Characterizing learning in anorexia nervosa: What do we know, and what is left to learn?

A growing body of research suggests that anorexia nervosa (AN) may be characterized by subtle alterations in learning that could relate to symptom maintenance and treatment outcomes. However, findings in this area have been mixed, and there have been few efforts to date to summarize this literature. The aim of the current systematic review was to analyze existing literature that tested different facets of learning in samples of individuals with AN compared to other groups. In line with PRISMA guidelines, we conducted an initial search in April 2024 on PSYCInfo and PubMed and an updated search in July 2025, yielding an initial sample of 1,831 articles, of which 194 were fully assessed for eligibility. Our final review included 130 articles; coding, extraction, and synthesis of results occurred from April 2024 to August 2025. Included studies predominantly were cross-sectional in nature, tested group differences between those actively ill with AN and control subjects who had no history of psychopathology, and used neuropsychological tasks primarily gauging episodic learning (e.g., verbal or visuospatial learning) or associative learning (e.g., reversal learning). Qualitative synthesis of these findings suggests that while there appear to be alterations in learning among individuals with AN compared to controls, the significance and specific pattern of learning differences is inconsistent. It is likely that heterogeneity exists within AN and other eating disorders across development, stage of illness, and clinical phenotypes. Future research that begins to characterize meaningful heterogeneity and adopts methods such as computational modeling may offer promising methods for improving the characterization of and treatments for this devastating condition.