A Single Dimension of Memory Strength Underlies Semantic and Phonological False Memories

Abstract

False memory can arise from semantically or perceptually similar experiences, as demonstrated in the Deese-Roediger-McDermott (DRM) paradigm. Although semantic and phonological DRM tasks yield parallel false-recognition patterns, researchers have argued that they reflect distinct memory processes, based in part on findings that semantic associates elicit more false “remember” (R) judgments, whereas phonological associates elicit more false “know” (K) judgments. We directly tested this claim using semantic and phonological DRM tasks with old-new recognition followed by RK ratings ( N = 227 English-speaking students from an Australian university). The R/K dissociation was replicated, yet both parametric and nonparametric analyses revealed a single underlying dimension. A signal-detection-theory model fitted the data nearly perfectly, where R judgments reflected a stricter response criterion than K. A state-trace analysis likewise showed a monotonic relation between old/new and R false alarms across tasks. These results challenge dual-process interpretations, instead suggesting that semantic materials simply produce higher rates of false recognition.