If color perception is associated with the non-orientated, color-selective cells in the blob regions, synchrony, or coincidence detection between blobs and other locations, would appear to be required in order to bind together features and different parts of the same object. In other cases, the binding problem might be solved by a hierarchical system and response integration.

Image analysis techniques that minimize mutual information in natural, colored scenes have identified statistically independent image elements, or filters, that resemble the receptive field properties and color tuning of simple neurons in monkey V1. This suggests that the decomposition of spatio-chromatic information into statistically independent luminance, red–green, and yellow–blue orientation selective channels offers an optimum means of coding natural images (Buchsbaum and Tailor, 2001). At the very least, the analysis indicates that simple cells in and outside the blob regions can be viewed as filters that simultaneously reduce spatial and chromatic redundancy, and that the non-oriented blob cells are part of an independent and separate pathway (Caywood et al., 2001).

Higher visual areas

After V1, information processing appears even more complicated. The signal pathways of MC cells project onto the areas V2, V3 and V5 (also called the middle temporal area, MT). Whereas V3 analyzes form and depth, and probably also some aspects of motion perception, V5 (MT) seems specialized for processing movement. Some of these areas are likely to also receive input from PC cells. A simplified schematic route of signals from the retina to cortical areas and the tasks their activity is associated with is shown in Figure 8.7 (van Essen et al., 1991).

A body of accumulated data provides evidence for two parallel pathways from striate cortex (V1) to extra-striate areas, an organization originally suggested by Ungerleider and Mishkin (1982). Figure 8.7 reproduces a proposed scheme for the division of labor between the two visual streams. One system, called ‘the ventral stream’, leads from striate cortex via areas V2 and V4 to the inferior temporal cortex (ITC). For instance, the spectral selective cells in the blobs of V1 project to the thin stripes of V2 and from there to V4. The ventral pathway has been thought to deal with aspects that are important for discerning shapes, color and texture and for object recognition. Its neurons help decide what the stimulus might be. At the highest level of processing in this pathway, we find cells that respond preferentially to faces.

The other system, the dorsal or parietal stream, involves areas V2 and V3, and the movement-sensitive MT, also called V5. This pathway deals with the localization of a visual stimulus in space, i.e. determining where the stimulus might be. Neurons in MT are sensitive to the direction and speed of movement. It has been conjectured that these two streams are continuations of the retino-geniculate-cortical PC and MC

Figure 8.7 Parallel pathways from the LGN to the visual areas V1, V2, V3, V4 and V5 (MT) of the cortex. The symbols are meant to indicate the attributes of the stimuli for which the neurons in these pathways are functionally specialized (prism ¼ wavelength; glasses ¼ binocular disparity; angle ¼ orientation of a contour; index finger ¼ direction of movement; modified from van Essen et al., 1991).

pathways, respectively. Even if there may be dominance of inputs from the PC cells over MC cells in the ventral stream, and a clear dominance of MC in the dorsal stream, interaction and cooperation between channels seem to appear in higher visual processing.

Cells associated with the ability to recognize faces of known persons seem to be located in a relatively small area in the ventral stream. People who suffer from prosopagnosia cannot even recognize their own face in the mirror, let alone the faces of friends or close family members, such as their own spouse or children. However, this does not prevent them from seeing that a face is a face and the attributes of a face, such as the nose, the mouth and eyes, and they are still able to interpret facial expressions. This would seem to demonstrate that the recognition of a familiar face and the analysis of facial expression as an expression of emotion are dealt with by different brain systems.

Place cells have been known to exist in the hippocampus of lower vertebrates (Moser and Paulsen, 2001), but until recently it was unclear whether this place coding

had a homolog in humans. Recent recordings from hippocampus and the parahippocampal region have now provided evidence for a human code for spatial navigation based on cells that respond at specific spatial locations and cells that respond to views and landmarks (Ekstrom et al., 2003). A recently discovered area, human visuomotor area V6A, appears to analyze absolute position inside a room. The receptive fields of these ‘position cells’ do not change in space with gaze shift, thus encoding space independently of retinotopic coordinates.

Highly selective neural activity does not necessarily require conscious awareness. In anesthetized animals, cells in V1 and the next higher areas respond with their usual selectivity to visual stimuli presented within their receptive fields. What does the abundant data on alert and anesthetized animals tell us about the influence of consciousness on responses at different levels of visual processing? The proportion of neurons that are inactivated by anesthesia seems to increase at higher levels in the visual pathway. At the earlier stages of cortical processing only a few cells seem to be influenced by the level of awareness and linked to perception. In contrast to V1 and V2, it seems that a great majority of cells in inferior temporal cortex (ITC) of the ventral stream, and in area MT of the dorsal stream, are sensitive to the state of consciousness. Logothetis (2002) suggested that the small number of neurons whose behavior reflects conscious perception are distributed over the entire visual pathway rather than being localized to a single area in the brain. Other experiments, reviewed in Gazzaniga et al. (2002), have shown that the processing of sensory inputs is influenced not only by the state of awareness, but also by selective attention, imagery and expectations based on previous experience.

Brain injury from a stroke or trauma can occasionally result in cerebral achromatopsia. This is a condition where the person has lost the ability to see colors. This color vision defect can manifest itself without any associated loss of visual field or visual acuity. To such people the world appears to be colored in shades of gray. Sacks (1995) has given a vivid description of such a case in the story of a painter who lost all chromatic vision after a car accident. He came to experience the world as if it were made from lead. Such cases support the idea that color is processed separately from other attributes of the visual image. According to Zeki (1993), this particular functional loss is usually associated with damage localized to the human color center, or human area V4. However, the deficits in patients with achromatopsia differ from those of monkeys with lesions in V4 (Kandel et al., 2000) in that humans cannot discriminate hues but can differentiate shape and texture, the latter being difficult for the monkey. V4 in the monkey is probably not directly comparable with the human color area that is being affected in achromatopsia.

Recently, optical imaging techniques have revealed a spatially organized representation of hues in monkey area V2 much like the orderly sequence of orientation selective cells in V1 (Roe and Ts’o, 1995; Xiao et al., 2003). The responses to different colors peaked at different locations in the thin stripes of V2, and the peak responses were spatially arranged in the order of the hues on a color circle.

Motion detection and direction sensitivity

The perception of form and movement probably arise from two independent neural networks. When something unknown approaches you at a great speed, you inevitably try to avoid it without stopping to determine whether it is a harmless piece of paper or something more dangerous. Physiological data indicate that the analysis of speed takes less time than the analysis of form or function. The neurons that deal with movement and speed are believed to project to other areas of cortex than those that handle form (Zeki, 1993). Damage to the ‘movement channel’ limits the ability to perceive movement, without affecting the perception of stationary objects. Damage to this channel may result in movement blindness.

The movement-sensitive neurons in V5 (MT) are clustered into columns of similar preferred directions. All movement directions are represented across the retinotopic map, and almost every cell in this area is direction-selective. In analogy with the organization of orientation-sensitive cells in V1, neighboring cells would be activated by a slightly different direction of movement, and as a consequence of retinotopic mapping, the same direction of movement for an adjacent position on the retina would be represented by other cells some small distance away. The perception of, for instance, the shape of a moving animal behind occluding trees would seem to depend on spatial interpolation of contours and temporal integration over the occluded body parts moving in synchrony (see Figure 1.4).

The strange phenomenon of movement blindness may be caused by damage to area V5. Zihl et al. (1983) reported a case of movement blindness in a woman who, after a stroke, had lost the ability to see moving objects, although she had no problem seeing them at rest. This caused significant problems in daily life. She would, for example, find it impossible to pour water into a glass. To her the water seemed frozen to ice, and she was unable to follow the water level as she filled the glass. She also had problems crossing streets. A car that was far away would suddenly be close without her having seen it approaching. This and other cases of cortical movement blindness are strong evidence that movement is, indeed, analyzed in a particular brain area.

In primates, a combination of orientation-sensitive cells and cells sensitive to the direction of movement is not encountered before the visual cortex. Neurons in the movement channel are selective to movement in a particular direction, for example in the direction of 2 o’clock, while they are unresponsive to movement in the opposite direction. Figure 8.8 gives a possible model for a possible structure of a neural network that is sensitive to the direction of movement. Owing to lateral inhibition operating in one direction only, a leading edge of bright light that moves from left to right on the retina sequentially activates the ON-cells (I-cells) in the intermediate layer that converge on the direction-sensitive cell at the bottom of the figure. Inhibition opposite to the direction of movement results in a transient increment response to the rightward passing of the dark/bright border and inhibition of the trailing bright/dark edge. A dark/bright border that moves towards the left inhibits the