ABSTRACT
Damage to areas of the brain concernedwith visual processing can lead to selective disorders of visual perception. In many instances, these disorders affect the processing of some aspects of the visual world but not others. For example, patients can have selective loss of color vision without necessarily having an impairment of form perception or motion vision (Heywood & Cowey, 1999). In contrast, patients can have a gross impairment in motion vision without suffering loss of color perception (Heywood & Zihl, 1999). Likewise there can be marked damage to form perception along with maintenance of the ability to reach and grasp objects using properties of form, or, contrariwise, impaired reaching and grasping from vision along with spared object perception (e.g., see Jeannerod, 1997; Milner & Goodale, 1995, for reviews). This neuropsychological evidence suggests that there is a fractionation of visual processing within the brain, with different regions specialized for coding contrasting properties of the world-form, color, motion, location, and so forth. The argument is further supported by evidence from neurophysiology, where cells in different cortical areas show response preferences for particular visual properties (e.g., color specialization in area V4, motion in area MT; see Cowey, 1994; Desimone & Ungerleider, 1989; Zeki, 1993), and from functional imaging data in humans, where there is selective activation of different cortical sites depending on the visual property being manipulated (e.g., see Tootell et al.,
1995; Ungerleider & Haxby, 1994; Watson et al., 1993). Indeed, even within a single visual domain, such as form perception, there will be multiple cells at early stages of vision that code the image independently and in parallel. However, for coherent representations of whole objects to be derived, activity must be integrated across these cells, so that, for example, local image features become grouped into the shapes that support recognition and action. I refer to this process of integrating image features, both between and within visual dimensions, as visual binding. In this chapter I use neuropsychological evidence on the fractionation of visual processing to argue that binding is not a single operation but instead can be decomposed into multiple stages, each of which may offer a probabilistic interpretation of how image features are interrelated. I link this argument to converging evidence from experimental psychology and neurophysiology, which suggests that forms of binding can be achieved at early stages of vision to be verified by later stages perhaps using reentrant feedback.
