Human single neurons in the medial temporal lobe encode inferred relational structures

Humans can infer never-experienced global relational structures (e.g., "A>D”) by observing and integrating multiple local relations (e.g., "A>B", "B>C", “C>D”). This capability for “transitive inference” (TI) depends on the medial temporal lobe (MTL). But the mechanisms by which the MTL performs TI remain poorly understood. Previous work suggested that “concept cells” (CCs), which are neurons that selectively respond to preferred concepts (e.g. “A”), might contribute to relational learning and potentially TI by widening their tuning to new concepts that co-occurred with initially response-eliciting concepts during learning of local relations (“A>B”). To examine this hypothesis, we recorded single-unit activity in the MTLs of 18 epilepsy patients implanted with depth-electrodes for chronic seizure monitoring while they performed TI. Contrary to our hypothesis, preliminary results suggested that CCs which initially responded to only specific items retained their tuning and did not extend their responses to new but related items (e.g., “B” in addition to “A”) following learning of local relations (“A>B”) during TI training. Instead, we observed the emergence of new CCs during TI training that mapped the global relational structure by tuning their responses to similar, adjacent local relations (e.g., preferred responding to “A>B”, “B>C”, and “C>D”, but not “D>E”). Importantly, this was only observed in sessions where patients successfully performed TI. These preliminary findings suggest that more complex mechanisms than a mere retuning of existing CCs in the MTL are involved when humans learn to infer global relational structures.