Review of Theories and Data on the Role of Top-Down Pathways in Visual Perception

Abstract

The cortical centers for visual perception are connected by a tangle of bottom-up and top-down pathways. While the role of the bottom-up pathways is well studied, the role of the top-down ones is not yet fully understood. The aim of this review is to describe recent neurodynamic models of visual perception that attribute distinct functions to top-down pathways, and to assess their merits in explaining and integrating neuroscientific and psychological findings on the interaction between perception and cognition. The first part of the review examines the role of top-down pathways in selective attention to relevant stimuli. Selective attention can be directed to a location in space, to an object, or to a feature. All three forms of selection take place in the same artificial neural network, which receives different top-down influences. The second part of the review examines the role of top-down pathways in mediating expectations about upcoming stimuli. This hypothesis is derived from the model of predictive coding, which is a general model of cortical information processing. Theoretical alternatives to predictive coding are also investigated. The results of neuroscientific and neural modeling suggest that top-down pathways convey different types of information depending on the specialization of the neural network in which they are embedded.