PROJECT SUMMARY/ABSTRACT Evidence indicates that disruptions in loss and reward valuation exist across traditional psychiatric diagnostic categories, and these elements are featured in the NIMH Research Domain Criteria matrix. However, validating these features of the RDoC matrix and determining the translational utility of loss and reward valuation requires at least three critical advances: i) understanding the elements’ relational structure (i.e., to what extent are loss and reward valuation linked or distinct), ii) establishing the functional relevance of valuation measures (i.e., which features of loss and reward valuation are related to which symptoms), and iii) determining the stability or lack thereof of the elements and relationships between the elements (i.e., determining which valuation features are state-like vs trait-like). To work toward validating valuation elements and their relevance to psychopathology, we respond to RFA-MH-19-242 (Computational Approaches for Validating Dimensional Constructs of Relevance to Psychopathology). Specifically, we take a data-driven, computational psychiatry approach merging clinical and experimental data to delineate relationships among computationally derived components of loss and reward valuation and with symptoms in a large sample of participants with clinically significant mood, anxiety, or anhedonia (Aim 1). In Aims 2 and 3, we incorporate a mechanistic trial to assess whether components of and relationships between loss and reward valuation are sensitive to change a) over time, b) following 12 sessions of instructed valuation (Aim 2), or c) following cognitive behavioral therapy (Aim 3). If successful, we believe there is immense opportunity to bridge behaviorally-oriented clinicians and computational (neuro)scientists and advance the field by mapping symptoms to neuromechanistic disease processes and spurring the development of new neurobehaviorally- guided treatment approaches. As required by the RFA, this application assesses multiple constructs (loss and reward valuation constructs and learning subconstructs) in the Negative and Positive Valence RDoC domains, using multiple tasks and levels of data.