Hippocampal subfields support human novelty detection via distinct signals
Detecting novelty is fundamental to adaptive memory, yet the neural comparisons that allow new events to be distinguished from prior experience are not fully understood. Hippocampal novelty signals are often framed as associative comparator responses that register violations of learned relationships. However, recognition memory models also predict a broader form of comparison: a global, item-based mismatch between an incoming stimulus and aggregate memory traces. Whether such global matching is implemented within human hippocampal circuitry, and whether it contributes to memory decisions, remains unknown. Using ultra-high-field 7T fMRI during dense sampling of a continuous object recognition task (76,800 trials across five sessions), we estimated both univariate novelty responses and multivariate global mismatch signals across hippocampal subfields. CA1 carried the novelty response that most reliably predicted accurate novelty detection, whereas the subiculum expressed a global mismatch signal that predicted detection independently of CA1. This subicular signal also shaped subsequent recognition behaviour, in which low global mismatch was associated with false alarms to novel objects that propagated into later hits. Finally, subicular global mismatch scaled with the experienced similarity of past events, consistent with an item-based comparison against accumulated memory. These findings identify distinct hippocampal mismatch signals for novelty detection and extend the human hippocampal comparator role beyond local associative mismatch to include global, item-based matching.