"Attention is a control system that optimizes behavior by altering the way that incoming sensory information is processed. Determining how attention influences the way that information is represented throughout the brain is a core problem in sensory neuroscience. Recent evidence from primates suggests that attention can change tuning in single neurons at higher stages of visual processing. Tuning changes imply that many visual representations are not fixed, but rather can be changed dynamically according to task demands. However, it is currently unknown whether attention causes tuning changes in the human brain.This proposal investigates how attention changes tuning for a broad range of semantic features across the entire human brain. To address this issue, the proposal uses functional MRI to record blood-oxygen level-dependent (BOLD) signals while humans search for specific object or action categories (e.g., “Humans” or “Vehicles”) in natural movies. The proposal leverages an innovative voxel-wise modeling framework to characterize how attention changes tuning for hundreds of semantic features. This framework will enable separate characterization of attentional influences on response baseline, response gain and tuning.The proposal consists of two broad Aims. Aim 1 focuses on how attention changes the way that objects and actions are represented in the brain. Aim 2 addresses the computational principles that govern attentional tuning changes. These experiments will provide crucial information for understanding how the brain dynamically changes representations to optimize behavior during natural vision, and they will provide the first quantitative semantic models that accurately predict BOLD responses during natural visual search. The results will advance our understanding of cognitive processes such as learning and memory that depend on attention, and they will have implications for developing treatments and interventions for attention-related mental disorders."