Polysemanticity in human hippocampal neurons

During podcast listening, human hippocampal single neurons show dense semantic codes with multiple tuning peaks, whose overdispersed isotropic geometry suggests interference-minimizing repulsive coding without grid-like hexagonal structure.

• Why it matters: The work addresses whether hippocampal neurons use polysemantic representations—multiple semantically distant features per neuron—to support capacity, robustness, and contextualization.
• What they did: Using single-neuron recordings while participants listened to podcasts, the study quantified semantic tuning by identifying multiple stimulus-driven peaks per neuron and analyzing the geometry of the resulting population code across contexts.
• The result: Neuron responses were strongly context-modulated while population geometry stayed stable and generalizable, implying pattern separation can occur without losing generalizability under contextual changes.

AI summary of the abstract · may contain errors · not medical advice · read the paper.