The real world, as man currently perceives it, is a complex mass of potential stimuli: various sound and light waves speeding by, molecules of different odors drifting in the air, hot and cold sources producing varying interactions of temperatures, and things to be touched, to name just a few. Man is limited by the capabilities of his sense organs and central nervous system to being able to perceive only an extremely small percentage of these stimuli. Man’s visual spectrum covers only a small part of the electromagnetic continuum. Dogs can hear sounds that man misses, and many animals can smell things that man cannot detect. Other things cannot be perceived because they are too small, too far away, too fast, and so forth. Thus man perceives the world through very small windows. He can, however, enlarge his scope somewhat with mechanical devices such as microscopes, infrared photography, slow- motion photography, and amplifiers.
Another source of stimuli is from man’s own body, such as from some internal organs, muscles, and substances in the blood. Again, whole sources of stimuli are outside of man’s conscious sensory capabilities. For example, try to perceive the activity of your spleen.
Given the range of stimuli that man can potentially perceive, what explains why he receives some stimuli and not others at any given time? Part of the answer lies in the properties of the stimuli themselves. Some stimuli simply push their way through the windows. It is fairly probable that a sudden loud noise or a bright light will be perceived in some fashion. This chapter, however, is concerned with the interaction between learning and the sensory-perceptual-attention mechanisms that select, filter, and interpret the stimuli. Figure 3—1 shows a model of this interaction. Some of the potential stimuli are selected and perceived and sent into memory storage. This memory storage then affects the later selection and interpretation of stimuli.
In the 17th and 18th centuries, philosophers debating about how our mind comes to have certain ideas often divided into two camps — the nativists and the empiricists. Nativists, such as Descartes, argued that man is born with the ability to perceive certain basic phenomena. The empiricists, such as Locke and Berkeley, argued that man learns to perceive. Thus the nativists stressed that learning depended on perception, whereas the empiricists stressed that perception depended on learning. Most theorists now reject both of these two extremes and emphasize the continually evolving interactions between perception and learning.
There are many examples of man’s learning to perceive. The first time a person looks through a microscope at a slide of part of the brain, he perceives very little of what he might later learn to see in the same slide. The Russian missile bases on Cuba were discovered by a man who had learned how to identify them from aerial photographs, whereas other observers of the photograph’s did not see them (Gibson, 1969, p. 8). The wine connoisseur has learned to perceive fine discriminations in the taste, smell, and color of various wines.
A key question is whether learning ever affects the way an object is initially perceived or whether it affects only some later stage in the processing, of the information obtained from the object. Does the wine actually taste different to the wine connoisseur, or has he simply learned to do more (make different responses) to the same taste that most of us experience? Perhaps learning doesn’t affect the original perception of the object, but rather affects what type of information about the object can be retrieved from short and long term information storage centers (see Chapter 4). Perhaps these retrieval processes affect what parts of the information go into the person’s consciousness. Alternatives such as these may be greatly expanded depending on the number of stages assumed to exist in the processing of information. The point is that although it is clear that learning interacts with the information-processing system, it is seldom clear at what point in this system learning has its effect. One of the early stages in the system is attention.
Of all the stimuli a person is capable of perceiving, he attends to some stimuli at the expense of others. Considerable research (Norman, 1969, Chap. 2) has been devoted to determining the variables that affect attention. Consider the “cocktail party phenomenon.” A room is filled with people talking to each other in small groups. Thus for any one person there are many voices and sounds impinging on his ears. Yet he is able to filter out of all this noise the conversation of the person he is talking to, and selectively attend to this conversation alone. The person can also easily switch his selective attention so that he tunes in on the conversation of two people nearby or to the words of the music in the background. Such selective attention is a truly remarkable process, which, as Norman (1969, p. 14) points out, has not been duplicated by electronic devices.
Since we can’t attend to all possible stimuli, it is important that we attend to those stimuli that are important to us at the time. But how do we know if a stimulus is important if we don’t first perceive it? Such reasoning usually leads to discussions of different levels of perception. For example, at a basic, probably non-conscious level, stimuli are processed and some decision mechanism or filter determines which stimuli will be further attended to. A different level of perception then might deal only with these selected stimuli. But on what basis does the decision mechanism or filter pick some stimuli over others? Part of the answer is past learning. What is important to the person now usually depends on prior learning experiences. An experienced automobile driver has learned what stimuli should be attended to for efficient and safe driving. Thus we see one place where learning and perception interact: memory storage partially determines what stimuli will be attended to.
A similar argument can be made for the phenomena of perceptual vigilance and perceptual defense. Perceptual vigilance refers to the hypothesis that events of particular importance to the individual are easier for him to perceive, while perceptual defense refers to the hypothesis that it may be possible for an individual to not perceive some events which are psychologically unpleasant. Although there is controversy about these phenomena, they may, to some extent, involve processes similar to those described above for selective attention. That is, some decision mechanism, influenced by memory storage, must first classify stimuli as important or dangerous before perceptual vigilance or perceptual defense can occur. Perhaps at one (non-conscious) perceptual level a stimulus is perceived, interpreted, and classified as dangerous. On the basis of this classification the stimulus is not permitted into the perceptual level that involves consciousness, so that the person has no subjective experience of ever having perceived the stimulus at all.
‘Phenomena such as hysterical blindness (functional blindness due to psychological rather than physiological factors) may often be extreme cases of perceptual defense. Consider the man who has a series of tragic experiences such as the death of his wife and loss of his job. Many of the stimuli that he encounters, such as things that remind him of his wife, will elicit excessive anxiety in this man. If the anxiety is strong enough, the decision mechanism may, via perceptual defense, stop certain stimuli from entering consciousness. In the extreme, with many stimuli eliciting an unbearable degree of anxiety, the decision mechanism may simply shut almost all visual stimuli from consciousness. This, then, would be a case of hysterical blindness, for although there is nothing anatomically wrong with the man’s visual system, he is, to a large extent, functionally blind. Such an explanation is, of course, oversimplified, but perceptual defense may be a significant factor in cases of hysterical blindness.
The next question is how learning affects the perception of stimuli that are attended to.
From the stimuli attended to, man constructs some idea of his environment. Although some of the processes involved in this construction may be innate, many appear to be learned, and this learning how to perceive is called perceptual learning.
There are basically two theoretical approaches to perceptual learning. The first emphasizes that the environment supplies most of the needed information for perception; perceptual learning is described in terms of learning how to effectively use information from the environment. (This is the approach of E. J. Gibson.) The second approach sees the environment as supplying inadequate information, and therefore views perceptual learning in terms of learning to make extrapolations from this limited information. This is the approach of the transactionalists and theorists such as Bruner. These two approaches are, of course, different points along a continuum of theories on how much information is supplied by the environment. Research will have to tell us which points on this continuum offer the best explanation of perceptual processes in different situations.
A few examples of the two approaches will be given below. (For a more complete coverage see Gibson, 1969.) Keep in mind that the specific experiments and arguments given with each theory are not unique to that theory but could be interpreted differently to fit a different theory.
Gibson’s Theory. Gibson’s (1969) theoretical position, a perceptual differentiation theory, is that perceptual learning is learning to extract information out of the sensory data of the environment. The environment is seen as supplying an abundance of information. To make sense of the sensory input a person must learn how to respond to distinctive features of the stimuli. For example, when a young child first hears people speaking to him it probably sounds like an undifferentiated mass of meaningless sounds. With time the child learns to pick out distinctive features, basic sounds that he can use to discriminate words, meanings, and other aspects of the language.
Perceptual learning, then, according to Gibson, has two components: First, the person must learn what the distinctive features are (e.g., according to what criteria does a good wine differ from a bad wine?). Second, the person must learn how to use the distinctive features to discriminate different relevant objects. At the start of a task a person might already be able to identify the distinctive features. If so, then perceptual learning is facilitated, as the person has only to learn how to use these features. Memory of an object is conceived as being stored in terms of distinctive features and invariant patterns, as opposed to just an unanalyzed copy of the stimuli.
Learning to identify and to respond to distinctive features involves processes such as abstraction, filtering out irrelevant variables, and selective attention. Abstraction involves distinguishing common elements or relationships. For example, in language learning we learn to identify (abstract out) certain basic sounds independent of differences in pitch, loudness, or speed. Filtering out irrelevant variables is learning to ignore those parts of the stimulation that are not essential to the required perception, as pitch might be irrelevant to the understanding of some parts of language. Selective attention, according to Gibson, refers to the exploratory activity of the sense organs, such as turning the head toward a sound or rolling a liquid over the tongue.
Let us consider the implications of Gibson’s theory in a practical situation: teaching. To facilitate perceptual learning, the teacher should emphasize distinctive features, for example, through the use of clearly contrasting examples. This technique would apply equally to elementary children learning to read, to high school students learning to discriminate among the sounds of different musical instruments, and to a medical student learning to make sense out of electroencephalogram records. First the teacher would help the students to identify some of the distinctive features; for example, pointing out cues for the student to discriminate between the string instruments. Then through the systematic use of contrasting examples the students would be given practice using the distinctive features to make the required discriminations.
Transactional Theory. An example of a theory that emphasizes extrapolations from limited information is transactional theory (Ittelson & Cantril, 1954). Perception is considered to be dependent on the person’s past transactions with the environment; it is an active process of interpretation of environmental events in terms of the person’s purpose, values, and past learning (e.g., expectations and assumptions). A stimulus pattern on the retina could have come from a variety of different objects; hence there are a number of different possible “perceptions.” The actual perception that the person has thus depends on his past learning: how functional and useful were the different possible perceptions in the past?
A number of impressive demonstrations were generated in support of transactional theory (Ittelson & Kilpatrick, 1951). One of the more outstanding of these is the trapezoidal window. This is a trapezoid frame with window-like panes in it and shadows painted on it to make it look like a window (a schematic is given in Figure 3—2). When seen in the proper setting a person can perceive the window in at least two different ways: (1) It could be seen for what it is, a trapezoid-shaped window; or (2) It could be perceived as a rectangular shaped window seen from an angle. Because most people’s transactions with windows in the past have been primarily with rectangular windows, the trapezoid window is generally perceived as a rectangular window.
Now consider what happens when the window is slowly rotated about the post shown attached to the middle of the base of the window. If the person perceived the window as being a trapezoid, he would simply see a rotating trapezoid. But he perceives the window as a rectangle. A rotating trapezoid does not produce the same types of retinal images as a rotating rectangle would. In order to fit the actual retinal images into the rectangle “hypothesis,” the person’s perception becomes distorted. What he immediately sees is an oscillating window whose speed, shape, and direction of turning seems to keep changing. It may seem to slowly move around in one direction and then to suddenly dart around in the other direction. If a rod is hung through the window pane, it often appears to move in a direction opposite that of the window. Sometimes the rod appears to bend, break, or pass through solid parts of the window; Thus in order to maintain seeing the trapezoidal window as a rectangle, most people will literally perceive the window and rod as doing things that they know are impossible. For instance, this author always sees these perceptual effects even though he “knows” the window is a trapezoid and “knows” it is turning at a constant rate.
The transactionalists have made a good case for the influence of learning on perception. However, as Gibson (1969, p. 45) pointed out, they have not included a clearly formulated explanation of how this learning takes place.
Categorization. Some theorists, such as Bruner (1958), have conceived of perception as influenced by categorization. The categories by which things are classified are generally a result of learning. Thus when some object or event is first perceived it is classified according to a system of categories. The final perception then depends on how the sense data were categorized at the time they were first perceived. Any difference between the actual object or event and the category under which it was classified may result in a distortion in the perception of the object or event, in order to make it fit the category. This “distortion” may often be nothing more than the selective perception of some features of the object over other features. Also, differences between categories and environmental events may produce modifications of the categories.
Consider playing cards in which the colors of the suits are reversed: spades and clubs are red, while hearts and diamonds are black. These cards do not fit into the card categories of an experienced card player. If quickly shown a red ten of spades, the card player’s perceptual system tries to map the sense experience into a category such as black spades, red hearts, or red, diamonds, but there is no category for red spades. Thus the person might misperceive the red ten of spades and perceive it as the black ten of spades or the red ten of hearts. In an experiment with such cards, Bruner and Postman (1949) identified two frequent types of errors: dominance reactions and compromise reactions. A dominance reaction occurred when the subject forced the suit to match the color or the color to match the suit. Perceiving the black five of hearts as the red five of hearts is a dominance reaction. In a compromise reaction the subject perceives some compromise between the actual object and a dominance reaction. For example, a red spade might be perceived as a purple spade (purple being a compromise between red and black), or as a black spade with red edges.
The author has introduced such a color-reversed deck into bridge games with experienced bridge players. The usual response by the players when they first pick up the cards is that there is something funny or unpleasant about the cards, but they don’t know exactly what it is. Many players play an entire hand without seeing what is actually different about the cards. Such players often have trouble sorting the cards into suits, suddenly noticing that they have five suits in their hands. Even when the color reversal is noticed, following suit during the play of the hand is often difficult. This author still gets a slightly unpleasant feeling from looking at such cards as a black jack of diamonds. On the other hand, the author’s wife, who does not play cards (and hence does not have as set a group of categories) never had any trouble immediately seeing the cards as they actually were.
As another example of categorization, consider prejudice. The roots of prejudice are many and varied. Aronson (1972, p. 180) lists four basic causes of prejudice: (1) economic and political competition or conflict, (2) displaced aggression, (3) personality needs, and (4) conformity to existing social norms. From the perspective of the preceding discussion it could be argued that an important variable in many cases of prejudice is the type of categories the prejudiced person uses in perceptually classifying people.
Consider a person with the following three categories for automobile drivers: good male drivers, bad male drivers, and bad female drivers. Now assume that a female driver passes by this person, but so fast that he doesn’t get a clear look at the sex of the driver. Assume also that the female driver displays particularly good driving skills. Because our prejudiced person has no category for good women drivers (she’s a black queen of hearts in his world), he misperceives the situation and perceives the driver as being male (a dominance reaction).
We can see how such a categorization argument can be applied to many forms of prejudice. Bruner (1958, p. 86) suggests. “We see a Negro sitting on a park bench, a Jew or Texan changing a check at a bank window, a German dressing down a taxi cab driver, and allocate each experience to an established and well-memorized stereotype: lazy Negro, mercenary Jew, rich Texan, bullying German.” Now there are many reasons why a black might be sitting on a park bench, few of which are because he is lazy. Perhaps he is on a break from a ten-hour-a-day job. But if the prejudiced person has only one category for blacks, a category that includes being lazy, then the prejudiced person’s perception of the black may be distorted. Worse still, when this person remembers the scene of the black in the park, his memory includes all of the distortions he originally added to the perception. Such a person perpetuates his own prejudice because his misperception and distorted memories are proof to him of the validity of his stereotypes.
This type of distortion is illustrated in an early study that Allport and Postman (1945) did on rumors. (Please remember that the study is 30 years old, and some of the specific findings might be different today, although the psychological processes are assumed to be the same.) In their study they would show one subject a picture which he would describe to a second subject, who then told a third subject, and so forth. This way the experimenters could observe the types of changes that took place as the story was passed on. Some of the results could be interpreted in terms of our categorization model. One picture was a subway scene that included a white man holding a razor while arguing with a black man. In over half of the final stories the black ended up holding the razor, Perhaps for many of the people a razor during an argument better fit the black category than the white man category. In some cases the number of blacks increased to four or “several.”
Although the preceding examples might be fairly extreme, it can be argued that everyone has a limited number of categories and so must be misperceiving some events. One purpose of education then is to increase the number of categories a person has and uses in order to decrease the amount of misperception.
The Hebbian Model. Hebb (1949) offered a provocative theory of perceptual learning in his book Organization of Behavior, The theory suggests that there are neural representations that correspond to environmental stimuli, and that learning involves neuronal associations between such representations. According to Hebb, simple visual perception can be broken down into small units such as lines and angles. With learning, these basic units form into simple figures and then into more complex perceptions. Hebb explains this learning in terms of associations between neural units. For example, one set of neurons might respond to a particular angle, while another set responds to a particular line. The perception and memory of a figure that includes this angle and line require a learned association between the two sets of neurons. In other words, we start seeing very simple things and gradually learn to be able to see more complex perceptions.
In developing his theory, Hebb drew heavily on the work of the German ophthalmologist von Senden (1960). Von Senden studied adults who had been virtually blind since birth and then were suddenly given sight by an eye operation such as removal of cataracts. Hebb distinguished two processes of perceptual development: figural unity, the simple detection of an object against its background, and identity, identifying an object as a member of class of objects. Von Senden’s patients, when given sight, were generally capable of figural unity, but seldom capable of identity. They could fixate on objects and follow moving objects with their eyes, but at first could not identify objects. In the beginning the patients relied a lot on color. If the shape of an object were changed but the color left the same, the patients often still identified the object as being the same.
Although von Senden’s subjects could detect a square or triangle against its background, they could not at first tell one from the other unless they counted the number of corners. Similarly they could not tell which of two sticks was longer unless they felt the sticks. Even when they learned to identify some objects by sight, a change in the physical orientation of the object might make it unknown again. With time the subjects learned how to visually identify more and more objects, but for many of the subjects their visual skills never became “normal.” Two years after the operation one patient could identify only four or five faces.
Hebb argued that the type of perceptual learning seen with von Senden’s subjects corresponded to what occurs with normal infants. The advantage of von Senden’s subjects is that they could verbalize what was taking place. Although correspondences between von Senden’s patients and infants may exist, there are too many differences to enable us to draw any firm conclusions. For example, von Senden’s subjects may have experienced some side effects when suddenly given sight (e.g., dazzle of bright lights or cramps in eye muscles) that impaired their visual progress, which wouldn’t be the case in a normal infant. It should also be remembered that von Senden’s subjects had spent their entire lives learning to interpret the world through sense modes other than vision. Thus it would be expected that these other learned responses related to handling the environment would interfere with the acquisition of new visual responses. In fact, many such patients often prefer their old sense modes to their now confusing vision. Another possibility is that there are certain critical periods in human visual development similar to the critical period discussed in the first chapter regarding imprinting. That is, there might be certain critical periods in the development of an individual in which he is particularly predisposed for some type of learning, such as perceptual learning. If this critical period is bypassed, as with von Senden’s subjects, the learning may be significantly more difficult,
Later support for theories such as Hebb’s came from the neurophysiological studies of Hubel and Wiesel (Hubel, 1963). Using recording electrodes in the visual cortex of cats (striate area of the occipital cortex), they studied what types of external visual stimuli would cause different nerve cells to fire. They found some cells that maximally fired to simple lines presented at one orientation to the eye, while other cells fired maximally to lines at other orientations. Some cells fired to movement of a stimulus in one direction in the visual field, but not to movement in the other direction. This type of cell function appears to be innate in that it can be demonstrated in newborn kittens.
Hubel and Wiesel .categorized the nerve cells they studied into two basic groups: simple cells and complex cells. Simple cells are ones that respond only to line stimuli of a specific orientation and position, while complex cells are more general in what they respond to. Hubel and Wiesel suggest that complex cells receive input from a number of simple cells.
For example, one simple cell might respond only to a dark vertical line in a specific part of the left visual field, and another simple cell might respond only to a dark vertical line in part of the right visual field; whereas a complex cell that receives input from these two simple cells might respond to the vertical line in either place. With more simple cells feeding into a complex cell we can imagine a complex cell that responds to a vertical line anywhere in the visual field. As complexity of the cell increases, we might find a cell that fires to figures of triangles if it gets the right input from cells that respond to horizontal lines plus cells that respond to slanted lines of a certain orientation.
It is easy to see how we could slowly build up a model of vision this way with more and more complex cells. Such a model could be compatible with Hebbian theory, as Hebb also sees perceptions building up from simple components such as those of simple cells. Such models, however, go far past the basic findings of Hubel and Wiesel, and hence should be considered quite speculative. We also have to be careful to not construct a model of perception that simply provides firing neurons to correspond to visual stimuli. A model of perception must be more flexible in order to include such intricate phenomena as visual illusions and the effects of learning on the interpretation of sensory events.
Gregory (1966, Chap. 9; 1968) has shown how past learning might account for a number of visual illusions. Consider the illusions given in Figure 3—3. In the Ponzo illusion the top of the two horizontal lines usually looks longer, although both lines are actually the same length. In the Müller-Lyer illusion the shaft of the first “arrow” with the ends turned out usually appears longer than the shaft of the second arrow with the ends turned in.
Gregory suggests that both illusions suggest depth to the viewer, and that those features of the figures “assumed” to be more distant appear larger. The Ponzo illusion corresponds to experiences the viewer has had with similar figures, such as railroad tracks. Our experience with railroad tracks has taught us that when our eye gets the image of the Ponzo illusion, the top of the figure is actually farther away from us than the bottom. Hence the top of the two horizontal lines is farther away, but both lines produce the same sized image on the retina. If two objects produce the same sized retinal image and one object is farther away, the farther object must be larger, and often will appear larger. So the argument is that somewhere in the perceptual processing of the Ponzo illusion the actual retinal images are compared with information about how far away the different parts of the figures are, and the results of this comparison determine our subjective experience of the relative sizes of the different figure parts. And it is past learning that affects the distance estimation.
Similarly the Müller-Lyer illusion might be explained in terms of our past experience with corners of rooms and buildings. If you look at the inside corner of a room, the line edges formed by the walls, ceiling, and floor, you will see a three-dimensional representation of the MullerLyer arrow with the ends turned out. Note that in this situation the shaft of the arrow is the part farthest away. Now if you look at the outside corner of a flat-roofed building, such as a phone booth, the line edges form a three-dimensional example of the arrow with the ends turned in. Here the shaft is the nearest part of the figure. If we look at the illusion, both shafts produce the same retinal image. But the one shaft is “assumed” to be the nearest part of the figure, while the other shaft is “assumed” to be the farthest part. Therefore we perceive the farther shaft to be larger.
If these illusions can be explained by the depth they suggest, why don’t the illusions look more three-dimensional? This is probably because the figures are printed on paper which superimposes a two-dimensional effect, but not a strong enough effect to offset the illusion. If instead the Müller-Lyer figures are constructed out of wire, painted with luminous paint, and viewed with one eye (to avoid stereoscopic depth information) in the dark, then they do look three-dimensional, like corners.
We have now seen a number of ways in which learning might affect perception. The next question is how the particular language a person learns affects his perception.
The Arabs have a multitude of different ways of naming camels, and the Hanunoo people in the Philippines have names for 92 different varieties of rice (Bourne et al., 1971, p. 285). Do these languages affect the way the person actually perceives and thinks about his environment? In other words, does learning a language affect later perception of the world? Is the Arab’s perception of camels different from ours, or does he just use the available information differently?
Whorf offered an interesting theory relative to these questions, called the Whorfian hypothesis or the linguistic relativity hypothesis (see Bourne et al., 1971, Chap. 13; Carroll, 1956). According to this theory, language is not simply a medium of communication and thought. In addition to these generally accepted functions of language, Whorf contends that the structure and semantics of any particular language mold the way a person perceives, understands, and responds to his environment. Similar to the process of categorization, language provides a framework for the person’s perception and storage of information. According to Whorf we dissect nature along lines laid down by our native language.
For example, Whorf noted that English grammar tends to divide sentences into noun phrases and verb phrases. He suggested that an effect of this grammar was that English speaking people have a tendency to analyze all of their experiences in terms of one of two categories — things or actions. (If your counterargument is that this is the only or best way to divide experiences into categories, then you are proving Whorf’s point.) Whorf spent considerable time studying the Hopi Indians, and many of his examples come from these studies. It appeared to Whorf that the Hopi did not have tenses for their verbs. From this Whorf concluded that the Hopi perception of the world must then be timeless. The Hopi language also had no word for imaginary space, which suggested to Whorf that the Hopi could not even imagine something like a missionary’s hell. However, many of Whorf’s conclusions about the Hopi have since been questioned.
Although the Whorfian hypothesis may be true to some extent, there are too many confounding variables to determine its exact status. Consider the problems in showing how language affects thinking. First of all, most of what we know about another person’s thinking processes comes from what the other person tells us, and so we are using the person’s language as a measure of the effects of language on thinking. This boils down to showing the effects of language on language, which isn’t too revealing. Secondly, there is the complication that most of thinking revolves around language. (Try to “think” about some topic without using words.) Therefore language must affect thinking, since it is one of the components of thinking.
But our question is whether language affects perception. Does one of the Hanunoo people, who can discriminate 92 varieties of rice, literally see rice differently than we do? Perhaps not. It may be that the importance of rice to these people plus their greater experience with different types of rice simply allows them to make more and finer judgements. The fact that they have more words for rice than we do in English simply reflects their ability to make more discriminations. That is, rice “looks” the same to them as it does to us, but they know more things to look for and have more words to classify what they see.
Thus one of the Hanunoo can code in one word (one of the 92 varieties) a lot of information about the rice. When he tells one of his friends which variety of rice he is dealing with, considerable information is exchanged. An English speaking person who knew what to look for might be able to code the rice with all of the same information, except that instead of a single word, his identification of the rice might involve a number of short descriptive phrases. One-word coding, being more efficient after you have mastered the coding process, may facilitate learning, remembering, and thinking about rice. But that is quite different from saying that the original perception is affected, So at this time we can’t say whether language affects original perception and/or only affects other processes, such as how the information is coded. The distinction between perception and coding is also far from clear.
In a relevant experiment, Brown and Lennenberg (1954) categorized a set of color discs according to “codability.” A color disc with a high codability score was one which most of their subjects gave the same name (e.g., “red”), while a disc with a low codability score was one given many different names and descriptive phrases (e.g., “dirty reddish green”). Next, a different set of subjects was shown some of the individual discs and had to match each color with one of the discs from a large display of all the color discs. When these subjects had to find one disc at a time, it made no difference whether the disc was of high or low codability. However, when the subjects were asked to find four discs simultaneously, they were faster and more accurate at finding discs of high codability. When looking for just one disc the subjects simply kept a picture of the disc in their mind as they scanned the display of discs. Doing four at a time, however, depended more on how the subjects coded the colors.
Gibson’s (1969) position on language and perception is that the person first learns to perceive objects and their features and later learns names for these objects, as opposed to Whorf’s position that language affects the original perceptual development. Gibson questions whether perceptual learning is appreciably affected by language categories, although she does allow that perceptual learning can be facilitated by calling attention, as with language, to distinctive features of the objects.
The discussion in this chapter has shown that stimuli do not fall on passive receivers. Rather, each person is predisposed to perceive stimuli in specific ways. This predisposition has sometimes been referred to as set, an ambiguous, generic term that encompasses a range of variables, including past experiences, motives, context, rewards, and instructions (see Dember, 1960, Chap. 8; Haber, 1966). In fact, the concept of set includes everything discussed so far in this chapter.
Think about the answer to the following question before reading further: Why are 1972 pennies worth almost twenty dollars? The answer is that you need two thousand pennies for twenty dollars and so you are only twenty-eight cents short. Most people have some trouble with this question because they are in the set of perceiving 1972 as a date, not as a number of pennies. But is this effect really on the perception of the 1972 or is it on some other stage in the processing of the concept of 1972, or possibly both? This is a key, and as yet unanswered, question.
The next example will illustrate how set-like effects can affect problem solving processes. Quickly answer the following question before reading further: If polk is pronounced poke with the l silent, and folk is pronounced foke, also with a silent l, how do you pronounce the white of an egg? Most people who answer quickly do not realize the white of an egg is the albumen, not the yolk, for they are in the set of words ending in lk or ke.
Figure 3—4 gives two other examples of set. Read each example fairly quickly and then go back and look for the set. The following are the answers for those who want them: In the first example the word “the” is printed twice, but most people read over one of the “the’s” because “bird in the hand” is such a common phrase. In the second example the last word may be read as a Scottish name because of the set established by the first three words, but it can also be the common word “machine.”
Let us now consider how set can influence people’s perceptions in a classroom situation. Kelley (1950) used students in a college psychology class. The experimenter came into the class one day and told the students that their regular instructor was to be replaced by a substitute for the day. The students were then given a written description of the substitute. However, the descriptions were not all the same. Half of the descriptions referred to the substitute as being a rather cold person, while the other half were the same except the word “warm” was used instead of “cold.” During the class that the substitute taught, the students who got the “warm” description participated more in class discussion with the substitute than did the students who got the “cold” description. After the class, the “warm” description students evaluated the substitute higher than did the “cold” description students, in terms of being more considerate, better natured, and so forth. All students were in the same classroom with the same substitute, but, according to the set they were in as a result of the written descriptions, they perceived the substitute differently and interacted with him differently. The substitute’s actual personality and teaching style, of course, affected the students’ ratings and perceptions in an absolute sense, but set made the difference between the two groups of students. It is easy to see how this type of phenomenon takes place all the time in classrooms, as students tell other students what they think or have heard about a particular teacher. Similarly, people perceive and respond differently to political figures (or anyone else for that matter) and to their speeches, depending on their particular “set.”
Set also affects a teacher’s perception of his students. Rosenthal and Jacobson (1968) told elementary school teachers that they had a test that would identify “spurters” — sudden fast learners. In fact, the students they identified as spurters were chosen randomly; thus any difference between spurters and non-spurters was purely due to the set of the teachers’ minds. Over time, the “results indicated strongly that children from whom teachers expected greater intellectual gains showed such gains.” The teachers also described the spurters as being happier, more curious, more interesting, more appealing, better adjusted, more affectionate, less in need of social approval, and as having a better chance of being successful in later life. The non-spurters also improved intellectually, but the more they improved, the less favorably they were rated by the teachers. This effect of set was powerful, particularly on first grade teachers.
How did the teachers’ sets influence the students’ intellectual gains? Rosenthal and Jacobson argue that it was not simply that the teacher spent more time with spurters but that the effect lay in more subtle interactions: “Her tone of voice, facial expression, touch and posture may be the means by which — probably quite unwittingly — she communicates her expectations to the pupils. Such communication might help the child by changing his conception of himself, his anticipation of his own behavior, his motivation or his cognitive skills.”
(e.g., Clairborn, 1969) have criticized the methodology of Rosenthal and Jacobson
and have failed to replicate their findings. O’Leary and Drabman (1971) conclude
that “At most, the evidence
Keeping Rosenthal and Jacobson’s theory in mind, consider what might be happening to students classified as slow learners, problem students, or special education students. Suppose a student does something that could be perceived as either creative curiosity or bothersome digression. The “spurter” might be rewarded and encouraged, whereas if the teacher is in the set of thinking of the student as a problem student, the student might be verbally punished and discouraged. What a shame this would be if the behavior had elements of curiosity that could have been encouraged.
Those who work with teacher training try to minimize such effects of set by discouraging the labeling of students and encouraging well-specified behavioral objectives and systematic keeping of behavioral records. That is, if the teacher decides exactly which behaviors should be encouraged and which should be discouraged (regardless of who does them) and how he can objectively determine which kind of behavior has occurred in a given situation, then the effects of set will be dramatically reduced.
Many clinicians utilize projective tests to aid in the personality assessments of their clients. These tests consist of relatively unstructured stimuli that the client must organize or interpret in some way. For example, the person might be shown an inkblot and asked to tell what it looks like. The assumption is that the responses that the subject makes to the projective test are some measure of his personality. Unfortunately such responses might be affected by set. The clinician could, although subtly and unintentionally, influence the subject’s set so that the subject will respond to the projective test in ways that fit the clinician’s expectations or theoretical bias. It is possible that many clinical phenomena, such as the types of symbols that a person reports as having occurred in his dreams, are influenced by set.
As mentioned earlier, we cannot at this time say for sure whether set affects the original perception or some later stage of information processing, or perhaps both. Haber (1966) discusses two contradictory hypotheses about set: (a) set affects the percept of the stimulus while the person is actually viewing it, and (b) set affects the report of the stimulus without affecting its percept. Haber summarizes as follows: “This review must conclude inconclusively with respect to a choice between the two hypotheses. Some evidence exists to support each of them, and some exists which favors one over the other. But there is none that supports one while disproving the other.” It does appear, however, that learning is a major variable affecting set, and that set may affect perception. This suggests again the possibility that learning plays a role in perception. Next we will consider how learning and perception interact in the learning of verbal material.
Different types of verbal material vary in their ability to elicit images. Does this image-eliciting ability affect how easily the material can be learned and remembered? This is the question investigated by Paivio (1969). Paivio classified verbal stimuli along a dimension from concrete to abstract. A concrete stimulus, such as the word “house,” is more likely to evoke images than an abstract stimulus such as the word “truth.” (Say each of these words to yourself and see which elicits more images.) Paivio suggests that concrete stimuli derive their meaning through association with specific objects and events as well as through association with other words. Learning of concrete material, then, could utilize the images or verbal associations, or both. Abstract stimuli, on the other hand, derive their meaning largely through associations with other words. Thus learning of abstract material would primarily utilize verbal associations.
Paivio often used a form of paired associate learning of noun pairs. Paired associate learning involves presenting the subject a number of paired items that the subject must learn to associate together. In Paivio’s task the subject, when presented with the first noun of a pair (the stimulus), had to learn to say the second noun (the response). Paivio found that learning was faster if the stimulus noun was concrete than if it was abstract. For example, it is easier to learn an association to the word “house” than to the word “truth.” Paivio suggests that images serve as “conceptual pegs” to which responses can be conditioned. That is, concrete nouns elicit more images than abstract nouns, and these images form the basis for learned associations. If given the pair “house — dog” the subject can easily conjure a scene involving a house and a dog, which facilitates learning and/or memory. It is not so easy with the pair “truth — dog.” Thus a person learns images to certain stimuli and these images facilitate later learning that involves the stimuli.
Paivio showed that the effects of noun-imagery were greater on the stimulus side than on the response side. Although having a concrete response noun might yield better learning than an abstract response noun, it is more important to have a concrete stimulus noun. The effects of noun-imagery were also found to be relatively independent of how meaningful the material was to the subject, i.e., how many associations the subject already had to the specific material. Although meaningfulness and noun-imagery often go together, in Paivio’s tasks the imagery had a greater effect on learning.
The rest of this discussion of verbal learning centers on how the subject encodes stimuli. Whether or not this should be called “perception” depends on the definition of perception. This author includes under perception all processes of cataloguing information. Memory and retrieval processes begin after this. Others have defined perception so that its domain stops earlier in the information processing.
Consider paired associate learning in which the subject is presented with one or more pairs of items and must learn to associate the members of the pair. Paivio’s task above was a form of paired associate learning. Another task might have pairs such as LUF—ZIJ, where the subject must learn to make the response ZIJ when presented with the stimulus LUF. A critical part of paired associate learning must be learning to tell the stimuli apart, a process called stimulus discrimination. Early theories of paired associate learning (e.g., Gibson, 1940) emphasized the role of stimulus discrimination. Now this emphasis has shifted somewhat to how the subject encodes the stimuli.
According to stimulus encoding theory the subject translates stimuli into forms that are easier to use in the current task. A visual stimulus might be coded into a verbal phrase and stored verbally. Or the stimulus LUF might be encoded as “love” for easier processing. Martin (1971) has argued that “a major portion of learning is perceptual learning—learning an effective identifying response to the nominal stimulus situation.” Let us say that a person learns one set of paired associates and then has to learn another set which consists of the same stimuli but different responses. (This two-stage learning is referred to as the A-B, A-C paradigm.) According to Martin the subject in learning to do this might learn to code the stimulus differently the second time; that is, he might make a different identifying response to the same stimulus.
Perception, learning, and motivation all come together in explanations for animals’ (including humans) apparent need for sensory complexity. Animals strive for variety, novelty, and complexity in their environments even though such striving does not seem to serve any immediate biological need. There is a vast literature on such phenomena (Berlyne, 1960; Dember, 1960, Chap. 8; Eisenberger, 1972; Fiske & Maddi, 1961), which includes the following examples. Bees prefer those flower shapes with the longest outline with respect to surface area. Some fish will learn mazes just to look in a mirror. In mazes with many correct paths to the goal, rats vary their paths and often prefer those paths with greater variety. Rats will also learn a simple maze where the reward is the opportunity to explore another maze. A rat will press a bar simply to turn a light on and then will press another bar to turn it back off again. Monkeys like to handle objects, and show preference for the more heterogeneous objects. Monkeys will also open windows or pull levers to see outside their cages, and will keep doing this, particularly if the environment keeps changing. Coming home from work, a man might decide to take a different, perhaps even longer route, just for a change in his routine. Women rearrange their living room furniture for similar reasons. Teachers find that almost any significant change in the classroom (painting the walls, installing new blackboards, introducing a new audio-visual device, making new seating arrangements) seems to improve learning for a while. In Chapter 7 we will see how need for variety might be a major personality variable.
Phenomena such as those just noted have been described under many names: exploration, novelty, curiosity, stimulus change, and stimulus satiation. Is there something common to all these phenomena, some theoretical construct under which they all fall? One answer to this question centers on stimulus complexity and animals’ attempts to experience a certain degree of complexity: a novel stimulus is often more complex than a familiar stimulus. Exploration and curiosity are simply attempts to increase the stimulus complexity of the environment. Unfortunately there is no good independent measure of complexity, although there have been attempts to measure it in terms of information theory, conflict, or specific stimulus attributes. A useful way of thinking about complexity, following Dember (1960, p. 352), is that “the more complex stimulus is the one the individual can do more with: it affords more potential opportunities for responding than does the less complex stimulus.” Most experiments, however, simply use stimuli that intuitively differ in complexity. We are nowhere near the point where we can take any two stimuli, particularly if they are of different sense modalities, and say that for a given organism one stimulus is more complex than the other.
To explain the effects of stimulus complexity many theorists use the concept of arousal. Arousal is a general excitatory process—a nonspecific drive — perhaps related to the activity of the reticular formation, a neural system in the brain stem. Although most studies measure arousal in terms of some physiological phenomenon (skin resistance, pupil size, EEG, heart rate, respiration, blood pressure), there is poor correlation among changes in the various measures. This raises the questions of which measure is best and whether there is more than one type of arousal.
Many theories have been offered to interrelate complexity, arousal, and behavior (see Eisenberger, 1972). A few of these will be mentioned under the following categories: minimum arousal theories, arousal induction theories, and optimal arousal theories.
Minimum Arousal Theories. The general orientation of minimum arousal theories is that arousal is a measure of deviation from an optimal state, so the less arousal there is, the better. Arousal might be produced by many different variables, including states of deprivation and noxious stimulation. Malmo (1958) suggested that drives could be broken down into general arousal plus a directional component. Since reduction in the drive was considered by Malmo to be rewarding, Malmo would be a minimum arousal theorist.
Dember (Dember, 1960; Dember & Earl, 1957) describes a pacer theory of complexity that doesn’t mention arousal per se, but that can be considered akin to arousal theories. According to this theory each animal has a preferred level of stimulus complexity. The animal will seek out that stimulus situation whose complexity is near his preferred level. Being forced to attend to stimuli that are too complex, or not complex enough, causes emotional disturbance. A pacer is a stimulus whose complexity is slightly higher than the animal’s complexity level. When an animal interacts with a pacer the preferred complexity level of the animal moves toward that of the pacer. Thus the animal’s complexity level keeps rising as its experience with pacers increases.
A baby’s complexity level is very low at first. He is very content with low complexity stimuli that might bore an adult, and can be overwhelmed by fairly complex stimuli that are pleasing to an adult. To “protect” himself, the baby might not attend to complex stimuli or he might screen them out early in perceptual processing. As the baby grows and encounters pacers, his complexity level rises and he seeks stimuli of greater and greater complexity.
Rather than speaking of a single complexity level for each animal, Dember suggests that there may be different complexity levels for different types of stimuli. Thus a person’s complexity level for music might be significantly higher than his complexity level for literature, probably because in his life he has had more experience with music and thus encounters more music pacers.
Pacers are usually pleasing because they are different enough to be not boring, but are not so complex that they are disrupting. Humor often involves situations that are somewhat unexpected, but not so strange that we strain to make sense of them. Beethoven is said to have remarked that in music everything must be at once surprising and expected.
Although pacer theory emphasizes the continual rise in complexity levels, one wonders if complexity levels ever decrease. Are there negative pacers which when encountered lower the animal’s complexity level? It may be that although an animal’s complexity level along some dimension generally rises, it does fluctuate back and forth, including many short term decreases.
If we assume, as some theorists after Dember have done, that discrepancies between an animal’s complexity level and the stimulus complexity produce arousal, then pacer theory is a minimum arousal theory, for animals work for minimum discrepancies between their complexity level and that of stimuli.
Pacer theory assumes that if the complexity of the stimulus situation is too far from the animal’s preferred level, there will be emotional disturbance. Chimps are frightened by a model of a chimp’s head without a body. They have always seen heads on bodies; a head without a body is too complex. Human infants are often distressed if they hear a strange sound coming from a familiar face or a familiar voice coming from a strange mask. Many unpleasant experiences with hallucinogenic drugs such as LSD result from the person’s being overwhelmed by sensory and thought experiences unfamiliar to him.
The effect of stimuli of too little complexity may be simply boredom. In more extreme situations, such as sensory deprivation, the effects are more pronounced. Sensory deprivation is not so much the depriving a person of stimuli as it is a drastic reduction in stimulus complexity. In some of the first studies (Bexton, Heron, & Scott, 1954), college students were paid to stay in a room lying in a bed. To reduce visual complexity their eyes were covered with translucent goggles. Auditory complexity was reduced by the person keeping his head in a U-shaped foam rubber pillow and hearing the hum of the air conditioner. Tactile complexity was reduced by having the person wear cotton gloves and cardboard cuffs that extended beyond the fingers. The students stayed this way 24 hours a day with time out only to eat and go to the toilet. Most subjects lasted only two to three days, although a few were able to last longer.
The effects of sensory deprivation on the subjects were quite disturbing. The subjects were restless, often displaying constant random motion. Their feelings vacillated between anger and mirth. Perhaps most distressing was that they could not think clearly for any length of time. It appears that the functional organization of the mind that we call rational thinking requires a certain amount of environmental support in the way of stimulus complexity. The subjects also had hallucinations, similar to drug-induced hallucinations, that ranged from simple objects to complex scenes, such as rows of yellow men with caps on or squirrels with packs on their backs marching purposively along. The hallucinations weren’t totally visual but also often included specific voices or sounds, and specific feelings. This seems to be the same type of phenomenon as the hallucinations reported by aviators during long flights, by truck drivers during extended trips, and by radar screen watchers whose shifts are too long. The explanation is probably that when the environment does not provide enough complexity, the mind draws from other sources.
While some of the subjects were in sensory deprivation they heard a recording of a talk arguing in favor of the existence of ghosts and supernatural phenomena. For some reason the talk in this situation was particularly persuasive. Some of the subjects said that for days after they left the sensory deprivation they were afraid they would see ghosts. This finding suggests that sensory deprivation might be a powerful influence technique, but almost no one has seriously investigated it. Adams (1965) put a hospitalized mental patient in mild sensory deprivation and presented to him a taped message discussing his particular case and the aims, procedures, and rationale of psychotherapy. Adams reports that this message resulted in general improvement in the subject. Part of the effect of sensory deprivation is that it focuses the subject’s attention on the message, minimizing most sources of distraction. Unfortunately Adams had only one subject in this report and the results are confounded with other parts of the treatment program.
Arousal Induction Theories. There are no true arousal induction theories — theories that argue that animals seek to experience the most arousal they can. Such a theory would be at variance with too much data. But several theories are almost arousal induction theories. Some theorists, such as Sheffield (see Chapter 6), argue that stimuli that elicit arousal have a greater determining effect on the animal’s behavior. Thus the animal may often behave in ways that, at least temporarily, raise his arousal, for the stimuli that increase arousal have a greater effect on the behavior than other stimuli. This will be explained in more detail in Chapter 6.
Miller (1963) suggested that there are one or more go-mechanisms in the brain which intensify ongoing responses to cues and traces of immediately preceding activities. These go-mechanisms are activated by events such as rewards, drive reduction, and the removal of discrepancy between intention and achievement. Thus if a rat learns to press a bar when a light comes on in order to get food, the reward of the food activates a go-mechanism which intensifies the response of pressing the bar to the cue of the light. Miller also suggested that the go-mechanism can become conditioned to the occurrence of the response so that future occurrences of the stimulus will elicit both the response and the excitatory state. If one equates the excitatory state of Miller’s go-mechanism with arousal, then Miller’s theory comes close to being an arousal induction theory.
Optimal Arousal Theories. According to optimal arousal theories there is a level of arousal that is ideal for each animal. If its arousal is too low, the animal will seek stimuli to increase arousal, whereas if arousal is too high, due to fear or hunger, for example, the animal will try to do what is necessary (e.g., flee or find food) to reduce the arousal to its optimal point.
Walker (1964) suggested a theory stating that the more complex a stimulus is the more arousal it elicits. An organism seeks those stimuli that elicit optimal arousal. However, according to Walker, as an animal experiences a stimulus, there is a decrease in the complexity of the stimulus relative to the animal. Thus stimuli that may have maintained optimal arousal for a while lose complexity, and the animal seeks out other stimuli. (In pacer theory the complexity of stimuli stays the same, and it is the animal’s complexity level that changes. In Walker’s theory the animal’s complexity-arousal level is constant while the complexity of stimuli changes.)
Walker equates the effects of rewards with arousal. This yields the following interesting prediction: “The rewarded event should undergo progressive and selective reduction in psychological complexity. Eventually it should reach a level of psychological complexity that is lower than that of the unrewarded alternative.” That is, rewards affect behavior as they do because of their effects on arousal. But as the animal encounters the same reward for the same response over and over again, the reward reduces in complexity, thus reducing its effect on arousal. If this were continued long enough the animal should abandon the rewarded response in favor of another response — probably non-rewarded — that produces more arousal. Most experiments, however, are terminated long before this point would be reached and observed. Unfortunately, there is little research on this prediction of the theory, although Walker does give some suggestive data. Partial support of the prediction occurred in an experiment (Walter & Mikulas, 1969) in which rats were tested each day for over 5 months in an operant chamber where they pressed a bar for food. The rates of bar-pressing decreased over time, which would seem to indicate that the food had lost some of its rewarding value.
Berlyne (1967) has offered an optimal arousal theory in which variables such as novelty, complexity, and ambiguity produce conflict, which in turn produces arousal. For example, seeing a novel stimulus, such as a picture of a dog’s body with a bird’s head, produces some conflict which in turn increases arousal. The greater the conflict, the greater the arousal. According to Berlyne, moderate increases in arousal (or decrements if the animal is already highly aroused) activate a reward system. This system underlies learning based on approach responses and pleasant feelings (e.g., positive reinforcement, appetitive classical conditioning). Thus an animal will seek out and be rewarded by a stimulus which produces a moderate increase in its arousal. High increases in arousal activate an aversion system which underlies learning based on avoidance responses and displeasure (e.g., punishment, defensive classical conditioning). Too much conflict and too much arousal create aversion and will be avoided. Activation of the aversion system is assumed to inhibit the reward system.
On a practical level, complexity theory might be helpful in many disparate areas. For example, in education it might help us to match the complexity of material to be learned with the student’s optimal complexity level. Or workers in a plant might be shifted among various positions in order to maintain sufficient complexity for optimal performance.
Having seen some possible ways in which learning and perception interact to affect information entering the system and how it is interpreted, we turn in the next chapter to a discussion of the possible stages the information goes through while being processed.
Man does not passively perceive his environment. Rather he selects, filters, and interprets environmental stimuli, largely on the basis of his past learning. Thus man’s subjective perception of everything from simple objects and visual illusions to complex social interactions is based on the interplay between his sensory-perceptual mechanisms and what he has already learned. Even what man attends to is partly determined by learning. On a broader level is the concept of set — a predisposition to perceive and respond to stimuli in specific ways. Set is influenced by a variety of factors, including past experiences, motives, context, rewards, and instructions.
An important and unresolved question is whether learning ever affects the actual initial perception of an object or if it affects only the responses made to the object. The Whorfian hypothesis suggests that the way a person perceives his environment is molded by the nature of the language he has learned. But do people with different languages actually see environmental objects differently or do they merely respond to the objects differently or process the information differently?
To a large extent people must learn how to perceive, which is called perceptual learning. There are basically two categories of theories of perceptual learning. The first category, which includes Gibson’s perceptual differentiation theory, assumes that the environment supplies most of the needed information and that we must learn how to use this information. The second category, which includes transactional theory, assumes that the environment supplies inadequate information, and thus we must learn to make extrapolations from this information.
In verbal learning studies it has been shown that it is easier to learn associations to words that elicit a number of images than to words that elicit fewer images. Also, the way a person codes the stimuli he is perceiving affects the ease and nature of associations he learns to the stimuli.
Animals, including humans, strive for variety, novelty, and complexity in their environment, even though such striving does not necessarily satisfy any biological need. The complexity of stimuli is assumed to affect the animal’s arousal — the amount of general excitation. Theorists differ on how much arousal they believe an animal will seek out. Minimum arousal theorists assume that the less arousal there is, the better. Arousal induction theorists, on the other hand, emphasize how stimuli that elicit arousal have a greater determining effect on the animal’s behavior. Finally, there are the optimal arousal theorists, who assume that the animal tries to maintain some intermediate optimal amount of arousal.
Dember, W. N. The Psychology of Perception. New York: Holt, Rinehart & Winston, 1960.
Gibson, E. J. Principles of Perceptual Learning and Development. New York: Appleton Century-Crofts, 1969.
Gregory, R. L. Eye and Brain. New York: McGraw-Hill, 1966.
Vanderplas, J. Perception and learning. In Marx, M. H. (ed.) Learning: Interactions. Toronto:Macmillan, 1970.