It is now believed that the visual system is organized into many more or less independent channels, not only for colour (which uses three channels) but also, and perhaps more surprisingly, for different orientations of lines or edges, directions of movement, brightness, and texture size. Isolating and studying channel characteristics is an important part of current experimental work on how the senses function. Just how the channels converge to a single unified perception (which used to be called the common sense) is not yet understood.
What use are channels? If information transmission was the only requirement for designing sensory channels, then just one, covering all the physical information, might suffice. However, the information also has to be analysed and, if there are many different channels, they can sort the information into relevant categories to start the perceptual processing. Many of the objects in the environment are only recognizable in terms of a combination of different physical attributes (think how you would recognize bacon cooking). Likewise many of our actions are determined by the information from various different physical sources. These different physical attributes and sources have therefore to be distinguished and it is convenient to allow the selectivity of channels to make initial distinctions. Conversely, sometimes a particular object can be partially specified by one source of physical information (think how you can search a crowd for a friend wearing a scarlet hat), and in this case a channel more or less specific to that information is very useful.
If I have a private channel from A to B, the receiver at B knows a great deal about the incoming messages in advance, such as their type and range, because these are simply inherent to the channel. The selectivity of a channel is a type of advance knowledge which allows the receiver to correct any systematic errors in the channel. Imagine a channel transmitting edge curvatures, which are physically as often curved one way as any other. If the receiver simply compares each incoming message with the distribution of messages that it has received in the past it will automatically calibrate the channel for any distortions that it introduces.
All channels have physical limits on the amount of information that they can transmit, but some physical attributes, for example luminance contrast, can take a wide range of values. In such instances, it is useful to be able to split the whole range into a number of separate bands and assign one channel per band. This arrangement allows the transmission of a wide range of values with high accuracy.
Luminance contrast is an interesting case. Imagine a diffuse shadow lying across a finely chequered surface. The luminance contrast between any adjacent black and white squares will always be the same because the shadow will equally affect the two. However, the luminance contrast between a black and a white square that are separated will be affected by the shadow, which may lie over one but not the other. Long-range contrast is different in this case from short-range contrast, even though they physically overlap in space. A channel for contrast must work over some predefined spatial range, and a long-range channel would be indifferent to short-range contrast and vice versa. Therefore it is necessary to have several contrast channels, each selective for a different spatial range.
Many physical aspects of our environment have this complicated overlap property. The wavelength of light is another instance: any physical surface reflects a wide range of different wavelengths. Sound and speech is another interesting case. In a room where several people are speaking at the same time, the air is vibrating in response to all of them together. The perceptual process has a problem, the 'cocktail party problem', because it has to break the air vibrations down in portions that will allow it to select one speaker's message. The hearing system has channels that are selective to frequency of sound and direction and these are thought to provide the basis for this selection.
It is sensible to have many channels for several reasons. They organize the overall sensory information into a structure which is suitable for subsequent cognitive processing. They also allow for automatic calibration and adaptation to prevailing conditions in the physical environment.
Empirically, it has so far proved easier to identify channels than to discover how their outputs are combined or used. Channels can be discovered by seeking interactions between the processing of very similar physical stimuli. Gazing at a periodic pattern of high-contrast black and white stripes will make the visual system less sensitive to very similar patterns, but does not affect sensitivity to patterns where the stripes have a considerably different size or direction (see after-effect, perceptual). Similarly, similar patterns will interfere when present simultaneously if they are processed by the same channel. In these observations the important point is the selectivity of the effect for a particular range of physical stimuli, and this range is thought to correspond to the selectivity of the channel itself.
(Published 1987)
— Richard L. Gregory/Roger J. Watt
- Bibliography
- Andrews, D. P. (1964). 'Error-correcting perceptual mechanisms'. Quarterly Journal of Experimental Psychology, 16.
- Braddick, O., Campbell, F. W., and Atkinson, J. (1978). 'Channels in vision: basic aspects'. In Held, R., Leibowitz, H., and Teuber, H. L. (eds.), Handbook of Sensory Physiology, vol. viii.
