Depth of Processing and Congruency Interact to Shape Electrophysiological Signatures of Recognition Memory

Recognition memory is influenced by both the depth of processing during encoding and the congruency of incoming information with existing knowledge. Although both factors have been shown to affect memory performance, their combined influence on the electrophysiological mechanisms underlying recognition remains unclear. The present study investigated how level of processing (LOP) and semantic congruency interact to modulate behavioral performance and event-related potentials (ERPs) associated with recognition memory. Thirty-six healthy young adults completed a recognition task in which encoding required either shallow perceptual judgments or deep semantic judgments, while congruency was manipulated independently. Behavioral performance was assessed using Linear Integrated Speed–Accuracy Scores (LISAS), and ERP components were identified through temporo-spatial principal component analysis. Deep encoding produced significantly better behavioral performance and larger late positive component (LPC) amplitudes than shallow encoding, indicating enhanced recollection-related processing. In contrast, semantic congruency did not significantly influence behavioral performance or LPC activity. However, congruency strongly modulated the FN400 component, with incongruent items eliciting more negative responses than congruent items. Importantly, this effect interacted with depth of processing, such that congruency-related differences were pronounced following shallow encoding but attenuated following deep encoding. These findings suggest a functional dissociation between early and late stages of recognition memory. While recollection-related activity appears to be primarily driven by goaldirected semantic processing, early neural responses are sensitive to the interaction between encoding demands and schema compatibility. The results support accounts proposing that recognition memory emerges from the dynamic interplay between mismatch detection and semantic integration processes.