Broadbent, Donald Eric
BROADBENT, DONALD ERIC
(b.Birmingham, United Kingdom, 6 May 1926; d. Oxford, United Kingdom, 10 April 1993)
experimental psychology, attention and perception, decision making.
Broadbent, often considered one of the founders of cognitive psychology, was best known for his experimental and theoretical work on attention and short-term memory. His “filter theory” of attention accounted for a wide range of phenomena, particularly in the auditory domain, and served to reawaken interest in the relation between attention and perception. His most influential work, Perception and Communication (1958), also served as a model of scientific method for the new orientation in psychology. It presented what may have been the first “flow chart” of a cognitive model. Among his many honors and awards, Broadbent was elected as a Fellow of the Royal Society in 1968 and a Foreign Associate of the U.S. National Academy of Sciences in 1971. He received the Distinguished Scientific Contribution Award from the American Psychological Association in 1975.
Origins and Education. Broadbent’s father was a business executive whose success led to a brief period of family affluence that ended when his father lost his job and disappeared shortly before World War II. While Broadbent grew up in Wales (and always considered himself to be more Welsh than English), his mother insured that he was educated at Winchester, an exclusive English public (that is, private) school. Lack of funds, and the privations of wartime restrictions and rationing, made his school years grim in many respects, a factor he credited, perhaps whimsically, for his “dour and puritan” attitude toward psychology. A very private person, friends in fact described a “puritanical streak” in Broadbent, mixed with exceptional generosity toward friends and students and a deeply moral attitude toward public service. Pressured at school toward the humanities, he felt more affinity for the sciences, and his interest in “real world” problems led him toward the social sciences.
In 1944, Broadbent joined the Royal Air Force (RAF) and applied for pilot training. After completing a short course in engineering at Cambridge University, he was then sent to Florida, where RAF pilots were given flight training far from predatory Luftwaffe fighters. In the United States, Broadbent first became aware of psychology in the context of noticing the importance of psychological problems in practice:
The AT6, which I was flying, carried two identical levers close together under the seat, one of which pulled up the flaps.... The other lever pulled up the wheels. With monotonous regularity, one or another of my colleagues would pull the wrong lever, drop an expensive airplane onto its belly ... , and after a harrowing interview with our superior officer, disappear to England.... The technology was fine, but it seemed to be badly matched to human beings.” (Broadbent, 1980 pp. 43–44)
Since psychology was widely taught in the United States (in contrast to the United Kingdom, where only a few programs were then in existence), Broadbent came to realize that psychology was a possible career choice, one that could combine his scientific affinities with social concerns. The end of the war meant a hasty return to England, and Broadbent, intrigued by the down-to-earth character and potential applications of psychological testing procedures, opted for a ground job with the RAF’s personnel selection branch. Following the end of his military service in 1947, Broadbent applied to Cambridge University for undergraduate training in psychology where, as it turned out, Sir Frederic Bartlett was head of the Psychology Department. Bartlett was well-known for his work on constructive processes in perception and for the view that memory was reconstructive; both aspects heavily influenced Broadbent. In addition, the young Kenneth J. W. Craik, a Scottish experimental psychologist familiar with sophisticated engineering approaches, had been working closely with Bartlett. In spite of Craik’s premature death in 1945, his influence on Bartlett’s lab was immense, and he left behind a number of incomplete reports that sketched an agenda for research using a cybernetic approach to human behavior. Broadbent thus became part of an enthusiastic group of psychologists involved in extending such an approach. Beginning during Broadbent’s school years, the advent of information theory, and subsequently the striking finding by William Edmunc Hick in 1952 that reaction time varied linearly with the amount of information in bits led the Bartlett-Craik approach in the direction of what would later be called cognitive psychology. Broadbent was to be a major contributor to this development.
Career. Following graduation in 1949, Broadbent worked for the Royal Navy on the psychological effects of ambient noise, a project administered by the Medical Research Council. Broadbent carried out this research at Cambridge as a member of the Applied Psychology Unit (APU), where he remained until 1974. The APU, founded in 1944 as a result of Bartlett’s efforts and initially led by Craik, provided advice and carried out research for government and military offices. When Craik died, Bartlett took over as director, assisted by Norman Mackworth, who became director in 1952, following Bartlett’s retirement. At this point (and with a staff of more than twenty), the APU moved from the Department of Psychology to its own quarters away from campus.
Bartlett’s research on noise led easily to an accommodation with the “vigilance” research of Mackworth and to Broadbent’s focus upon the problems of attention, then a neglected topic in psychology. Concentrating during the 1950s on his laboratory work, which was closely tied to applied concerns, Broadbent’s thinking increasingly involved problems of communication. New technology, especially in the form of tape recorders with separate channels for each ear, made possible experiments on the problems involved when one individual is receiving messages from several sources at one time. Watching flight controllers, studying aircraft landings on carriers at sea, conducting a communication experiment on the pilot while in-flight—such experiences led Broadbent to remark, “I could see human feats of perception, decision, and control that were clearly highly admirable, in which error might mean death, and yet that lay outside the view of human beings normally put forward by academics” (Broadbent, “Donald E. Broadbent,” p. 55).
During these years, Broadbent enlarged his contacts with individuals in other disciplines, notably Colin Cherry of Imperial College and Peter Ladefoged of the University of Edinburgh. Exposure to the work of psychologists and others in the United States was slow and hampered by the expense of travel, but Broadbent gradually came to know of the work carried out at such nontraditional centers as the Harvard Psychoacoustic Laboratory and the Aviation Psychology Laboratory at Ohio State University, where, as at the APU, scientists from various disciplines worked on common problems. For Broadbent, such contacts confirmed the value of multidisciplinary approaches and revealed the limitations of what he saw as a frequently insular academic psychology.
By the mid-1950s, Broadbent was beginning to feel constrained by the narrow confines of the usual technical journal. To range more widely, he wrote Perception and Communication (1958), a book that remained his major achievement and that enhanced his already prominent reputation. In the same year, Mackworth left for a new position in the United States, and Broadbent became the director of the APU, a position he held until 1974. His work as director increasingly carried administrative duties, although he remained active as an experimenter and writer, publishing, besides many articles, the semi-popular work Behaviour in 1961; a major restatement and enlargement of his theories, Decision and Stress (1971); and a collection of lectures, In Defence of Empirical Psychology(1973). Moving in 1974 to Oxford University as a professor of experimental psychology provided more time for his own research. Retiring in 1991, Broadbent remained active utill his sudden death from a stroke in 1993.
Research and Theoretical Contributions. Broadbent’s “filter theory” of attention and perception, which receives notice in nearly every textbook of cognitive psychology, began as an attempt to understand certain curious results in his communication research. When multiple messages arrive at one time, how can a listener separate out the meaningful ones from those that are nonmeaningful? Broadbent’s research on the problem, which began with studies of the effects of noise and of masking on speech (for example, his “The Role of Auditory Localization” in 1954), soon made clear that the difficulties of perception were not simply peripheral but were central as well. For example, ship operators sometimes appeared to be overloaded with information rather than hindered by purely sensory factors. Using the new technology of tape recording, separate messages could be delivered to each ear of a listener. Most ways of separating the messages spatially, even presenting stereo recordings with a message in each of the two channels, helped performance. Broadbent concluded that a selective mechanism was at work to reduce the information flow through a limited capacity perceiver
In a 1957 paper, “A Mechanical Model for Human Attention and Immediate Memory,” he analogized the selective mechanism to a Y-shaped tube with a hinge at the junction of three pipes. If incoming information is represented as a flow of balls being dropped into one of the tubes, then the first ball can push aside the hinge and reach the bottom pipe. The subsequent flow of balls in one pipe leaves the hinge pushed aside so that balls from the other pipe of the mechanism cannot pass the junction. Similarly, human attention to a selected message seemed to block out an unselected message. The story is too simple in this form, however, because it leaves out important aspects of the human mechanism, namely, the fact that the mechanism (unlike the simple hinge) can switch from one channel to another, and also the fact that over time the balls in the unattended channel would be lost. Broad-bent’s initial model identified the switching with a central attentional process keyed to spatial location, and the loss (and the implied retention over short time intervals) to a memory mechanism, but one with very limited capacity. Invoking attention was a radical step, since the concept had disappeared almost entirely from experimental psychology for four decades. As Broadbent put it,
The reason for its disappearance had been the lack of suitable terminology: common-sense language has great difficulty in discussing the problems of the air-traffic controller because, in one sense, he does more than one thing at the same time, and in another sense, his capacities are limited. But the new notions of information theory provided a suitable vocabulary. Instead of thinking of a “stimulus” in the outside world, which did or did not produce a response, one could speak of an event that produced some representation in the person. (1980 pp. 55–56)
Experimental studies by Broadbent and others showed that material that had gotten through the hinge initially could be recycled back to the original arm of the Y; it could be “rehearsed” and kept from being lost. In this respect, Broadbent’s filter theory of attention requires a memory mechanism, a short-term store. In Perception and Communication, Broadbent dropped the mechanical analogy (which he had not represented as anything more than a mnemonic device in the first place) and emphasized instead the analogy between the flow of information and the function of attention and memory. Because information theory provided a metric for measuring the amount of information in “bits” (the logarithm of the number of alternatives, weighted by the probability of each alternative), quantitative studies became possible.
Broadbent’s final representation of his filter model used a flow chart (Figure 1), that is, a figure showing “boxes” connected by arrows to indicate the separate processing stages and the flow of information from one stage to another. Note that the figure provides a short-term store preceding the selective filter mechanism, including also a feedback loop such that information could be recycled through the store (“rehearsed”) as needed. Information in the short-term store could be held only for a short time (hence the need for a rehearsal mechanism). In addition, it appeared to have another unusual property, namely, it did not seem to have a limited capacity that could be defined in terms of information theory as such. It was not the number of bits of information in the technical sense that mattered. Rather, as Irwin Pollack, George Miller, and others had shown, its capacity was limited by
the number of items in store—and each of the items could be a “packet” of very high information value.
In the late 1950s, experiments carried out by Anne Treisman at Oxford provoked Broadbent and Margaret Gregory (then his research assistant and later his second wife) to carry out a series of experiments that greatly modified the original filter theory. Treisman had found that in fact some information could get through an unattended channel. Thus, if a subject were asked to “shadow” (that is, repeat back out loud) words in the left ear channel while ignoring words in the right ear channel, and the shadowed message were then changed from left to right, subjects would often continue shadowing without pause, especially if the words in the unattended channel were a highly probable extension of the original message. Thus, some information, partial and fragmentary though it was, was getting through the unattended channel. Treisman’s “double theory of attention” argued that incoming information was selected from something like a probabilistic “dictionary,” as well as from a particular channel (left or right ear, say). Broadbent’s filter theory was clearly discon-firmed by these results; “Attention does not simply switch off some sense organs and leave others switched on,” he wrote (In Defence of Empirical Psychology, p. 137).
In revising his own account, Broadbent also credited Treisman for an important suggestion, namely, that in prolonged vigilance studies, some of the results could be due to changes in subject confidence about a percept rather than to changes in subjects’ sensitivity to the incoming information. “Margaret and I,” Broadbent recalled, “began to look at confidence ratings and, thus, to think about the theory of signal detection put forward by Spike Tanner and John Swets” (“Donald E. Broadbent,” p.66). Signal detection theory (SDT) was developed to analyze psychophysical changes in the threshold for detecting the presence or absence of a stimulus. By analyzing “hits,” (correct detections, that is, “yes” responses when the stimulus is present or “no” responses when it is absent), “misses” (failures to detect when the stimulus is present), and “false alarms” (“yes” responses when no stimulus is present), the theory allowed changes in observer sensitivity to be separated from changes in the observer’s criterion (or “bias”) in reporting. Across a large number of tasks, Broadbent and Gregory were able to show that changes in error rate in complex tasks could be due to changes in bias, changes in sensitivity, or both. This resolved the apparent paradox presented by Treisman’s results; a listener was not insensitive to information in the unattended channel but was in fact responding with greatly reduced bias toward that channel. When highly probable stimuli occurred in the unattended channel, this meant that the “bias setting” could change, thus changing the apparent “gain” in that channel. In this fashion, Broadbent’s “filter” became instead a semi-permeable one, with two parameter settings, one for bias and one for sensitivity. Selective attention could then be understood as selection of the relevant parameter settings. Furthermore, Broadbent’s new account in effect extended his model of attention and memory into a decision-making context. This further enhanced its ability to account for the effects of stress, noise, and the like on performance.
The new account required Broadbent to reassess completely all of the areas of research with which he had been associated. His Decision and Stress(1971) summarized the changes and widened the scope of his theory to include decision making, vigilance, and the choice of actions in voluntary behavior. In addition to SDT, Broad-bent, like many others in the field of judgment and decision making, made extensive use of Bayesian formulations to understand how limited capacity organisms could function in probabilistic and uncertain environments. In 1973, he collected a series of lectures under the title In Defence of Empirical Psychology, which presented a simpler version of the new theory. This and Decision and Stress, however, enjoyed much less success than the 1958 book, Perception and Communication.
Views on Psychology and Science. Throughout his career, Broadbent’s thinking centered on three principles of research, each of which he attributed to Bartlett’s influence: (1) research problems should derive from real life whenever possible; (2) experimental situations should remain complex rather than be oversimplified; and (3) data should determine theory rather than the other way around. Ironically, he did not accept what has become Bartlett’s most famous principle, that memory was organized in “schemas,” organized complexes of stored information and templates for the generation and regeneration of specific recall, although Broadbent, like Bartlett, did rely heavily upon the overall notion of a dynamic perception and memory system.
In all of his books and many of his articles, Broadbent stressed his criticism of the hypothetico-deductive method, which he informally characterized as the “guess and test” approach. Drawing many of his illustrations from the work of Clark Hull, Broadbent felt that it was an ineffective scientific strategy to develop a formal theory (the “guess”) and then assess it by drawing specific predictions and testing to see if the predictions were confirmed. Hull’s system failed, in Broadbent’s view, because of its excessive use of a priori theorems of great generality but little specificity; so much by way of ad hoc assumptions and theorems had to be added to the theory to make a specific prediction that the failure of a prediction told one essentially nothing about where the problem might lie; was it the theory itself or the ad hoc assumptions? And if the latter, which ones? Instead, he argued, science should begin with more narrowly construed theories based upon reliably observed phenomena. These should then be subject to as many alternative explanations as possible; experiments could then be designed that would rule out specific formulations. In this fashion, the theoretical side of psychology would not outrun the empirical side; “The proper road for psychology is by way of more modest theoriz-ing,” wrote Broadbent (Perception and Communication, p. 313).
In Perception and Communication, Broadbent illustrated his approach by beginning the book with extensive accounts of experiments, arranged in a compelling sequence of tests, alternative explanations, refined explanations, further tests, and so on. Only as these provided a base of reliable understandings of specific results did he seek extensions toward more general accounts. Thus, the famous flow chart that summarized his filter theory appears at the end of the book, on page 299, not at the beginning, as a theory to be “tested.” In fact, read in the early 2000s, Perception and Communication remained a sterling example of an inductivist approach to experimental psychology
Identified as one of the founders of cognitive psychology, Broadbent was never comfortable with all of the aspects of that designation. Giving full due to behavioral approaches (especially those of Ivan Pavlov and of Hull), Broadbent’s invocation of unobservables was, he claimed, a natural development of those earlier systems, not a revolutionary new ideal. The unobservables he invoked were tied closely to observables, and he never lost the belief that behavior constituted an essential starting point for all of psychology and that the explanation of behavior was the ultimate goal of all of psychology. In critiquing a distributed memory model, for example, Broadbent argued in his paper, “A Question of Levels,” that James L. McClel-land and David E. Rumelhart’s connectionist model had implications for a physiological level of explanation but not for a psychological level; for Broadbent, a neural network model is simply too far removed from the relevant behavioral data.
Broadbent’s , Influence. By bringing the problem of attention back into experimental psychology, Broadbent significantly changed the direction of that field, opening one of its most fruitful lines of inquiry in the cognitive era. Furthermore, his tireless efforts to enhance public perception of the new psychological results, and to extend their applications to real world problems, had a lasting influence on psychology, particularly in Britain. He was an important advocate and contributor at a time when British psychology moved from a marginal academic specialty to a thriving enterprise with connections to many aspects of British industry, government, academia, and public discourse. In contrast to American psychology, British psychology had never fully embraced behaviorism, retaining instead close ties to biological traditions of thought, especially those deriving from Darwinian influences. In this context, Broadbent extended the unique efforts of Bartlett and Craik to establish a purely psychological science, one tied closely to experiment and to real-world applications.
L. Weiskrantz, “Donald Eric Broadbent,” Biographical Memoirs of the Fellows of the Royal Society 40 (1994): 31–42, contains a microfiche with a complete bibliography of Broadbent’s published works.
WORKS BY BROADBENT
“The Role of Auditory Localization in Attention and Memory Span.” Journal of Experimental Psychology 47 (1954): 191–196.
“A Mechanical Model for Human Attention and Immediate Memory.” Psychological Review 64 (1957): 205–215. Perception and Communication. New York: Pergamon Press, 1958. His most influential and important work, a sophisticated example of the power of experimentation, as well as the basic exposition of his “filter theory.”
Behaviour. New York: Basic Books, 1961.
“Attention and the Perception of Speech.” Scientific American206 (April 1962): 143–149. A good introduction for the general reader.
Decision and Stress. London; New York: Academic Press, 1971. The extensive revision of his 1958 theory along decision-theoretic lines.
“Levels, Hierarchies, and the Locus of Control.” QuarterlyJournal of Experimental Psychology 29 (1977): 181–201.
“Donald E. Broadbent.” In History of Psychology inAutobiography, edited by Gardner Lindzey. Vol. 7. San Francisco: W. H. Freeman, 1980. Warm, witty, and insightful, the best source for an appreciation of Broadbent as a person, and for the major themes in his work.
“A Question of Levels: Comment on McClelland and Rumelhart.” Journal of Experimental Psychology: General114 (1985): 189–192. Preceded by a paper of McClelland and Rumelhart and followed by their rejoinder.
“Effective Decisions and Their Verbal Justification.”Philosophical Transactions of the Royal Society of London, series B—Biological Sciences, 327 (1990): 493–502. Broadbent’s critique of the usefulness of verbal protocols in psychology.
Baddeley, Alan, and Lawrence Weiskrantz, eds. Attention:Selection, Awareness, and Control, A Tribute to Donald Broadbent. Oxford; New York: Clarendon Press, 1993. Published just after Broadbent’s sudden death, this edited collection of original papers by many of his students and colleagues contains much information about his influence and the direction of his ideas in the late twentieth century.
Berry, D. “Donald Broadbent.” The Psychologist 15 (2002): 402–405. Also available from http://www.bps.org.uk A warm remembrance by a former student, with an account of his work on implicit learning.
Lachter, J., K. I. Forster, and E. Ruthruff. “Forty-Five Years after Broadbent: Still No Identification without Attention.” Psychological Review 111 (2004): 880–913. A revival of the 1958 filter model, with experiments suggesting that the original model remained valid in some circumstances.
Weiskrantz, L. “Donald Eric Broadbent.” In BiographicalMemoirs of the Fellows of the Royal Society40 (1994): 31–42. The fullest of several obituaries.
Ryan D. Tweney