The Dual-Consciousness Mechanism of False Memory Guided Decision Making

Abstract: The constructive nature of memory makes it susceptible to retrieval-related distortions that give rise to false memories. Whether—and how—false memories directly influence value-based decisions remains unclear. We integrated the Deese–Roediger–McDermott (DRM) paradigm with a reward-learning task to test whether false reward memories bias decision preferences via conscious or unconscious processing. Experiment 1 combined the DRM paradigm with a reward-learning task to examine the effect of false reward memories on decision preferences. Experiment 2 introduced a divided-attention task to reduce attentional resources; if false reward memories impact decision-making primarily via conscious processing, divided attention should attenuate false memory-based decision preferences. Experiment 3 employed the process-dissociation procedure (PDP) to dissociate the conscious and unconscious components of false memory. This further tested the predictive role of both components in decision preferences, clarifying mechanisms by which false memory shapes value-based choices.
We tested these hypotheses in three experiments. Experiment 1 used a within-subjects Reward Condition (Reward vs. Non-Reward; n = 37). Experiment 2 introduced a divided-attention task, yielding a 2 (Reward: Reward vs. Non-Reward) × 2 (Attention: Full vs. Divided) mixed design (n = 73). Experiment 3 used PDP to dissociate conscious and unconscious contributions of true and false memory in guiding decision-making (n = 50). Across experiments, during reward learning, some DRM-list items were paired with reward and others were not. In the memory phase, participants made old/new judgments and, for ‘old’ responses, indicated whether each picture had been rewarded. In the decision-making phase, participants maximized earnings by choosing, between two pictures, the one more likely to yield reward.
The paradigm reliably elicited false reward memories for critical lures, which consistently biased decision preferences. In Experiment 2, the divided-attention manipulation during the encoding phase significantly reduced participants' true reward memory, and diminished preference for rewarded items. Trial-level generalized linear mixed-effects models (GLMMs) examined whether true and false reward memory could predict decision preferences. Results showed that false reward memory predicted decision preferences under both full- and divided-attention conditions. Simultaneously, decision preferences for critical lures were higher under full attention than under distraction, indicating that attentional distraction reduced participants' decision preferences. Under the PDP assumption that conscious and unconscious processes operate independently, performance in the inclusion and exclusion conditions was used to obtain operational estimates of the two components. By introducing PDP, Experiment 3 showed that false memory contained a larger unconscious component than conscious component. Models incorporating both components accounted for decision preferences for critical lures. Both conscious and unconscious components of false memory positively predicted decision preferences of critical lures, indicating complementary contributions to decision preferences.
These findings are consistent with a spreading-activation account in which reward associations propagate within semantic networks, fostering false memories that subsequently bias value-based decision-making. Together, the results show that false memory effects on decision preferences cannot be explained solely by limited attentional resources and instead reflect complementary conscious and unconscious contributions. This work clarifies how constructive memory processes shape economically relevant decisions and motivates neuroimaging research to delineate the neural pathways linking false reward memory to biased decision preference.