Human Gaze Behaviors Track Abstract Stimulus Categories

Categorization, or the ability to group stimuli according to behavioral relevance, is a cornerstone of abstract cognition. Neurophysiological studies in nonhuman primates have revealed that category-selective signals are robustly encoded in oculomotor structures, including the lateral intraparietal area and superior colliculus, and that this encoding produces small, uninstructed eye movements that reflect learned category distinctions. Whether a similar phenomenon exists in humans is unknown. Here, we show that human gaze behavior encodes abstract categorical information independent of physical stimulus features. In three experiments, participants learned to classify oriented stimuli according to an arbitrary rule while their eye movements were recorded. Category identity could be reliably decoded from records of gaze position, particularly on trials with accurate categorization. A delayed match-to-category task confirmed that category-selective gaze patterns emerged before any motor response could be planned, ruling out response-related confounds. A further experiment demonstrated that gaze patterns tracked both orientation and color categories, confirming that the decoded signal reflects the observer's behavioral state rather than stimulus-evoked oculomotor biases. These findings establish that human oculomotor behavior carries abstract cognitive signals, paralleling recent nonhuman primate results linking incidental gaze shifts to the multiplexed encoding of category information within oculomotor networks.