Fixation-locked hippocampal activity reflects semantic content and temporal order of visual exploration during scene encoding

Memory formation relies on the hippocampus and unfolds over time across experience, such as during the visual exploration of complex, naturalistic scenes. Eye movements evoke hippocampal activity, including fixation-locked field potentials and phase resets of theta oscillations. This suggests that hippocampal encoding is temporally structured by the sequence of visual fixations. Because eye-movement sequences sample semantically meaningful portions of scenes, they provide temporal structure to semantic content in memory. However, it remains unclear how the semantic content and temporal order of fixations jointly shape medial temporal lobe activity. We therefore tested whether intracranial EEG recordings from human hippocampus and amygdala reflect the semantic content and temporal order of individual fixations during encoding of naturalistic scenes. Relative to other semantic content, fixations on people were particularly relevant for memory, with the first fixation on a person predicting subsequent scene recognition. Fixation-locked hippocampal responses were enhanced for fixations to people relative to other semantic content, expressed in both larger fixation-evoked potentials and stronger theta phase locking. These effects were strongest for the first fixation relative to subsequent fixations. Theta phase locking was also enhanced in both hippocampus and amygdala for first fixations on people relative to later fixations and to other semantic content. These findings indicate that medial temporal lobe activity is structured by discrete fixation-level events during scene encoding, suggesting that theta-paced sampling contributes to the transformation of semantic and temporal components of visual experiences into memory. Significance StatementThis study shows that the semantic content and order of eye fixations jointly influence human hippocampal activity during memory encoding. Combining intracranial recordings, eye-movement tracking, and deconvolutional modeling, we show that the first glance at a person within naturalistic scenes is a privileged event, associated with increased hippocampal activity, theta-phase resetting in hippocampus and amygdala, and subsequent memory success. These findings recast eye movements not as mere motor acts, but as an important process that helps medial-temporal structures prioritize and integrate behaviorally relevant information into episodic memory.