Perception as a Constructive Process:
Why should we need a perceptual theory? Firstly, it helps choose topics for research, such as ocomotion or illusions. Secondly, it gives a clearer idea of the questions we are trying to answer. In David Marr's terminology, a "computational theory", or what we are trying to explain.
"Indirect" or "Constructivist" Theories vs. "Direct" or "Ecological" Theories:
The origins of the distinction between these two types of theory are partly in the 19th century debate between Hermann von Helmholtz and Edward Hering. Helmholtz was an empiricist and thought that we come to see the world through experience, interacting and experimenting with the world. Hering, on the other hand, was a nativist and believed that innate ideas and innate mechanisms were used to perceive and we do not need to learn to perceive.
To illustrate the differences, a good aspect of perception to look at is visual direction.
The receptors in the eye can only measure three things - intensity, wavelength, and oculocentric direction (direction with respect to the eyeball). All other things that are perceived, such as motion, orientation, and depth, are derived and are secondary.
Visual direction is determined using the concept of "local sign", that is, the idea of receptors being tagged or labelled with a particular oculocentric direction. Helmholtz and Hering agreed that in adults we use local sign to judge visual direction, but they disagreed as to how local sign information was gathered in the first place. Hering believed it was innately given, whereas Helmholtz thought it needs to be learned, perhaps by learning based on the direction and magnitude of eye movements to foveate a target stimulus.
Some evidence in support of the constructivist view is that of Gregory's patient SB, who was blind from birth. SB, when he had his sight restored, could not identify objects that he had not touched.
The Indirect Approach:
Helmholtz thought that perception needs to be learned, and not only that, it must go beyond what is available to the senses. The essence of constructivist theory is seeing perception as unconscious inference. This allows us to seek a meaningful explanation of a pattern of stimulation. An important feature of this is that the inferences that are made implies that there is some sort of choice to be made between alternative interpretations of a stimulus.
Helmholtz (1909) said; "The sensations of the senses are tokens for consciousness, it being left to our intelligence to learn how to comprehend their meaning...The only psychic activity required for this purpose is the regularly recurrent association between two ideas which have often been connected before".
Richard Gregory has put forward a similar view, seeing perceptions as hypotheses, using induction to create hypotheses from sensory data in a similar way to the generation of scientific hypotheses. Another prominent indirect theorist is Irvin Rock, who focussed on the "logic of perception". He sees perception as a result of "intelligent, thought-like processes" and being inferential like problem solving.
There are several key ideas in constructivism:
1. Perception is seen as a construction, not a direct, mechanistic process
2. There is an underlying assumption that sensory information alone is insufficient to attain perception
3. There is an idea of a "meaning" being added to "sensations"
4. Key to the whole indirect approach is Helmholtz's distinction between "sensation" and "perception"
The richness of what we experience when we see is just not present in the input to the eyes. Perception is rather seen as an achievement by empiricists.
In Support of the Indirect Theory:
Rock characterises Gibson's idea of "direct" perception as a "stimulus theory". Gibson's direct theory leads to each stimulus eliciting a single percept. However, Gregory, Rock, and Helmholtz all point to situations where there is not a direct link between a stimulus and the percept produced to support the idea of indirect perception. One such situation is where a single stimulus leads to multiple percepts - for instance, ambiguous figures such as the Necker Cube, which alternates in depth, Rubin's Vase, which spontaneously alternates between a pair of faces and a vase, or Boring's ambiguous mother-in-law picture, which alternates between an old lady and a young woman. In such instances, one cannot see more than one percept at a time. Perception cannot be direct in these cases.
Another situation where perception cannot be direct is when multiple stimuli elicit a single percept. Some examples of this are perceptual constancies - things like size constancy (seeing an object as being the same size regardless of the size of its image on the retina as a function of distance), shape constancy (seeing an object as being the same shape regardless of the angle of viewing), colour constancy (seeing objects as being the same colour regardless of the wavelength properties of the light incident on it), lightness constancy (seeing an object as being the same lightness regardless of the intensity of the light incident on it), and depth constancy (seeing something as the same depth despite changes in disparity). Rock explains such constancies as "taking into account" of these changes. Helmholtz sees it in terms of our knowledge of the world.
There is a third case that suggests that perception is indirect, and that is situations where there is a single stimulus, but it elicits a false percept. Many optical illusions do just this - what is on paper is one thing, but we see something false.
Gregory believed that the most interesting type of illusions were what he called cognitive. Such illusions include the Müller-Lyer illusion, and Ponzo illusions. There are two possible explanations to these illusions. It may be because the length of the lines are the same on the retina, but we see the lines as being of different lengths. Or, it may be that the length of the lines on paper are the same, but we see them as different length. In either case, the illusions are a result of misapplied size scaling. If we imagine the Ponzo illusion in the real world by looking at a road going into the distance, the two sides of it converging, with two lines appearing on it, equal in length on the retina, the lines would have to be of different lengths in order for them to create the same size image on the retina. The Müller-Lyer illusion can be explained similarly by imagining the fins on the lines as perspective lines.
The Moon illusion, where the moon appears bigger when it is lower in the sky, also can be explained by misapplying size scaling. When the moon is high is the sky, there is nothing nearby to compare it with. However, when it is near the horizon, the objects of the environment cause us to see the moon as being bigger. The moon now appears behind the horizon, so the brain scales up the size of the moon to what it believes would cause the same size image on the retina.
Other cognitive illusions include illusory contours. But such illusions exist outside the visual modality - the size-weight illusion shows that we tend to think of a bigger object as being heavier than a smaller one.
Physiological Evidence of the Indirect Theory:
Perceptual aftereffects, such as brightness, motion, and colour aftereffects, show that we are constructing things that are not actually there. More evidence comes from contrast effects seen in such illusions as Mach bands, the Herman grid, the Zöllner illusion, and the Fraser spiral.
Summary of the Indirect View:
The key point is that hypotheses about perception allow us to go beyond sensory data, and also interpolate and extrapolate. Rock, although acknowledging that some illusions are due to the properties of built-in mechanisms, attributes the other illusions to "intelligent, thoughtful, rule-following processes".
Key ideas of the indirect theories:
Perception is constructive
Insufficiency of sense data
Meaning is added (not contained in the stimulation)
Often, but not exclusively, empiricist
Illusions are readily accounted for because there is not direct link between stimuli and percepts
Perception is seen as a consequence of the perceiver
Emphasis on putative mechanisms to explain effects
One might say: "There is more to perception than meets the eye."
Direct Perception and Ecological Optics:
James Gibson has been the main protagonist in putting forward the direct view of perception. From his role in training World War II pilots, Gibson formulated two major ideas in visual perception - the role and importance of gradients, and the idea of optic flow.
There are several different gradients in visual perception - there are gradients of texture element size, motion, and disparity. They arise through a simple consequence of geometry - Euclid's Law (double the distance, halve the angular size). So, a surface covered with similar sized elements creates a gradient of image size - grass and carpet fibres, for example, close to you are quite visible, but as they recede into the distance, they become much less defined. However, we do not see a gradient, the gradient is just a property of the retinal image.
Another key idea in this is that we do not live in an arbitrary world. Instead, we live in a world of continuous surfaces. This means there may be problems with much experimental evidence in perception, since the real world is not always like the conditions that are found in the laboratory. Gibson would argue that it is not surprising if the visual system exploits the properties, or utilises the constraints, of the particular world we live in.
Gibson's second important idea was in contrast with much of the previous work, which had been based on static images. Gibson took into account in his ideas that we do not ever really view a static world, and our eyes do move around. In normal life we are constantly moving around and generating patterns of image motion on the retina. It turns out that these motions follow very lawful and systematic transformations. This optic flow follows a certain pattern, which the visual system uses to attain much information about the world. This pattern is basically that the rate the image moves across the retina is an inverse function of its distance from the viewer.
Gibson argued that there are two sorts of information in the transforming optic array. Exterospecific information gives information about the depth and distance, and the layout of the surrounding world. Propriospecific information gives information about self-movement and is used for such things as balance, ego-motion, and the direction of heading.
Gibson believed that we underestimate the richness of the information that is available to us when we look at perception in artificial or impoverished laboratory situations. This shows the importance of "ecological validity". The distinction between optics and ecological optics is that optics focuses on the properties of projected lines, angles, and points based on abstractions of the world, whereas ecological optics is more concerned with the properties of projected surfaces and other naturally occurring features. Such ideas were novel when Gibson published them in 1950, but no-one would disagree with either the role of gradients or the importance of optic flow nowadays.
The Optic Array:
Irvin Rock, a constructivist, claimed that Gibson's ideas were a "stimulus theory". However, Gibson (1966) rejected the view of the stimulus as the appropriate way to describe the input. Gibson argued that input should be described in terms of information - stimuli excite the receptors, but information excites the perceptual system. Information is contained in the spatio-temporal patterning of light, not in the light itself. This would perhaps make the traditional problem of how we integrate information from the different senses more manageable. Different types of receptor deal with different types of energy, but the information is of the same sort.
As a consequence of this, Gibson rejected the retinal image as the appropriate starting point for understanding perception. Instead, Gibson thought the important thing was the information contained in the "ambient optic array", which is the description of how light is structured at a particular vantage point, and can be described in angular terms.
So why an optic array, rather than the retinal image? Optic arrays are not limited to a particular species - so, there is no image in compound eyes (such as in the Limulus and in Insectoids), yet there is an optic array. Also, it separates out the computational theory from whether we use it.
The idea of optic flow led Gibson to highlight the concept of "transformation", which is defined a "particular style of change". Gibson claimed that the visual system has evolved to detect particular "styles of change" that contain useful information, for instance, expansion patterns contain information about the approach of objects. The significance of this idea is that it avoids the need for stored representations. So, we do not have to think of detecting motion as a process in which the current retinal image is compared with a stored past image, but instead, motion is seen as the extraction of a spatio-temporal change.
Gibson also argued that, so long as we accept the logic of extracting change over a short interval of time (as in detecting expansion), we should accept the possibility of extracting change over a much longer time period, which could be weeks, or even years. This could be used for estimating the age of a person.
Ecological Theory of Perception:
Gibson's fifth main idea was concerning the interrelationship between the perceiver and the environment - hence, an "ecological theory of perception". Gibson said; "If we describe the characteristics of the perceiver, we are effectively describing the characteristics of the animal's particular environment". He also said; "If we describe the characteristics of the particular environment, we are effectively describing the characteristics of the perceiver. There is a mutuality, or synergy, between the perceiver and the environment, so perception is not solely to do with the perciever.
The Link Between Perception and Action:
Traditionally, we think of perceiving in order to act. Gibson's argument is that we also act (i.e. move) in order to perceive. In psychology, processing is divided into sensory systems and perception, cognition, and motor behaviour. The point Gibson was making is that we would only have evolved sensory systems if they provided us with meaningful information which allowed us to act. The traditional distinctions are, according to the Gibsonian view, artificial and misleading. Motor skills and vision have been shown to be tied up together by David Lee and his swinging room, which he used to manipulate subjects' stance.
The Notion of Invariants:
With respect to constancies, the traditional constructivist theory emphasises how the stimulus can change and yet we perceive things as being the same. The constructivist view is that we have to take other information into account in order to correct for changing retinal information. However, Gibson offers the alternative view that perceptual systems have evolved to extract the properties which do not change, that is, the invariant properties. So, in the case of size perception, constancy is achieved by the perceptual system extracting relative size. Relative judgements are nearly always much better than absolute judgements.
The Notion of Affordances:
So what is perception actually for? The constructivist view would say that it is for the creation of representations. However, Gibson suggests that perception is to allow us to act. For Gibson, meaning is the essence of perceptual processing, not something that is "added on", as the constructivists believe. Gibson (1979) argued that the transforming optic array contains meaning, for example, the pattern of light created by a tree stump indicates that it is "sit-able on". This sounds unlikely, but it deserves consideration, especially when we consider animal or infant perceptual systems.
Summary of the Direct View:
Importance of gradients
Information in optic flow
Information as the input rather than stimuli
Concept of transformation
Ideas of mutuality and synergy
Senses considered as perceptual systems
Idea of invarients
Notion of affordances
Role of past experience? Attunement and differentiation rather than laying down memories
So, Who Is Right?
This direct vs. indirect debate has not been completely resolved. There are good arguments for both. However, perhaps the question should be whether they are really mutually exclusive. With so many things in Psychology, the truth may lie somewhere in the middleground.