perception

perception Our senses probe the external world, and they also tell us about ourselves as they monitor the positions of our limbs and the balance of our bodies. Through pain they signal injury and illness.

How we experience and know about external objects is a question that was discussed by the Greek philosophers and has been ever since. Planned experiments on perception, in the spirit of the physical sciences, were hardly attempted before the mid nineteenth century. They have revealed a surprising complexity of physiological and cognitive (knowledge-based) processes, of which we are normally unaware, though many can be demonstrated simply and dramatically, especially through the phenomena of illusions.

There is a long-standing tradition in philosophy that perception, especially touch and vision, gives undeniably true knowledge. For philosophers have generally sought certainty and often claimed it, whereas scientists (who are used to their theories being upset by new data) are more ready to settle for today's best bet. Many scientific instruments have been developed because of the unreliability or inadequacy of perception.

Perceptions are separate, and in several ways different, from conceptual understanding, for perception must work very fast (whereas we may take minutes or hours to ‘make up our minds’, and years to form new concepts). Also, it would be impossible for perception to draw upon all of our knowledge; and perceptions are of individual objects and events in present time, while concepts are abstract and generally timeless.

The evolution of mechanisms for the perception of objects and events at a distance (most completely through vision and hearing) freed organisms from the tyranny of reflex responses to immediate situations, and no doubt was a necessary precursor of all intelligence. It is a fairly new notion that perception itself is an intelligent activity, requiring still only partly understood problem-solving to infer the objective world from sensory signals. Earlier accounts, especially British Empiricism, portrayed sensory perception very differently, as a passive, undistorting window, through which the mind accepts sensations directly from objects. This is not consistent with physiological knowledge of the senses and the brain, nor with many phenomena, such as illusions of vision and hearing and touch. The notion of ‘direct perception’ is, however, still maintained by some followers of the American psychologist J. J. Gibson, perhaps by taking this aspect of his important writings too literally. Perception is not traditionally thought of as an intelligent activity; though the power, especially of vision, to probe distance gains the time needed for intelligent behaviour and for the intelligence of perception itself.

Are perceptions simply picked up by the senses passively, or are they created actively by the brain, or mind? This issue between passive or active perception is a long-standing debate, with significant implications, such as: what is ‘objective’ and what ‘subjective’? The philosopher John Locke (1690) suggested that there are primary characteristics, such as hardness and mass and extension of objects, in space and time — in the world before life, existing apart from mind — and secondary characteristics, created in minds or brains. Thus colours are not in the world, but are created within us, though related in complex ways to light and to the surfaces of objects. Sir Isaac Newton (Opticks, 1704) expressed clearly that red light is not itself red, but is: ‘red-making’:
…there is nothing else than a certain power and disposition to stir up the sensation of this or that colour. For as sound in a bell or musical string … is nothing but a trembling motion.Then (in Query 23 of Opticks) Newton speculates on something like a neural mechanism of vision:
Is not vision perform'd chiefly by the Vibrations of this (Eatherial) Medium, excited in the bottom of the eye of Rays of Light, and propagated through the solid, pellucid and uniform Capillamenta of the optic Nerves in the place (the ‘Sensorium’) of Sensation?The Empiricist school (of which, in their different ways, Locke and Newton were founders) also rejected the notion that minds can receive knowledge by direct intuition quite apart from sensory experience. Mind was now regarded as essentially isolated from the physical world — linked only by tenuous threads of nerve and by fallible inferences of what might be ‘out there’. Some people find this too unsettling to be true. But it is now generally accepted that perception depends on active, physiologically based, intelligent processes. This is not intuitively obvious, since perception seems so simple and easy and we know nothing of the processes in our brains by introspection. Seeing happens so fast and so effortlessly that it is hard to conceive the complexity of the processes that we now know are needed to interpret the nature of the visual world from sensory signals — processes that remain largely beyond the capabilities of the most advanced computers.

Paradoxically, this takes us to concepts familiar to engineers and useful for physiology. We may describe the organs of the senses as ‘transducers’, which accept patterns of energy from the external world, signalling them as coded messages to be read by the brain, which uses these patterns to infer the state-of-play of the surrounding world, and something of the body's own states. Another useful engineering concept is that of ‘channels’. The various senses feed specialized ‘brain modules’ through neural channels, discovered by physiological and ‘psychophysical’ (perceptual) experiments. Thus, as Thomas Young suggested in 1801, colour vision is created from information about the wavelength of light transmitted through three channels, red, green, and blue, responding to light of long, medium, and short wavelength, respectively. All the hundreds of colours we can see are interpretations by the brain of the relative activity of these three colour channels. The three colour channels correspond, initially, to three kinds of light-catching photopigment in the photoreceptors, called cones, in the retina.

There are similar neural channels representing the orientation of lines and edges, and for movement, as first shown by direct physiological recording from nerve cells in the visual cortex of cats by the physiologists D. H. Hubel and T. N. Wiesel in 1962. There are channels for many other visual characteristics: stereoscopic (3-D) depth, texture, spatial size, etc. The ear has many different frequency channels, and there are scores of channels for the sense of ‘touch’, including those for various kinds of pain, for tickle, and for monitoring the positions of the limbs and the stretch of muscles in order to control movement. We are unaware of activity in these sensory channels themselves. Somehow outputs from the many channels are combined to give consistent perceptions. Small discrepancies — such as the delay in sound between seeing a ball hit a bat and hearing the impact — are rejected or pulled into place to maintain a consistent world. Equally, whole objects are somehow assembled from the many signals in different sensory channels that define them. But how this (‘the binding problem’) is done is not understood.

The theory that perception is ‘cognitive’, depending on inferences from essentially inadequate sensory signals, was first clearly proposed by the German polymath physicist, physiologist, and psychologist, Hermann von Helmholtz (1821–94). He called perceptions ‘unconscious inferences’. We might say that they (our most intimate experiences and knowledge) are simply hypotheses, essentially like the predictive hypotheses of science — though not always agreeing in particular accounts.

More recently, attempts to program computers to see (an important component of artificial intelligence) has shown how hard it is to infer objects from sensed data. The most influential attempt, by physiologist David Marr, suggested that object shapes are derived from the retinal images via three essential stages: (i) the ‘primal sketch(es)’, describing intensity changes and locations or critical features and local geometric relations;(ii) the ‘2¹/₂-D sketch’, giving a preliminary analysis of depth, surface discontinuities, and so on, in a frame that is centred on the viewer;(iii) the ‘3-D model representation’, in an object-centred co-ordinate system, so that we see objects much as they really are in 3-D space, though they are presented from just one viewpoint. Marr supposed that this last stage is aided by restraints on the range of likely solutions to the problem of what is ‘out there’. These information-processing constraints are set by assuming typical object shapes; for example, that the shapes of many objects, such as other human beings, are modified cylinders. Interestingly, the painter Paul Cézanne came close to this notion in 1904:
Treat nature by the cylinder, the sphere, the cone, everything in proper perspective so that each side of an object or a plane is directed towards a central point … nature for us men is more depth than surface. David Marr stressed the importance of immediate, passive processing of sensory signals, over active, cognitive ‘top down’ application of knowledge gained from the past. This is a central controversy, currently moving towards greater cognitive ‘top down’ contributions, especially for vision. When computers (or the form of computing known as ‘neural nets’) can access vast amounts of knowledge appropriately, in real time, they might share our miracle of perception.

Artists and scientists can teach each other secrets of perception (as by Gombrich, 1960), though such cross-cultural communication is not easy for most of us.

Richard L. Gregory

Bibliography

Gibson, J. J. (1950). Perception of the visual world. Houghton Mifflin, Boston MA.
Gibson, J. J. (1966). The senses considered as perceptual systems. Houghton Mifflin, Boston MA.
Gombrich, E. (1960). Art and illusion. Phaidon, London.
Gregory, R. L. (1966, fifth edn 1998). Eye and brain. Oxford University Press.
Hubel, D. and and Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. Journal of Physiology, 160, 106–54.
Marr, D. (1982). Vision. W. H. Freeman, San Francisco.
Zeki, S. (1993). A vision of the brain. Blackwell, Oxford.

See also illusions; sensation; senses, extensions of; sensory integration; vision.