Decisions often depend on representations about the probability and value of potential outcomes. However, maintaining accurate representations of these variables can be difficult in a dynamic environment. Most strategies for maintaining accurate representations in such an environment update them after experiencing unpredicted outcomes. A key challenge for these approaches is to decide how much influence that unpredicted outcomes should have on existing representations. In principle, this decision should take into account at least two forms of environmental variability. Persistent environmental stochasticity, or noise, leads each outcome to be a bad predictor of the next suggesting that each new outcome should have only a minimal influence on an existing representation. Another form of variability occurs due to sudden environmental changes, or change-points. Such change-points can render historical outcomes irrelevant to future ones, suggesting that representations should be highly influenced by a new outcome. Both forms of variability lead to deviations from expected outcomes, however the two types of variability suggest opposite courses of action. Previous work has shown that people and animals are capable of updating representations nearly optimally in noisy and changing environments, suggesting that the brain has a mechanism for using environmental variability to assign influence to new outcomes. However, little is known about the underlying neural mechanisms. One prominent hypothesis implicates the brainstem nucleus locus coeruleus (LC) in providing an uncertainty signal that can be used to adaptively adjust the influence of incoming sensory information on perceptual processing. However, this theory ¿ and its relationship to more general forms of belief updating ¿ has yet to be tested empirically. The goal of this proposal is to provide me with training on state-of-the-art experimental techniques that combine quantitative behavioral and neurophysiological measurements. This training will allow me to test the hypothesis that LC encodes key computational variables related to perceived noise and change-points that are used to assign influence to incoming information. The proposed experiments are based on behavioral and computational approaches that I developed previously in my graduate work. The first specific aim is to characterize the relationship between pupil diameter and LC activity. The second aim will test whether LC activity reflects behavioral and computational metrics of outcome influence in the same subjects while they perform a representation updating task. Together these Aims will provide new insights about the role of LC in complex, adaptive behavior.