A recent experimental study examined whether the brain continues to respond to food cues as rewarding even after an individual is full. While hunger is biologically regulated by hormones such as ghrelin, which stimulates appetite, and leptin, which suppresses it, overeating cannot be fully explained by these homeostatic mechanisms. Food-related cues such as images or advertisements can capture attention, elicit cravings, and even prompt eating, even in individuals who are already satiated.
Learning Food cues are also conditioned stimuli, which develop after repeated association with food. In the long run, they can trigger the habitual and model-free reward mechanisms that do not depend on the prevailing hunger. The experiment tested whether neural reward signals persist after outcome devaluation, specifically, after participants consumed one of the foods to the point of satiety. In case the neuronal responses were not decreased after the devaluation, it would indicate that the initial reward cues are not affected by the current motivational value.
There were 90 participants from the University of Plymouth. The exclusion criteria were dieting, eating disorders, neurological disorders, taking medications, age above 29 years, and BMI out of 18.5-45. Neural analyses on 76 participants were made after 14 subjects were rejected because of excessive EEG artifacts.
Participants first evaluated eleven different foods based on how desirable they found them. From these, two items with closely matched ratings falling between 5 and 8 were chosen for each individual, with one sweet option and one savory option included. One of the selected foods was then randomly designated for devaluation and was eaten to satiety halfway through the study.
Huvermann and colleagues likewise used food images as feedback within a probabilistic paradigm; however, their design involved a guessing task rather than a reinforcement learning task, and the outcomes varied along a graded scale instead of being simply binary. Subjects were made to complete a probabilistic two-step reinforcement learning experiment. In every trial, they selected fractal images with a probability of resulting in an intermediate stimulus and a food or no-food outcome image. Every food result received one point for future access to that food.
Before the meal, the participants had to complete around 180 trials. After consuming the devalued food until they no longer wished to eat it, participants completed approximately 60 additional trials. In the post-meal phase, food and monetary rewards were inversely linked: the fractal associated with a 0.68 probability of winning food carried a 0.32 probability of winning money, and vice versa.
EEG was taken from 61 electrodes. The principal neural effect was a reward positivity (RewP), which can be defined as the voltage difference between food and no-food at FCz in 240-340 ms. Task accuracy was used as the measure of behavioral devaluation, and desirability ratings were used as the measure of subjective devaluation. Mixed learning strategies were demonstrated by the use of computational modeling (Omega mean 0.52, SD 0.31).
Debasement success was obtained on the behavioral front. Accuracy was not significantly different (drift: 0.53 vs. 0.55; fixed: 0.65 vs. 0.65) in drift and fixed stages, both in valued and devalued foods. The devalued food (0.47) and the valued food (0.56) performed worse post-meal than the valued food, and the interaction was significant (F2,146 = 6.36, p < .001, η²p = .08).
The same was the case with subjective ratings. There was a similarity in the pre-meal ratings (valued 6.69; devalued 6.82). The food rating was also devalued to 4.21 after food and was higher at 6.00 during the post-meal (F1,75 = 50.20, p <.001, 2p =.401).
Neurally, results diverged. The RewP demonstrated a high pre-meal food-no-food effect (t(75) = 5.64, p <.001). There was, however, no interaction of food value category and meal stage (F1,75 = 0.003, p =.954). There was no support for neural devaluation sensitivity in the Bayesian tests across the EEG, and moderate to strong support for devaluation insensitivity.
Neural measures were not correlated with behavioral performance, subjective rating, and omega. Even as participants consciously minimized their desire for, and behavioral approach toward, the sated food, there was no reduction in early neural reward responses. The absence of devaluation effects in the RewP or in the EEG suggests that early neural reactions to food cues may reflect habit-like, model-free value signals that persist regardless of the individual’s current state of hunger.
This dissociation suggests that resistance to food cues may be contingent on subsequent, goal-oriented control mechanisms and not immediate neural valuation. Even when people are full, their brains may temporarily continue to encode food cues as rewarding, offering insight into why overeating can occur despite satiety.
Reference: Sambrook TD, Wills AJ, Hardwick B et al., Devaluation insensitivity of event related potentials associated with food cues. Appetite. 2026;218:108390. doi:10.1016/j.appet.2025.108390
