NEURAL DYNAMICS OF SEQUENTIAL LEARNING OF STIMULUS–LOCATION MAPPINGS AND RULES

Humans often perform complex tasks through sequential learning, in which basic stimulus-response mappings are acquired first and later modified by abstract rules. However, the temporal neural dynamics underlying this process remain poorly understood. In this study, we investigated how stimulus-response mappings and rule-based transformations are learned and dynamically represented over time. Participants first learned stimulus–location mappings and subsequently acquired color–based rules that transformed these learned locations. In each trial, participants fixated on a central point, briefly viewed an image, and selected its associated location from five options arranged in a pentagon. During an intermediate delay, a color cue (red, blue, or yellow) indicated the applicable rule: red signaled the standard mapping, blue required a 72-degree clockwise rotation, and yellow required a 144-degree rotation. These cues required flexible adjustment of previously learned spatial associations. Behavioral performance was initially low but improved significantly with practice, reflecting successful formation of stimulus-location mappings. Introduction of new rules led to a temporary decrease in accuracy, followed by rapid recovery, indicating effective integration of rule-based transformations with existing mappings. To identify the neural correlates of this adaptive behavior, we trained a spatial position classifier using support vector machines and applied it to time-resolved EEG data. Temporal generalization analysis revealed that location-specific neural representations emerged approximately 600 ms after image presentation and were flexibly updated around 800 ms after the rule cue. Together, these findings demonstrate that participants formed stable stimulus-location associations and dynamically transformed them in response to rule changes.