Exposure therapy is the most effective evidence-based treatment for a variety of anxiety disorders, but relapse is common. Failures of exposure therapy can be explained in part by laboratory studies using aversive conditioning and extinction, where the extinction phase can be viewed as a model of exposure therapy. These studies show that the “safe” association learned during extinction is tenuous, and the original negative association can sometimes return – a process known as spontaneous recovery. Recent work from our group shows that during extinction, some participants update the original association from fear to safety while others remember the fear association. The latter participants show spontaneous recovery (relapse) of fear at a later test. The overall goal of Project 3 is to understand these individual differences in spontaneous recovery, their relationship to transdiagnostic mental health symptoms of anxiety and to generalized anxiety disorder, and their neural underpinnings. To accomplish this goal, we will leverage the computational framework of latent cause inference; according to this framework, spontaneous recovery arises when an old latent cause associated with an aversive outcome is inferred to be active in a harmless situation, a form of overgeneralization. We will fit a model (from Core C) to behavior of individual participants to precisely quantify individual differences in the computational process underlying spontaneous recovery and correlate these to symptoms. We will also identify the neurocognitive systems that are implicated in this overgeneralization: we hypothesize that participants show spontaneous recovery because of deficient memory-control processes – these participants are failing to suppress retrieval of the original memory, even in the presence of evidence that the situation is now “safe”. We will relate this to clinical symptoms: based on prior work, we hypothesize that spontaneous recovery will be associated with low memory-control ability and high clinical symptoms of anxiety. Using Core B, Aim 3.1 will test for relationships between memory control, spontaneous recovery, anxiety symptoms and parameters of latent cause inference, taking a dimensional approach in a large-scale online study and also comparing a clinical sample with generalized anxiety disorder to healthy controls. Aim 3.2 will leverage Core D to identify the neural correlates of spontaneous recovery using multivariate predictive modeling of fMRI, focusing on brain networks previously established to relate to memory control, and also using data-driven methods to identify new biomarkers. Last, Aim 3.3 will causally test our hypotheses about brain networks that mediate spontaneous recovery by intervening on these networks with real-time fMRI neurofeedback, with the goal of reducing spontaneous recovery in individuals who are prone to it. Taken together, these studies will substantially advance our understanding of why spontaneous recovery occurs an...