Lectures: Visual Perception 3

Selectivity and Perceptual Constancy

Categorization is a key 'top-down' process which is involved in perception. Categories simplify. Categorization has a number of functions:

The cost of these advantages is a loss of particularity and uniqueness in perception and recall. For Romantics, it is also regarded as inducing a sense of distance from the world. The way we categorize phenomena seems to be a 'natural' 'reflection of reality', leading us to forget the role of categorization in constructing the world.

Probably the most well-known example of the cultural diversity of categories is that Eskimos have dozens of words for 'snow' - an assertion which is frequently attributed to Benjamin Lee Whorf. Actually, Whorf seems never to have claimed that Eskimos had more than five words for snow (Whorf 1956, 216). However, a more recent study - not of the Inuit but of the Koyukon Indians of the subarctic forest - does list 16 terms for snow, representing these distinctions:

I do not intend to discuss the controversial issue of the extent to which the way we perceive the world may be influenced by the categories which are embedded in the language available to us. I have discussed the so-called Sapir-Whorf hypothesis elsewhere. Suffice it to say that words can be found in English, as above, to refer to distinctions which we may not habitually make, but that this does not rule out the possibility that the categories which we employ may not only reflect our view of the world but may also sometimes exercise subtle influences upon it.

There is research evidence that verbal labels may influence the recall of visual images. In a well-known experiment by Carmichael, Hogan and Walter (1932), observers were shown simple line drawings each of which was associated with either of two verbal captions - e.g. a drawing of two circles linked by a straight line bore either the caption 'eye-glasses' or the caption 'dumb-bells'. The observers were then asked to reproduce the drawings. Their reproductions showed a strong tendency to distort the original image to make it closer to the verbal label which had been attached to it.

The Stroop Colour-Word Test can be used to illustrate the difficulties which we can experience in separating labels from what they refer to (see below). Try counting the number of green words, for instance.

red green blue green red yellow blue
yellow red blue yellow green red blue
blue yellow yellow blue red blue yellow
red green green red green green green
green blue blue yellow yellow yellow
yellow red green yellow blue green red
blue green red red green red green blue
red yellow yellow red blue yellow blue
yellow blue red blue green green yellow
green red yellow blue yellow blue red
blue red blue green red yellow blue
green green red yellow blue yellow blue

A well-known study demonstrating the influence of language on recall is that of Elizabeth Loftus (1974 & 1979). She showed observers a short film of a traffic accident and asked how fast the cars had been going. However, the wording of the question differed between the two groups asked. Those in one group were asked 'About how fast were the cars going when they hit each other?' whilst those in the second group were asked 'About how fast were the cars going when they smashed into each other?' Those who had been asked the question with the term smashed gave higher estimates of the cars' speeds than the others did. A week later, the same observers were asked whether they had seen any broken glass (there had been none). More than twice as many of those questioned with the word 'smashed' reported seeing the nonexistent glass as those questioned with the word 'hit'.

It is not only with respect to categorization thay perception is described as 'selective'. Recovering his sight after 30 years of blindness, one man reported:

Perception is unavoidably selective: we can't see all there is to see. There are of course physiological limits (both for the human species and for individuals); some argue that there are limits to cognitive capacity. And then there are the constraints of our locational viewpoint: we can't see things from every angle at once. But in addition to such physical limits we focus on salient features and ignore details which are irrelevant to our current purposes or interests. Selectivity thus involves omission. Some commentators use the 'filter' metaphor - we 'filter out' data, but this suggests a certain passivity: we may also 'seek out' data of a certain kind.

Selective attention is assisted by redundancy: we don't always need much data in order to recognize something. Often we can manage with minimal visual data, making use of what is called 'redundancy'. You may know those 'blocky' pictures of famous people in which you can just about recognize who it is. Our schemata allow us to 'fill in gaps' because we know what should be there. So selectivity also involves addition.

Selectivity also involves organization: foregrounding, backgrounding and rearranging features. Objects, events or situations are 'sized up' in relation to our frames of reference, and these influence how perception is structured (Newcomb 1952, 88-96).

Gordon Allport and Leo Postman (1945; Newcomb 1952, 88-96) offered a classic account of the selectivity of perception in their study of rumour. Whenever an event is open to divergent interpretations, reporting it involves transforming it. The selection, retention, reporting and retelling of events routinely involves several kinds of transformations. All of these involve simplifying events to make them more meaningful in terms of personal interests, needs and experience. The process is exaggerated where memory and retelling are involved, but it is already at work in the selectivity involved in the initial perception of an event.

Contrary to the popular idea that rumours ‘snowball’, becoming more elaborate in the telling, psychological studies suggest that retelling tends to make accounts shorter, more concise, more easily grasped and told. There is an increasing tendency to use fewer words. Levelling is the selective process by which certain details are omitted. However, items of particular interest to the reporters, which confirm their expectations or help to structure their reports, do tend to persist.

Sharpening is the reciprocal selective process of levelling. Alongside the loss of some details, there also tends to be a pointing-up of a limited number of details which caught the individual’s attention, often including attention-grabbing words. Temporal sharpening involves a tendency to describe events in the present tense. Movement is often emphasized or introduced. Items prominent because of their relative size or quantity tend to be retained. Labels tend to persist. Primacy effects may lead to the retention of items coming first in a series. Familiar symbols are also likely to be retained. Explanations may be introduced, especially to produce ‘closure’.

Underlying the selective processes of levelling and sharpening, and of transpositions, importations and other transformations involved in retellings, Allport and Postman (1945) argue, is the process of assimilation. This involves the influence of habits, interests and sentiments on reporters and listeners. Aspects of a story are sharpened or levelled to make them more consistent with what is seen as the principal theme of the story, thus making the story more coherent and ‘well-rounded’. Items relevant to the theme may be imported and those irrelevant to the theme may be omitted. Apparent ‘gaps’ may be filled. And some details may be changed to make them more consistent.

Selective perception is based on what seems to 'stand out'. Much of this 'standing out' is related to our purposes, interests, expectations, past experiences and the current demands of the situation. However, some seems more widely-shared - throughout a culture or even across the human species. For instance, we seem to have a general preference for features which are large and/or bright and/or moving, for the novel, the surprising and the incongruous, and for what is meaningfully complex ('looking like things'), and our fixation tends to be on discontinuities, corners and contours. We will turn to such apparently universal features of human perception in discussing Gestalt theories.

Some aspects of perception can be usefully discussed in terms of 'selectivity', but to see perception purely in terms of selectivity would be reductive. It would court the danger of implying that perception is relatively passive and would downplay the active construction of reality.

This is related to what psychologists refer to as perceptual constancy. Our perception of objects is far more constant or stable than our retinal images. Retinal images change with the movement of the eyes, the head and our position, together with changing light. If we relied only on retinal images for visual perception we would always be conscious of people growing physically bigger when they came closer, objects changing their shapes whenever we moved, and colours changing with every shift in lighting conditions. Counteracting the chaos of constant change in retinal images, the visual properties of objects tend to remain constant in consciousness. We are not usually conscious of people appearing to get bigger as they approach us or of things appearing to change shape according to the angles from which we view them. In relation to visual perception, key 'constancies' are: size, shape, lightness and colour.

The following illustration demonstrates how a door appears to change shape as it is opened. Shape constancy ensures that we are not typically conscious of this.

With regard to shape constancy, R H Thouless published a paper entitled 'Phenomenal regression to the "real" object' in the British Journal of Psychology in 1931. He reported an experiment in which he exposed a circular disc at various angles and asked observers to judge its shape each time. The observers did so by selecting a matching disc from a series of circular and elliptical ones which they had been given. When the disc was directly in front of them and in a vertical plane the judgement was easy, of course. But when Thouless rotated the disc away from the observer so that it appeared elliptical, the task was more difficult. Judgements of shape reflected a compromise between the shape as displayed at an angle (an ellipse) and the actual shape of the object (the circle). Observers did not see the shape as it would be on the retina but instead exhibited a 'phenomenal regression' - the phenomenal or apparent shape was inbetween the tilted shape and the vertical shape. This has been called a 'perceptual compromise'.

Familiarity of shape is also an explanation of the illusion generated by a special 'room' called 'the Ames Room' (students in Wales: note that there is an Ames Room at Techniquest in Cardiff). Observers peer through a single hole in a wall of this structure (thus having only monocular cues to depth, as in looking at a photograph or painting, rather than a space around which one could move). Two people of similar size within this special room would look very different in size to observers as indicated in the following illustration.

The reason for this strange illusion is to do with the extraordinary construction of the room, the lefthand wall of which actually goes far further back than the righthand wall, for instance. We are so used to rooms being rectangular that we intepret everything within it on this assumption. The Ames Room represents the mind making a habitual bet, and only getting it wrong because of a careful conspiracy.


home :: admin :: lectures :: reading :: assignments :: modules :: tutor