Theories of Visual Perception

Ian E. Gordon

• Publisher: Psychology Press
• Year: 2004
• ISBN:
  9781135424282
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

Theories of Visual Perception 3rd Edition provides clear critical accounts of several of the major approaches to the challenge of explaining how we see the world. It explains why approaches to theories of visual perception differ so widely and places each theory into its historical and philosophical context. Coverage ranges from early theories by such influential writers as Helmholtz and the Gestalt School, to more recent work in the field of Artificial Intelligence. This fully revised and expanded edition contains new material on the Minimum Principle in perception, neural networks, and cognitive brain imaging.

1 Theory and method

In this chapter we will offer some general remarks on the nature of scientific theories – a reasonable beginning to a book bearing the present title. Then we shall say a little about the role of new techniques in the discovery process and subsequent theorizing.

The calibre of scientific theories

Much has been written about the nature of scientific theories. Perhaps the best-known writer in this field is Karl Popper (1902–1994). Popper argued that no scientific theory can be proved to be correct; it can only be shown to be wrong, or at least flawed. It is the job of scientists to find ways of challenging theories by empirical means. This is only possible if theories are designed to be open to empirical validation – they must be falsifiable. That the moon is made of green cheese is a silly theory (or hypothesis), but it is at least falsifiable: we can go there and check. In contrast, Freud’s psychoanalytic theory is not silly, but many have doubted whether, for example, his concept of the Oedipus complex could ever be falsified by empirical observation.

Popper (1959) demonstrates how the process of empirical testing leads to an evolution of theories whereby earlier theories become corrected and often embedded within newer, more inclusive ones. Thus, Newton’s theory can still be used in many contemporary situations, such as sending a rocket to the moon. But it was the presence of certain weaknesses in the theory (facts it could not explain) that led to Einstein’s theory of relativity – which includes Newton’s laws as special cases.

It is now generally agreed that when Popper made his assertions he was thinking about what can be called ‘great theories’. Examples are Newton’s laws, Darwin’s theory of evolution, relativity theory and, most recently, the quantum electrodynamic theory of light. There is no need to write at length about these theories. However, here are a few examples of their power.

Newton’s formula, G = (m1 × m2)/d², where G = gravitational force, m₁ and m₂ are the masses of two bodies and d² is the square of the distance between them, has been described by the distinguished physicist Richard Feynman as the most powerful equation of all time (Feynman, 1999): it can be used to predict the movements of the planets around the sun, the orbit of the moon around the earth and even the heights of tides – and these are but a few examples of its power.

Darwin’s theory of evolution has been described by the philosopher Daniel Dennet as, ‘The best idea anyone ever had’ (Dennet, 1991). The theory explains the evolution of all species and has survived every test to which it has been subjected. For example, in the 140 years since the theory’s publication no single fossil has ever been found in the ‘wrong’ geological stratum.¹

Einstein’s famous equation, e= mc², where e = energy, m = mass and c = the speed of light, led eventually to the splitting of the atom: a landmark in the progress of knowledge.

The quantum electrodynamic theory is hard for the non-physicist to grasp. However, one fact demonstrates the theory’s power: the theory has been subjected to a series of ever more stringent experimental tests. For example, a number of experiments have found the value of what is known as ‘Dirac’s number’ to be 1.00115965246 with an uncertainty of around 4 for the last digit. The quantum electrodynamic theory predicts the value to be 1.00115965246 with a larger uncertainty for the last digit. If the distance from New York and Los Angeles (about 3000 miles) were to be measured to this degree of accuracy, the above mismatch would amount to the thickness of a human hair. This is a very powerful theory.

What we may call ‘good’ theories fall short of the extraordinarily high standards exhibited in those outlined above. They may explain some important phenomena, but they give rise to predictions that may not always be confirmed by empirical testing. Examples include: Mendelian genetics, Marx’s theory of the historical process, Keynes’s economic theory, Chomsky’s theory of syntax and, closer to the theme of this book, the Young–Helmholtz theory of colour vision.

There are no great theories in this book. It will be claimed in later chapters that Marr’s work and some physiological discoveries have led to ‘good’ theories, in the sense outlined above. However, most of the work that is to be described might better be described as coherent sets of ideas – ideas that have in fact prompted much high-quality research. We may call these ‘utilitarian’ or ‘working’ theories. For example, there are many perceptual researchers who believe that the essence of visual perception is that it is a knowledge-driven process (or sets of processes); in other words, perception is essentially a constructive process. Other workers have assumed that perception is largely the product of innate brain processes. Currently, there are many who claim that perceptual processes cannot be adequately understood until the intimate relationship between perceivers and the environment(s) in which they evolved has been made the focus of major research programmes.

These sets of ideas have not much in the way of formal structure. Few of them are capable of generating quantitative predictions and, as we shall see, even when flaws are found and awkward facts are discovered, this does not lead to the abandonment of the approach or even to major alterations to the theories. What such theories have done is, first, to provoke ingenious experiments, second to unite like-minded researchers and raise their motivation. To take but one example, infant perception research draws huge numbers of workers to international conferences. There they can exhibit their latest research findings, argue with others, learn about the latest theoretical trends and place their own researches into a contemporary context. Outside the lecture theatres they also have fun.

Given that there is no general philosophical agreement among vision researchers on what needs to be explained about perception – conscious experience, neurophysiological mechanisms, and so on – it is not surprising that theories of visual perception have so far lacked the rigour and power of the great scientific theories. We should not be depressed by this fact. The brain is the most complex system in the known universe. It may never be fully understood.

The importance of methods and measurement

In editions 1 and 2 of this book, the first substantive chapter was devoted to the concept of the threshold. Students tended to find this chapter dry and over-technical; some academic colleagues wondered whether the theory of the threshold was a true psychological theory. On the second point, we believe that we were right. However, in response to students’ complaints we have removed the chapter.

That said, we have found that students, particularly beginners, and also lay people who have read Theories of Visual Perception, often wish to know more about how certain phenomena in perception were discovered. This section says something about the relation between methods, discoveries, and the theories designed to explain the discoveries. We shall develop this point by describing a few examples.

In the eighteenth and nineteenth centuries, sensation was held to be instantaneous; that is to say, when a surface is touched and the touch is felt, there is no time lag between these two events. The reason for this belief is that for much of the period many thinking people believed that the nervous system was part of the soul. As such, it was part of God’s creation and was therefore perfect and instantaneous. We should not, in retrospect, deride these beliefs; they were part of the culture of citizens educated in Christian communities.²

In the middle of the nineteenth century, the great scientist Helmholtz (of whom we shall learn more later) carried out a reaction-time experiment. He touched the foot of one of his observers and instructed him to push a key on feeling the touch. Many trials were carried out and the average time to react was recorded. Then Helmholtz repeated the experiment, this time touching the observer’s thigh. On average, these reaction times were shorter. Helmholtz knew the distance between the foot and thigh sites. Thus, by subtracting the foot times from the thigh times, Helmholtz was able to arrive at an estimate of neural conduction time. This showed that the average conduction time was about 100 metres per second. Sensations are definitely not simultaneous with stimulus events.

This result changed many things. For example, it led to the measurement of reaction times under many different conditions (reaction times to pain, to heat, to cold, and so on). The development of measures of choice reaction times (‘press this switch when a red stimulus appears, press that switch when you see a green stimulus’) opened up all sort of possibilities of measuring human decision times: techniques that are used to this day in exploring subjective perceptual complexity and even the complexities of language. Reaction time research has been wonderfully fruitful. And many of the data gathered in this research have led to new and important theories. For example, reaction times to pricking pain are considerably shorter than reaction times to burning pain. This led eventually to the gate control theory of pain – a very important development.

Helmholtz was able to establish this breakthrough only because he was able to use a new precise timing device, the kymograph, on which were mounted two relay-driven pens. One was triggered by stimulus onset and the other marked the revolving paper drum when the observer responded. Simple reaction times (e.g., reacting to touch) average around 120 ms – less than one-fifth of a second. Until it was possible to measure with this degree of precision, human reaction times would have played no part in experimental psychology (see the Endnotes to this chapter).

Once accurate timers became available, it was possible to calibrate mechanical shutters. This led to the development of the tachistoscope, by means of which it was possible to present visual stimuli for very short periods of time (one-hundredth of a second, say). Immediately, it became possible to measure how much time it took to identify visual stimuli, to count clusters of dots and to read words. Use of modern electronic versions of the tachistoscope has allowed modern researchers to measure how long it takes to place a visual image on the retina (this proves to be nearly instantaneous) and how long it takes to feed information from the image back to the visual cortex (the rate here is about 10 items/s). Using even more sophisticated tachistoscopes, it has been found that an image can be ‘wiped off’ the retina by a second stimulus, provided the interval between the two exposures is sufficiently brief.

At start of the twentieth century anatomists and physiologists became interested in the functions carried out by different regions of the brain. Of necessity, many of these investigations were crude: a common technique, known as ablation, was to remove part of the brain of an experimental animal and then, after the animal recovered, it was tested to see which functions had been lost. Refinements in this technique eventually enabled researchers such as Sherrington and his colleagues (see, e.g., Creed, Denny- Brown, Sherrington, Eccles, & Liddell, 1932) to explore the working of the spinal cord in anaesthetized animals: it was in this manner that the first spinal reflexes were discovered.

A few decades after Sherrington’s researches were published, the microelectrode was developed. A typical electrode is formed by pulling a red-hot hollow tube into a fine capillary. The tube is then filled with a conducting fluid. By carefully lowering the tip of such electrodes onto the visual cortex of living cats (and connecting the electrode to a powerful amplifier – another invaluable invention), Hubel and Wiesel (1962) were able to record the activity of individual cortical cells and eventually to show the functional architecture of this region of the brain. Hubel and Wiesel received the Nobel Prize for this work.

The availability of digital computers can be said to have revolutionized the study of visual perception. For example, prior to 1950 a typical perceptual experiment required the experimenter to present a stimulus to which the observer was required to respond. Then the next presentation was made and the process repeated until all the experimental trials had been completed. This simple process yielded masses of important data. However, as the present author can testify, running such experiments was a lengthy time-consuming process, and it was often necessary to test 10–20 volunteers in order to obtain reliable statistical data.

When digital computers became available, the situation changed dramatically. Simple experiments could be run in the absence of the researcher, the resulting data could be analysed immediately and the entire procedure was much more efficient and took less time. More importantly, the computer enabled a completely different type of experiment to be run. For example, in studies of eye movements, the subject could be wired up to a computer, then the computer recorded the current position of the eye and delivered a visual stimulus. This was merely an increase in efficiency. However, because of the speed of the computer, the visual display could be changed during an eye movement. This technique yielded important insights into what goes on when we make rapid saccadic eye movements – whether the eye can take in information during such movements, and so on.

As will be shown in Chapter 2, the perceptual capacities of newborn infants are of enormous theoretical interest. The challenge has always been how to communicate with these infants. A wonderfully simple technique depends on a phenomenon known as habituation. Show an infant a visual pattern and the infant will look at it. Keep on showing the same pattern and eventually the infant will pay it little or no attention (measured by filming the infant’s direction of gaze). Now change the stimulus. If the infant starts to look at the new stimulus, this is a sign that the change has been detected. Careful manipulation of stimulus patterns in this research has led to some quite remarkable discoveries concerning what infants can see in the first days and weeks after birth.

Each of the technical discoveries outlined above has had a major impact on perceptual theory. Reaction-time data have led to theories of human decision making. Tachistoscopic experiments have taught us much about the nature of visual processing. Single cell recordings from the visual cortex have stimulated theories as to how the visual apparatus organizes inputs to create reliable representations of the external three-dimensional (3D) world. Eye movement studies continue to yield insights into the nature of language processing. There can be no doubt concerning the importance of new techniques in science.

Endnotes

• Some of the problems to be discussed in the remainder of this volume are philosophical ones. Readers who have not undertaken a formal study of philosophy will profit from reading sections in Gregory (1987) and the whole of Dennett (1991).
• In the following chapters there are numerous references to neurophysiological research and theory. Several good undergraduate texts contain excellent accounts of this material. Rosenzweig, Leiman, and Breedlove (1999) is a general account of biological psychology with good sections on the nervous system and perceptual mechanisms. The book is well illustrated and comes with a floppy disc suitable for Macintosh and PC computers.
• Although Helmholtz used human reaction time in an inspired manner to measure neural conduction speeds, he did not discover the reaction-time phenomenon. Many years earlier, astronomers found that different observers recorded different transit times when studying the motion of the planets. When a particular planet was known to be passing through a region of the sky, a telescope fitted with a graticule would be pointed towards the predicted position of the planet at a particular time. As the moving image of the planet was spotted in the telescope, the observer would wait until the planet met the edge of the graticule. As the planet’s image crossed the central line, the astronomer would react by pressing a k...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Preface to This Third Edition
• Acknowledgements
• 1. Theory and Method
• 2. The Gestalt Theory
• 3. Brunswik’s Probabilistic Functionalism
• 4. The Neurophysiological Approach to Visual Perception
• 5. Empiricism: Perception As a Constructive Process
• 6. Direct Perception and Ecological Optics: The Work of J. J. Gibson
• 7. Marr’s Computational Approach to Visual Perception
• 8. Some Final Remarks On Theories of Visual Perception
• References