Testing a Dynamic Spatiotemporal Normalization Framework for Human Vision

PROJECT SUMMARY/ABSTRACT Our perception at a given point in time does not only depend on instantaneous visual input, but also on what happened before and after. Yet most research on visual perception has focused on static image processing, and we lack unifying theoretical frameworks for understanding how vision unfolds in time. One promising theoretical building block is normalization. Normalization is a canonical computation in sensory systems that plays a key role in highlighting salient changes in visual scenes. Whereas the normalization framework has been vetted extensively in the spatial domain, extension of the normalization framework to the time domain has been limited. In addition, although spatial normalization is thought to be regulated by spatial attention, how temporal attention (attention to moments in time) interacts with temporal normalization has not been investigated. In this proposal, we introduce and test a flexible new dynamic model of spatiotemporal normalization, which has the potential to unify disparate effects of contextual and attentional modulation across space and time under a single computational framework for dynamic vision. The proposed behavioral and fMRI experiments are designed to test several predictions of the computational theory. These include effects of preceding and following temporal context on perception and neural responses, the dependence of temporal context effects on the time constants of temporal receptive fields, and the tight relationship between temporal attentional modulation and temporal normalization. The high-quality linked behavioral and fMRI datasets we collect will not only allow us to fill in gaps in our knowledge about basic aspects of temporal processing, but will be suitable for future researchers seeking to test dynamic vision models. We expect that the basic knowledge gained from these studies will provide a necessary foundation for the development of diagnostic tools and treatments for clinical disorders that involve deficits in central visual processing, especially in the context of dynamic, active vision.