Mental imagery as a dynamic network node in hunger, satiety, and eating behavior: a systematic review

Abstract Purpose Eating behavior emerges from dynamic interactions among homeostatic, hedonic, and cognitive processes. Food-related mental imagery has been implicated in craving, satiation, and eating pathology, yet its role has not been synthesized within a systems-level framework. Methods Following PRISMA 2020 guidance, PubMed, Scopus, and Web of Science were searched from inception to 30 September 2025. Eligible studies examined mental or chemosensory imagery in relation to eating behaviors and/or hunger–satiety regulation in general populations and clinically relevant samples. Two independent reviewers screened records and assessed risk of bias using RoB 2, ROBINS-I, and related tools. Results Twenty-eight studies (experimental, ecological momentary assessment, and neuroimaging designs) were included. Brief or vivid imagery often increased craving, whereas repeated mental simulation reduced food-specific desire/intake via sensory-specific satiety without consistently altering general hunger. Chemosensory imagery showed stronger links with trait craving and cue reactivity than visual vividness. Evidence also suggested bidirectional psychophysiological coupling (ghrelin eliciting imagery; imagery associated with cortisol/ghrelin responses) and context-dependent effects moderated by hunger state, cognitive resources, restraint, and environmental cues. Conclusion Mental imagery may act as an integrative component within eating-related networks, helping to reconcile contradictory findings—including the amplification-satiation paradox—through dose, modality, and contextual boundary conditions. The proposed higher-order network model is hypothesis-generating, but may offer an organizing framework capable of formalizing multi-way interactions among homeostatic, hedonic, and cognitive-affective components, with implications for understanding both normative and disordered eating. Direct empirical validation is needed, particularly through longitudinal and network-sensitive designs in clinically diagnosed samples. Level of evidence Level I, systematic review.