The nervous systems of living creatures are subjected to far more stimulation than can be used. On the one hand, the sense organs receive stimuli of great variety. On the other, memories, images, and ideas arise internally and must be considered from moment to moment. Yet it is a commonplace that we are consciously aware of only a limited amount of this information at any moment. The operation by which a person selects information in attention, and its study, have twice been seen as central to research on our understanding of how information is processed by humans and animals.
The first great period was around 1900, when
James,
Titchener,
Wundt, and W. B. Pillsbury all wrote on attention at length. William James, on this as on so many other topics, described the main characteristics of attention with precision. Attention was, for him, 'the taking possession by the mind, in clear and vivid form, of one of what seem simultaneously possible objects or trains of thought'. Titchener and his students, in particular, carried out an extensive experimental programme into such topics as 'prior entry' and the conditions of binocular rivalry and fluctuations of attention. The first of these was the observation that, of two simultaneous events, the one to which attention was diverted appeared to occur earlier than the other.
The fundamental property of attention was, for those writers, to make the contents of
consciousness appear clearer. It would have made little sense to them to discuss effects of attention of which the observer was not aware. Pillsbury's book
Attention, published in 1908, contains many observations which modern work has confirmed, and is a remarkably insightful volume in many respects.
With the rise of
behaviourism, attention was relegated to the status of a mental function which could not be admitted as a suitable object for research. Indeed, for some 30 years it disappeared from indexes and reviews. The second golden age of attention research dates from the early 1950s, and received a particular impetus with the publication of
Perception and Communication by
Donald Broadbent in 1958. A major reason for the renewed interest was the need for the solution of new practical problems, such as the design of control towers and communication networks in the Second World War. A controller might receive several messages at once from different aircraft or ships, and be required to make appropriate responses to each. With an attempt to understand how humans behaved in such situations, modern work on attention began. It was aided by the invention of the tape recorder, which for the first time allowed the ready control and replicability of speech signals, while the phenomenon of stereophony provided an easy way of varying the content and amount of information in competing messages.
Broadbent's filter theory tried to explain how the brain coped with the information overload caused by having many sense organs receiving information simultaneously. Drawing an analogy with electronic communication, he proposed that there is in the brain a single central information channel whose rate of information processing is rather limited. This channel could select only one sensory input channel at a time, and could switch no more than about twice a second between input channels. To accept an input was equivalent to paying attention to that source of information, and information on unattended channels could be held in a
short-term memory for a few seconds. Broadbent called the selection mechanism 'the Filter'. While he drew on many fields of research, the most direct line of evidence was the 'split span' experiment. If three digits are read, at a rate of two per second, to the left ear of a listener, and another three to the right ear, so that the listener receives three synchronous pairs, he will recall them ear by ear, not pair by pair. Broadbent interpreted this to mean that the listener attended to one ear first and then switched to the memory trace of material in the other ear. By finding the fastest rate at which the listener could repeat the message as pairs, he believed he had measured the rate of switching of auditory attention. This concept of a single-channel, limited-capacity information-processing system was central to much research in the next 25 years.
From about 1953 to 1963 'speech shadowing' was widely used by Broadbent, Neville Moray, Anne Treisman, and others. This technique had been introduced by Colin Cherry, and required a listener to repeat aloud a prose message in the presence of one or more distracting messages. It was found that major factors which aided selective attention included separation in space of the speakers, difference in voice timbre, and the statistical structure of the messages. In a series of elegant experiments Anne Treisman greatly extended our knowledge. Certain features of a distracting message proved to be potent sources of distraction, including emotional words (such as a listener's own name), contextually probable words, and — for bilingual listeners — the presence of a translation of the message to which they were listening.
However, these experiments also showed that some material, such as emotionally important words and contextually probable words, was perceived even when in the 'rejected' message. This led to a series of modifications to Broadbent's filter theory, notably by Treisman, Anthony Deutsch, and Don Norman. Although differing considerably in detail, they all attempted to account for the fact that the filter apparently did not block all information from the rejected channels, and that selection could be not only of sensory inputs, but of such features as language, class of word, colour (in the case of visual stimuli), and even classes of responses. Attention came to be seen as acting in a variety of ways, at a variety of levels, and on a variety of operations in the nervous system.
By the mid-1960s interest had grown greatly, and a wide variety of experimental techniques was developed. In addition to speech shadowing, simultaneous auditory messages were used, requiring much simpler responses than speech. In a series of studies Moray showed that, contrary to what the early shadowing experiments seemed to indicate, attention acted in the same way on non-linguistic as on linguistic material. Robert Sorkin and his co-workers in America explored attention to non-linguistic auditory material. Others, like Alan Allport and Peter McLeod in Britain, and Richard Shiffrin and Walter Schneider in America, found conditions where little or no interference between two messages occurred. These arose especially when messages were presented in different sensory modalities, where long practice had made performance almost automatic rather than conscious and voluntary, and when no competition between responses was required.
Visual attention was investigated by means of eye movements in such tasks as reading, by Paul Kolers, and in car driving and piloting aircraft, by John Senders. Very spectacular results were found using the 'Stroop test', in which attentional and perceptual conflict is induced by the nature of the stimulus. If the word 'red' is written in green ink, the word 'blue' in yellow ink, and so on, it is possible to read the word rapidly without the colour of the ink causing interference, but almost impossible to name the colour of the ink. Analogues of this effect provide a way of discovering which 'analysers' in the brain can be selectively biased by voluntary attention. (The word 'analyser' originated in a theory of learning and perception due to Deutsch, but its use in attention theory is largely due to Treisman.)
Some attempts were made to investigate the physiological mechanisms underlying attention. For example, the 'Expectancy wave', or CNV, was discovered in
electroencephalographic records, a change in electrical brain activity which appears when the observer is concentrating on the imminent arrival of a signal he knows is probable. In the late 1970s Emmanuel Donchin and his colleagues began work on the 'P-300' component of the brain-evoked potential, and this seems very likely to be intimately connected with attention in the sense of decision making. But to date our understanding of the physiological basis of attention lags a long way behind behavioural research.
More than a dozen theories of attention have been proposed since 1958, most of them strongly influenced by communication theory and computer technology. Their variety to some extent is due to the variety of phenomena which may be subsumed under the heading 'attention'. In addition to our ability to listen selectively to one message and ignore another, or to look at a picture in one colour in the presence of other colours, one may cite
vigilance (or watch keeping), in which an observer looks for very rare events, such as detecting the presence of a sonar or radar target. Some studies have been made of mental concentration on cognitive problems. The interference between internal images and incoming stimuli has been investigated, as has the ability of the brain of a sleeping person to respond selectively to the sleeper's name, even though the sleeper is not aware of the response.
Although no single theory has emerged as completely dominant, the influence of Broadbent's filter theory remains strong, and what follows is a conflation of theories based on his suggestions. He assumed that an observer can block or weaken the strength of incoming messages to the brain, and there is ample evidence that this can happen. It is not known whether this is done by reducing the intensity of the messages or by switching them on and off rapidly. But some such blocking definitely occurs. In vision it can be done by closing the eyes or averting the gaze. In hearing the mechanism is not so clear. It seems probable that all information which impinges on the receptors of the sense organs reaches the pattern-analysing mechanisms of the brain. The filter perhaps acts to prevent the output of these analysers from reaching consciousness, although behaviour may still be produced, as when we become aware that we have driven for some time 'without being aware of it' (see
time-gap experience). It seems likely that information from different sense modalities, or from different dimensions within a modality (such as colour and shape), can be attended to simultaneously, at least after practice, while tasks which are very similar (such as judging the loudness of two tones) cannot. One should note that in making these assertions we are far from James's definition. Very often the observer is conscious of only one message, but it can be shown that the second is producing behaviour simultaneously with that produced consciously. (See also
subliminal perception.)
A second way in which attention can operate is by biasing the interpretation of information proceeding from pattern analysis to consciousness. Thus a person expecting to see a bull in mist will see one, while a person expecting to see a rock will see a rock. This kind of bias is set by the probability of events, their subjectively perceived value, and contextual information derived from recent inputs and from memories. In earlier days this kind of bias was called 'mental set'.
The contents of consciousness as filtered by attention are very limited: attention is frequently modelled as a 'limited-capacity information channel'. But with practice, quite dramatic increases in performance are seen, and some writers, among whom Daniel Kahneman is particularly influential, have proposed a 'parallel processing' model of attention, in which the main limit is on the total effort available, rather than on competition between separate analysers. Such models make extensive use of the concept of
arousal.
Recently, renewed interest in applications of attention has become apparent. Most of the laboratory research has been directed to understanding the internal mechanisms of attention in the brain. But as large and complex man-machine systems appear, and more and more automation is introduced, there is a tremendous need for a good understanding of man as a monitor of complex systems. How should they be designed so as to optimize the use of attention? (It is fairly clear that in real-life tasks attention is never switched more than about twice a second.) How should a man be trained so as to combine to best effect his limited conscious attention with his unconscious control of
skilled behaviour? The solution of such questions is necessary if accidents in power stations, aircraft, and industry are to be avoided. Attention theory has advanced to a point where it can give a real insight into the solution of practical problems, and there is likely to be a third age in which the precise experimental work of the 1950s to the 1970s is extended in more complex ways to solve the problems of man-machine system design. We have come to see attention not merely as a single process concerned with enhancing the clarity of perception. Rather, it is a complex of skills. These include selecting one from several messages, selecting one from several interpretations of information, selecting one from several plans of action, and selecting one from several actions. But in our interactions with the rich dynamics of the world in which we live and to which we adapt, attention also models that world, and provides us with strategies and high-level control of our tactics of information sampling, optimizing our information processing in the face of our limited processing capacities.
(Published 1987)— Neville Moray
Bibliography- Broadbent, D. (1958). Perception and Communication.
- James, W. (1890). Principles of Psychology.
- Kahneman, D. (1973). Attention and Effort.
- Kanwisher, N., and Wojciulik, E. (2000). 'Visual attention: insights from neuroimaging'. Nature Reviews Neuroscience, 1.
- Moray, N. (1969). Listening and Attention.
- Titchener, E. B. (1908). Lectures on the Elementary Psychology of Feeling and Attention.
- Underwood, G. (1978). Strategies of Information Processing.