Odd Perceptions

Richard L. Gregory

• Publisher: Routledge
• Year: 2017
• ISBN:
  9781315516196
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

Richard Gregory was one of the major scientific thinkers of our time. Originally published in 1986, here he presents essays on the rich subject of perception. How we experience colours, shapes, sounds, touches, tickles, tastes and smells is a mysterious and rich inquiry. Wonderful as these sensations are, though, he argues that perception becomes really interesting when we consider how objects are identified and located in space and time as things we interact with, using our intelligence to understand them. Gregory's essays convey the crucial importance of the major scientists and their achievements in the study of perception; but they also show us how much we can learn from our surroundings, our language, our times, our successes and our failures. Why are we so often fooled, in scientific as well as everyday life?

II
Musing

8
Magical Mechanisms of Mind

It is most odd that scientists deal in knowledge – and exercise immense power through knowledge – yet view with deep suspicion explanations in terms of cognitive concepts. Thus psychology is hardly recognized as a respectable science, except when couched in the terms of physics-based sciences such as physiology. If one speaks, for example, to a molecular biologist – and of course they are very well worth listening to – in terms of perception being knowledge-based active processes, his eyes will glass over. They glaze with the kind of frosty glass that protects ladies’ modesty in showers. And as the lady can look out better than you can look in, just so one is trapped into a losing position where explanation becomes powerless. The objection, it seems, is that cognitive accounts in terms of more or less appropriate deployment of knowledge look spooky, as they are not in terms of causal mechanisms. I know this reaction as well as anyone, as I have been plugging cognitive accounts of perceptual illusions for many years. But I persist, for, like it or not, mechanisms can be controlled by symbols.

To put this a little differently: there is a general acceptance in science that mechanistic accounts art good and cognitive accounts are bad science. Yet, surely, it seems bizarre for scientists to scorn cognitive explanations, when what they do all their lives is bartering knowledge. Also, it has been known from the prehistoric abacus that patterns of pebbles (‘calculi’) arranged in strings (‘neurons’) allow mechanical operations to solve problems – so stones can think! It is indeed likely that ancient standing stone circles, such as Stonehenge, were observatories and computers for converting lunar to solar calenders and for the prediction of eclipses (Thom 1971; Hadingham 1975; Thom and Thom 1978). These were, surely, crucial steps towards autonomous computers. It was a most significant ancient discovery that dead stones, and not only organisms with brains, can handle and be symbols, and so be cognitive. We have hardly assimilated this prehistoric discovery even now. Rather than being ‘stoned out of their minds’, our prehistoric ancestors had minds in their stones: yet this looks too like magic for acceptance in science now.

To take this further we may enquire into the origins of, and what is meant by, ‘cognition’ and ‘mechanism’.

The word ‘cognition’ is descended from one of the most ancient of all words: the shadow-casting ‘gnomon’ of ancient sundials; Sumerian, dating from before 3000 BC. The concept is a special stick or column used for casting a shadow to represent, particularly, the positions and movements of the Sun and Moon. Ever since, ‘gnomon’ has been associated with recording, measuring, calculating and knowing. The many word records of Gnomes in poetry or prose date from the earliest Greek literature of the sixth century BC.

Not altogether unrelated are the much later and all too familiar garden gnomes, which were Scandinavian spirits of the Earth. Several current words, such as ‘gnomic’, ‘conning’, ‘cognition’ and ‘knowledge’, derive from the representing-by-shadows of the Sumerian sundial at the very start of science. Gnomons were used by Eratosthenes (276–194 BC) to measure the circumference of the Earth, from the different lengths of the Sun’s shadows at Aswan and Alexandria. And Aristarchos, a generation before Archimedes, measured the distances of the Sun and Moon – and realized that the Sun is the centre of our local universe. So, whatever the objections to gnomon-based notions of cognitive concepts, they do have a respectably ancient lineage! But beside the acceptable connotations of representing and measuring by shadows there are, in the magical sense, occult associations. Thus for Paracelsus, the sixteenth-century alchemist, Gnomes (and the female Gnomides) were inquisitive irresponsible spirits of our alchemical element – Earth. This is the basis, from ancient times, of the garden gnome: ‘know-man’.

More generally ‘gnomic’ meant rules for wisdom, or heuristic rules for gaining and using knowledge. This is very much how it is used in present-day cognitive psychology, even though it has magical occult origins in spells. This may be the stumbling block for scientific acceptance of explanations in cognitive terms: they look too like spells of magic. This is so especially when physical mechanism accounts are challenged by the claims of symbols affecting matter. Cognitive explanations of brain function suggest that symbols have highly significant powers to move: that patterns of neural activity control us and all we do. For these ever changing patterns in the brain are seen as symbols representing and conveying knowledge and fantasy in a kind of secret brain language. Seen in this way, the patterns of structure and activity of the nervous system may have the suspect magic of the astrologer’s account of patterns of the planets and stars as messages, to be read as picture writing in the sky.

From its beginning, ever since the Babylonians accepted gnomons as pointers to discovery through measurement, science has gradually moved away from reading patterns in nature as talismans of intention. So now symbol power is suspect to science, though symbols have been used for recording and calculating continuously since the wedge-shaped marks on clay of the Babylonian astronomers. It has taken a very long time to distinguish between spells and hieratic sacret-secret writings, such as those of the Egyptians, guarded by the ibis-headed scribe god Thoth, from symbols as freely shared workaday tools. Astrology and astronomy only very gradually separated, over whether the patterns formed by the Sun, Moon and stars are gnome-magic messages, with powers beyond appearance, or whether they are the visible parts of a vast, mainly hidden machine. What science has done is to reveal the machine as force and energy, obeying laws we express in our symbols. So the magic power of symbols has moved from the external world into processes of our understanding. Now that cognitive scientists claim that patterns of structure and activity in our brains are symbols having the power to confer perception and intelligence, to be consistent we have to ask – are we magical?

What are the origins of the concept of physical mechanisms? The word ‘machine’ first referred to stage devices for bringing gods to Earth in Greek plays. So it, too, has early magical associations – especially as the workings of these machines were often hidden, as they were used for conjuring mysteries. Many early machines were used to provide, as occasion needed, evidence of miracles and magical powers. In the long run, though, it was the experience of designing and using machines for everyday uses that gradually separated natural from unnatural. Aristotle thought of a falling object as finding its natural home on Earth, like a home-sick person going back to his own country. He then allowed that matter could be forced into unnatural motions, as in the moving parts of machines. The latter step was to see the entire natural world as a machine. At first, it was seen as a machine designed for our comfort and use, but part of the immense effect of the Darwinian revolution was to replace the notion of the universe as a life-support machine designed for us with the very different notion that we have, over millions of years, adapted to it.

Now we see physics and engineering as different because it is only engineering that has discernible design purpose. So goal-directed matter characterizes engineering, but not physics. In this light, living organisms have the odd status of being goal-seeking machines designed by blind processes, producing the wonderfully effective designs of life without preconceived intention. This dropping of design intention and the rejection of talismans in nature leave us in a moral vacuum: for though we can judge the successes and failures of our own machines by reference to our intentions and their effectiveness, we can no longer judge ourselves in this way.

Although machines serve our purposes, they are generally seen as causally determined by physical principles; so they are not magical. Curiously, when Darwin rejected the special status of the human species, by showing beyond reasonable doubt that we evolved with no special step from animal ancestors, he also forced a widening of the prevailing limited view of mechanical causation. For this took a knock with Darwin’s and Wallace’s concept that systematic change – indeed creative invention – can come about statistically, by selecting from randomly varying individuals. This statistical process, which can be highly creative, lacks the safe, familiar material links of traditional machines. So in this sense at least the organic world is not machinelike – unless of course we extend the meaning of ‘machine’ beyond traditional mechanical causation. Whether we see this as magic creeping back into nature will depend, of course, on just how we define ‘magic’. This is not at all easy to settle.

The blind yet highly creative Darwinian statistical processes, producing changes in species, may seem very different from the mechanisms of behaviour in individuals. But there seems no reason why man-made machines, or living organisms including ourselves, should not play much the same trick – to be creative by selecting from randomly occurring events. These might be either external events or within the machine or the brain. But the behaviour would not, in the traditional sense, be mechanically caused, any more than the statistical processes by which we evolved are traditional physical links between cause and effect.

Another challenge to traditional views of machines came with the control engineering of cybernetics. Feed-back control allows machines to seek goals and correct both internal and external errors. They thus have individual self-guided purposiveness, which is essentially more than the purposes built into precybernetic machines. Cybernetic devices are not, however, guided by selected knowledge from past situations and they hardly learn. This was the step taken, first with analog and then far more flexibly with digital computers. Control by structured knowledge is central to artificial intelligence, and machine learning is becoming important. AI machines are described by the structure and content of their internal knowledge as represented in their programs and stored data, rather than by the characteristics of their mechanisms. Provided their mechanisms are capable of carrying out the program instructions, the former may be almost ignored. If, for example, we are beaten by a chess computer, we attribute its ability to its symbol-handling power to assess positions and make better decisions than ours. But although these machines are controlled by selecting from internally represented possibilities and from knowledge-based rules which we also use, yet we still resist calling them brothers. Why should this be? Is it simply because, although they beat us at chess, they are too ignorant for acceptance in our club?

There is a pair of key though curiously unfamiliar concepts in experimental psychology – positive transfer and negative transfer of knowledge. The essential notion is that what has been learned in the past creates analogies for handling the present and predicting the future. This prediction may set up likely possibilities, allowing planning and decision-making. When the stored knowledge is adequate, and can be applied appropriately to the present situation, then all is well. But when there are significant discrepancies between the knowledge drawn from the past and what is needed, then things go wrong as systematic errors are generated. Negative transfer of knowledge is like having an out-of-date map; we get lost even though we have read it correctly, for it is representing a past reality which no longer exists. Such errors are not attributable to a mechanism: the mechanisms of map-making and reading are irrelevant for the error. Correspondingly, success from positive transfer is not mechanically (or for that matter physiologically) explained. All that matters is whether the necessary knowledge is available and is appropriately applied. This looks like magic, for how without reference to mechanical links of causation, or some statistical effect, can either knowledge (or fiction) affect behaviour – to produce successes or failures?

This looks like dangerous-to-science magic because, on this account, details, however complete, of physiological mechanisms do not tell us what crucially matters. For what matters is whether or not the guiding knowledge or assumptions directing behaviour are appropriate and adequate. This makes knowledge look causal – which is the fearful magic of cognition. But how can science reject cognitive concepts for describing what brains do, while at the same time science itself is dedicated to creating and applying knowledge? Let’s grasp the nettle, and assume that:

1. Brains are machines, but they are symbol-handling machines. They accept data, in neurally coded symbolic forms, to build memory generalizations which are represented internally by cerebral patterns, which are higher-order symbols organized as knowledge. These are internal in the brain, and are difficult or impossible to read from outside. So they are private to each of us.
2. Perceptions are symbolic representations of the world. They are not, however, pictures, but are more like linguistic descriptions in terms of the presence or absence of selected features.
3. Knowledge stored from the past (from genetic inheritance and gained by individual learning) is essential for perception. When it is not adequate, or is applied inappropriately, errors or illusions occur. All perceptions are mixtures of fact, distortion and fiction.
4. The knowledge upon which perception is based only partly overlaps conceptual understanding: so perception and conception may disagree. It is simply false to say that we necessarily, or even generally, believe what we see or see what we believe, for we are often surprised.
5. It can be useful to explain perception and behaviour in knowledge terms, even to the extent of disregarding physiological mechanisms. But this is rejected by much of science – even though science itself is knowledge-based.

We may have to accept that perceptions are, scientifically speaking, odd – because they depend on knowledge. What is spooky is that perceiving organisms are very different from other lumps of matter, because they are not driven so much by present events (stimuli) as by generalizations derived from the past, with an eye to the probable future. This is extremely different from accounts in physics where, for example, falling stones obey laws only of present forces. In fact cognitive-based behaviour is much more like Aristotle’s physics, itself derived from earlier magical accounts, which were like young children’s common sense. Ordinary objects do not predict or have intentions, although, like us, they move in time. A spookiness of cognition is that it transcends the present, by working from the past and what might be, to cope with the future. Thus the Moon obeys Newton’s laws; but an astronaut circling the Moon makes decisions affecting his orbit based on his training and his predictions. This element of prediction and acting on what might be, but is not, makes cognitive processes embarrassingly hard to describe or investigate by normal scientific means. No doubt this is largely why they are ignored and often actively rejected.

The approaching shadow of metaphysics may, however, be warded off here by considering that computers are effective symbol-handling machines. But the power of symbols in computers may look spookily magical, as it is their symbolic structure that counts. The teasing puzzle is that even in t...