PROJECT SUMMARY Anorexia nervosa (AN) is a devastating disorder with a mortality rate among the highest of any psychiatric illness. Current treatments are often inadequate, and no pharmacologic agents have proven effective. A better understanding of the pathophysiology of AN is greatly needed. Avoidance of high-fat food is a central behavioral disturbance contributing to morbidity and mortality of AN, yet little is known about the underlying cognitive and neural mechanisms. Anxiety and fear of fat are thought to play important roles: individuals with AN are highly anxious in general, with the primary focus of this anxiety being food or eating related. Extreme food avoidance and anxiety surrounding food are well documented in AN, yet a mechanistic framework for understanding these phenomena and their relationship has not been rigorously studied. The proposed research tests a novel mechanism potentially underlying the persistent avoidance of high-calorie, high-fat food—excessive generalization of food avoidance—and leverages recent advances in computational and cognitive neuroscience to elucidate this mechanism. With the use of a novel paradigm, we can operationalize avoidance behavior and quantify learning and generalization of threat and safety signals. Our specific aim is to examine the neural and cognitive mechanisms supporting generalization of avoidance behavior in AN. We will use functional magnetic resonance imaging (fMRI) and computational modeling to characterize, in AN and healthy comparison participants (HC), how active avoidance behavior is generalized both when the feared outcome is electric shock (aversive to AN and HC) and high-fat food (aversive to AN). Our central prediction is that individuals with AN will generalize more broadly when avoiding food than when avoiding shock, and that this will be related to 1) activity in brain circuits involving the insula, striatum, and amygdala and 2) real-world avoidance of high-calorie, high-fat food. The study proposes a novel mechanism to account for the severely maladaptive eating behavior in AN, which has proven very difficult to treat. The application of computational neuroscience tools will allow for more precise specification of mechanisms at both the cognitive and neural levels, which is essential for identifying novel treatment targets, characterizing biomarkers of illness vulnerability and trajectory, and translational research more broadly. This study will create a foundation for a large-scale R01 to investigate these mechanisms developmentally and development and testing of mechanism-based treatments in clinical trials.