SUMMARYSelective spatial attention, the ability to dynamically prioritize the most important spatial location, is essential for adaptive behavior. It has been studied primarily in head-fixed animals, and almost exclusively in primates. Here, we report the development of two human-inspired, discrimination-based behavioral paradigms for studying selective visuospatial attention in the freely behaving mouse: the spatial probability task, and the flanker task. In the spatial probability task, we found enhanced response accuracy, perceptual discriminability, and rates of sensory evidence accumulation at the location with higher probability of target occurrence, and opposite effects at the lower probability location. In the absence of systematic differences in sensory input, motor biases, and trial structure, these results demonstrated endogenous expectation-driven shifts of spatial attention. In the flanker task, we found that a second, ‘flanker’ stimulus presented with the target, but with incongruent information, caused switch-like decrements in response accuracy and perceptual discriminability as a function of flanker contrast, as well as a reduced rate of evidence accumulation. These results demonstrated exogenous capture of spatial attention. The innovation of behavioral tasks for selective visuospatial attention in unrestrained mice opens up a rich avenue for future research dissecting the neural circuit mechanisms underlying this critical executive function.