A Neural Theory of Visual Attention: Bridging Cognition and Neurophysiology Claus Bundesen, Thomas Habekost, and Søren Kyllingsbæk University of Copenhagen A neural theory of visual attention (NTVA) is presented. NTVA is a neural interpretation of C. Bundesen’s (1990) theory of visual attention (TVA). In NTVA, visual processing capacity is distributed across stimuli by dynamic remapping of receptive fields of cortical cells such that more processing resources (cells) are devoted to behaviorally important objects than to less important ones. By use of the same basic equations used in TVA, NTVA accounts for a wide range of known attentional effects in human performance (reaction times and error rates) and a wide range of effects observed in firing rates of single cells in the primate visual system. NTVA provides a mathematical framework to unify the 2 fields of research—formulas bridging cognition and neurophysiology. This article presents a neural theory of visual attention: NTVA. The theory proposes a close link between attentional function at the behavioral and at the cellular level. By use of the same basic equations used in TVA (theory of visual attention; Bundesen, 1990), the theory accounts for both a large portion of the atten- tional effects reported in the psychological literature and a large portion of the attentional effects demonstrated in individual neu- rons. NTVA thus provides a mathematical framework to unify these two fields of research. NTVA is a further development of TVA (Bundesen, 1990). TVA is a formal, computational theory that accounts for a wide range of attentional effects in mind and behavior reported in the psychological literature. At the heart of TVA are two equations, and NTVA is a neural interpretation of these equations. The equations jointly describe two mechanisms of attentional selection: filtering (selection of objects) and pigeonholing (selection of fea- tures). In NTVA, filtering affects the number of cells (cortical neurons) in which an object is represented, whereas pigeonholing is a multiplicative scaling of the level of activation in cells coding for particular features (see Figure 1). The total activation repre- senting a visual categorization of the form “object x has feature i” is directly proportional to both the number of neurons representing the categorization (which is controlled by filtering) and the level of activation of the individual neurons representing the categorization (controlled by pigeonholing), and Equation 1 of TVA essentially expresses this fact. Filtering is done in such a way that the number of cells in which an object is represented increases with the behavioral importance of the object (parallel processing with differential allocation of resources). More specifically, the probability that a cortical neuron represents a particular object within its classical receptive field (RF) equals the attentional weight of the object divided by the sum of the attentional weights across all objects in the RF. Equation 2 of TVA describes how attentional weights are com- puted, and logically this computation must occur before processing resources (cells) can be distributed in accordance with the weights. Accordingly, in NTVA, a normal perceptual cycle consists of two waves: a wave of unselective processing followed by a wave of selective processing. During the first wave, cortical processing resources are distributed at random (unselectively) across the vi- sual field. At the end of the first wave, an attentional weight has been computed for each object in the visual field and stored in a saliency map. The weights are used for reallocation of attention (visual processing capacity) by dynamic remapping of RFs of cortical neurons such that the number of neurons allocated to an object increases with the attentional weight of the object. Hence, during the second wave, cortical processing is selective in the sense that the amount of processing resources (number of neurons) allocated to an object depends on the attentional weight of the object. Because more processing resources are devoted to behav- iorally important objects than to less important ones, the important objects are more likely to become encoded into visual short-term memory (VSTM). The VSTM system is conceived as a (K- winners-take-all) feedback mechanism that sustains activity in the neurons that have won the attentional competition. This article contains three main sections. In the first main section, we review the equations describing the basic mechanisms of selection in TVA and summarize how TVA has been applied to a broad range of findings on human performance in visual recog- nition and attention tasks. The neural interpretation of TVA, NTVA, is presented in the second main section. We first develop the neural interpretation of Equation 1 of TVA in general terms. We then present a set of simple networks for performing the attentional operations of NTVA. We show how the computations of the networks correspond to the original equations of TVA, and Claus Bundesen, Thomas Habekost, and Søren Kyllingsbæk, Center for Visual Cognition, Department of Psychology, University of Copenhagen, Copenhagen, Denmark. This research was supported by grants from the Danish Research Coun- cil for the Humanities and the Carlsberg Foundation. We thank Kyle Cave, John Duncan, Glyn Humphreys, Axel Larsen, Gordon Logan, Tom Palm- eri, Antonino Raffone, Jeff Schall, and Werner Schneider for their com- ments on earlier versions of this article. Correspondence concerning this article should be addressed to Claus Bundesen, Center for Visual Cognition, Department of Psychology, Uni- versity of Copenhagen, Njalsgade 90, DK-2300 Copenhagen S, Denmark. E-mail: bundesen@psy.ku.dk Psychological Review Copyright 2005 by the American Psychological Association 2005, Vol. 112, No. 2, 291–328 0033-295X/05/$12.00 DOI: 10.1037/0033-295X.112.2.291 291