Abstract
Individuals with binge-type eating disorders (binge-EDs) repeatedly engage in binge eating despite negative consequences, suggesting altered decision-making. However, the specific cognitive mechanisms underlying these alterations remain poorly understood. In this longitudinal study, we applied the drift diffusion model (DDM) - a computational approach that isolates core decision-making components - to examine how these components relate to binge-eating frequency over time. Ninety-five adults with binge-EDs (69% binge-eating disorder; 15% bulimia nervosa) completed a probabilistic reward task at baseline and 3-month follow-up, with binge-eating frequency assessed concurrently and at 6-month follow-up. Results indicated that slower evidence accumulation (lower drift rate) consistently predicted greater binge-eating frequency both cross-sectionally (baseline p < .001; 3-month p = .018) and prospectively (6-month p < .001), highlighting impaired integration of decision-relevant information as a possible mechanism maintaining binge eating. A lower decision threshold, indicating less cautious decision-making, was cross-sectionally associated with greater binge-eating frequency (baseline p < .001) but did not predict symptoms over time (p-values >.552). In contrast, reward sensitivity (start bias) showed no significant relationship with binge-eating frequency (p-values >.357), possibly reflecting methodological limitations. These findings tentatively support the hypothesis that specific deficits in core decision-making processes contribute to binge-eating persistence, suggesting novel intervention targets. Additionally, our study demonstrates the utility of the DDM as a computational framework for unifying and interpreting diverse behavioral data within the binge-ED literature.