Saccades track visual associative memory processes with precision and sensitivity.

Humans primarily use vision to engage with and learn about the world. The hippocampus plays a crucial role in binding visual experiences of people, objects and contexts over time to create event memories. Thus, eye tracking could read out hippocampal dynamics in a precise and sensitive manner. Furthermore, eye tracking could potentially detect subjective memory decline reported by temporal lobe epilepsy patients that is missed by standardized cognitive testing. We asked whether eye movements could precisely and sensitively detect memory variability within trials and between subject cohorts. We predicted that (i) eye-tracking behaviour during visual retrieval could be validated against accuracy-based tests and that (ii) memory failures would be characterized by distinct spatiotemporal patterns of visual scanning. Fourteen healthy controls and 30 temporal lobe epilepsy patients participated in a visual object association task while eye movements and pupil size were recorded. We found a difference in accuracy during retrieval between healthy controls and temporal lobe epilepsy patients. Correct retrieval trials correlated with fewer saccades, early target preference, and a more organized search pattern. Eye-movement patterns could predict retrieval accuracy at the single trial level with outstanding performance, with percentage of gaze time on the target versus the lure as the most important features. Even during correct retrieval trials, temporal lobe epilepsy patients exhibited a more chaotic scanning pattern compared to healthy controls, suggesting a weaker memory trace. Healthy versus epilepsy diagnosis could be predicted with good performance, with trial entropy and pupillary changes as key predictive factors. Saccade patterns correlated with individual subjects' accuracy scores and performance on standardized cognitive tests but provided a greater range of performance. In summary, scanning behaviour provides a continuous measure of associative memory function that capture meaningful variability during trials, between trials, and between subjects. Thus, eye tracking could be a precise and sensitive method to detect subtle memory decline in temporal lobe epilepsy or other neuropsychiatric populations with memory impairment and may generate precise behavioural phenotyping in research settings.