Pii: S0960-9822(02)01593-0

A fundamental organizing principle of the brain seems to be that anatomically discrete regions perform separate tasks. The extent of this specialization is clearest in the visual system, where the cerebral cortex is subdivided into distinct areas, each of which makes a different contribution to the processing of visual images [1]. These areas were originally identified on anatomical grounds (and simply identified with numbers V1, V2, V3...), then physiological investigations indicated that different subdivisions have different properties [2]. The clearest example of such specialisation is provided by V5 (also called MT) in the primate brain, which plays a crucial role in processing moving images (reviewed in [3]), but has little to do with the processing of shape or color. One visual function which is not clearly identified with an anatomically distinct pathway is stereopsis — our ability to combine images from two eyes to perceive depth (Figure 1). If this aspect of visual processing is not localized in the way that motion processing is, then studies comparing stereo and motion may provide valuable insights into what principles dictate the need for anatomical localization of function. Such an endeavor depends critically on the view that there is no anatomical pathway specialized for stereopsis. While some recent physiological evidence adds support to this view [4–6], a recent imaging study in humans [7] indicates that there may after all be a degree of specialization for stereopsis in some brain areas. In order to reconcile these recent observations, it is useful to consider further the distribution of visual operations between pathways. A distinction that has been used for many years divides visual cortical areas into two groups: a ‘dorsal stream’, which is largely concerned with object location and movement, and a ‘ventral stream’, largely concerned with object shape and color [8]. It has been argued that stereopsis is a function of the dorsal stream, but this idea was largely derived from psychophysical experiments [9]. These provide (at best) only indirect evidence about the anatomical location of visual function. One physiological observation used to suggest a role for the dorsal stream in stereopsis was that the cortical areas of the dorsal stream all contained disparity-selective neurons. However, there was not a well-documented lack of disparity-selective neurons in the ventral stream. Rather, several areas within the ventral stream had not been examined for disparity-selectivity. Three recent studies have addressed this imbalance, and found disparity selectivity in cortical areas V4 [4,5] and TE [6], central components of the ventral stream. Another recent study [10] found disparity selectivity in area V3, part of the dorsal stream with connections to ventral stream areas. The emerging picture is that all parts of the visual cortex contain disparity-selective neurons, which suggests there is not an anatomical pathway specialized for the computations supporting stereopsis. Alternatively, this may only mean that simply demonstrating that a brain area contains disparity-selective neurons is not a reliable indicator of a role in stereopsis. A series of experiments from my own group (reviewed in [11]) has demonstrated that, at the earliest stages of cortical processing (area V1), the properties of disparity-selective neurons differ from the perceptual properties of stereopsis in several important ways. It is important to remember that binocular disparities may be used for several different functions — seeing depth, singleness of vision, control of binocular eye movements — and that simply measuring selectivity for disparity does not reveal what contribution (if any) a neuron makes to each of these. More sophisticated neurophysiological tests of binocular function might help identify a cortical locus for stereopsis. Two such approaches have successfully demonstrated a close link between the activity of neurons in area V5/MT and stereopsis. First, electrical Dispatch Current Biology, Vol. 12, R93–R95, February 5, 2002, ©2002 Elsevier Science Ltd. All rights reserved. PII S0960-9822(02)00669-3