision and experimental inactivation of the claustrum in mice impairs hippocampal-dependent memory reconsolidation, establishing causality in this perception-to-memory loop. We propose that aberrant PE calibration accounts for cognitive symptoms in Alzheimer's disease, schizophrenia and learning disorders. We also outline pharmacological and circuit-level strategies to normalize PEs, and a research agenda combining ultra-high-field imaging, causal manipulations and hybrid Bayesian-connectionist modelling. Clarifying how the hippocampus and claustrum cooperate to align perception with memory promises biomarkers and precision therapies for cognitive disorders. This article is part of the theme issue 'The role of hippocampal predictions in cognition: bridging perception and memory'.