Maladaptive processes of uncertainty are linked to cognitive dysfunctions in mental illness, such as anxiety, behavioral addictions, and schizophrenia. However, it is not clear what specific computations support this general notion of uncertainty, how they go wrong, and how these same computational constructs transdiagnostically influence many mental illnesses. We plan to address these questions by drawing on our recent theoretical work (Piray and Daw, 2021), which identifies specific computational hypotheses about how uncertainty processes may go wrong when organisms are faced with observations that are corrupted by two types of noise: moment-to-moment stochasticity of observations and volatility, i.e., how quickly they change. Using a novel task, we will address these questions both cross-sectionally (Aim 1) and longitudinally (Aim 2) in a large general population. Statistical principles indicate that volatility and stochasticity have opposite effects on the learning rate, a parameter that determines the reliance on each new outcome during learning. But earlier research in computational neuroscience and computational psychiatry failed to consider mutual dependencies in computing volatility and stochasticity, leaving open the question of how the brain separates these two types of noise in the real world in which they are both unknown. In recent work, we addressed this issue and introduced a model for the joint estimation of both factors. A key prediction of the model is that individuals who are less sensitive to stochasticity are more likely to mistake stochasticity for volatility, and vice versa. This situation might, in principle, arise in psychiatric conditions. Here, we propose a behavioral task that systematically manipulates both volatility and stochasticity. The task, together with the model, allows us to find two key parameters for each subject: sensitivity to stochasticity and sensitivity to volatility. In Aim 1, we will use the task to characterize the cognitive process supporting computations of volatility and stochasticity and link the process parameters to transdiagnostic constructs of psychopathology. In Aim 2, we will determine whether model parameters predict trajectories of clinically relevant symptoms. This project attempts to model complicated cognitive processes that are relevant for understanding learning and decision-making dysfunctions in mental illness by utilizing cutting-edge data collection technologies and by mapping subject-level parameters reflecting individual variations in this process to transdiagnostic self-report measures. This work makes it possible to study the neurophysiological underpinnings of volatility, stochasticity, and uncertainty in the future. This project also has the potential to pave the way for the development of biomarkers for transdiagnostic constructs related to uncertainty computations, which are relevant to anxiety, depression, behavioral addictions, and schizophrenia.