A Common Recognition Architecture Across Memory Scales

Abstract

Recognition decisions occur across memory episodes that vary vastly in scale, from compact single-item episodes to extended experiences involving hundreds of items. To test how memory scale shapes the neural mechanisms underlying old/new judgments, we recorded EEG while participants performed a recognition memory task across set sizes ranging from 1 to 128 items (spanning 2 to 256 recognition judgments). Classic electrophysiological signatures of recognition, including frontal and parietal old/new effects, persisted continuously across set sizes. Furthermore, multivariate neural classifiers trained on small memory scales generalized robustly across set sizes. Rather than revealing distinct neural formats, increasing memory scale produced systematic, continuous shifts in the timing and strength of a shared recognition response: signals emerged progressively later and weaker at larger memory scales. Temporal-alignment and cross-correlation analyses demonstrated that responses across scales followed a common waveform trajectory, with timing shifts propagating into motor response preparation. Finally, individual difference analyses revealed that this common recognition format exhibits scale-dependent behavioral relevance, tracking working-memory capacity at smaller scales and long-term memory performance at larger scales. Together, these findings demonstrate that recognition signals share a common neural format across scale while changing dynamically in timing, strength, and behavioral relevance, providing evidence for convergence onto a shared recognition architecture.