MRC Cognition and Brain Sciences Unit

You are in:  Home » History of the Unit

1. SUMMARY (Directors Overview)

General Notes. This material has been scanned from the original typescript. While we have done our best to remove errors, some may well remain. You can access other parts of this particular Progress Report either from the menu at the bottom of this entry, by using your browser's back function, by navigating back to the Unit History Timeline, or by accessing the relevant section of the electronic archive. Reference for this report are indexed by number and these can be found in a dedicated section also accessible from the menu at the bottom of this entry.

I would like to begin my last Progress Report with a brief explanation as to why I have chosen to relinquish the Directorship of the APU several years ahead of retirement. There are two major reasons. At a personal level, I find that I am continuing to enjoy research just as much as ever; in the normal course of events I would retire in five years time, and following the very sensible MRC policy of discouraging ageing ex-Directors from continuing to haunt their old Units, would find myself looking for a new home as a pensioner. I much prefer the prospect of moving now, in order to initiate what I hope will be an extended programme of research in a new environment.

My second reason is that I believe that such a strategy would be better for the APU. For the past year we have been engaged in drafting the Unit's Progress Report, and it is likely to be a year from now before the Strategy Committee makes its decision about our work. That would leave four years to my retirement, and the normal course of events would be for us to produce another Report two years before my departure, which would then be used as a basis for deciding the future of the Unit and the selection of my successor. By the time he or she arrived, the Unit would have been in a state of greater or lesser disruption for a total of six years, with a restriction on appointing tenured staff operating over at least half of that period. While I am sure that the Council would be as flexible as possible, it is difficult to see how a prolonged period of uncertainty could be avoided, with its inevitably negative effect on the Unit's morale.

After discussions with the Office and my colleagues at the Unit, I therefore decided that I should retire, thereby allowing the current evaluation of the Unit's work to be combined with the decision as to the Unit's future, and I trust, the consequent search for a new Director.

The Remit of the APU

This Progress Report coincides not only with my own retirement from the Directorship after 20 years, but also with the Unit's 50th anniversary. I hope I can therefore be forgiven for taking this opportunity to look at the Unit's remit from a rather broader perspective than would normally be justified.

Throughout its first 50 years, the Unit has consistently operated at the interface between psychological theory and practical questions, basing general models of human cognition on specific empirical evidence, from both naturalistic and laboratory-based studies, and using this evidence in the attempt to solve applied problems. The Unit's first Director, Kenneth Craik, was perhaps the first person to formulate clearly, and test empirically the idea of the computer as a model of complex human behaviour. He was tragically killed in a cycling accident one year after becoming Director, and was succeeded by Sir Frederick Bartlett who was also Professor of Psychology. Bartlett had a great talent for combining careful experimentation with the development of theory that was applicable to naturalistic as well as experimental data. As such he was a precursor of the subsequent so-called cognitive revolution of the 1960s, and his reputation has continued to be high both in Europe and North America. When he retired he was succeeded by Norman Mackworth, probably the least theoretically-oriented of the Unit's directors, but whose influence on the Unit continues to be strong, through the non-hierarchical management style he favoured, and not least because of his purchase of No. 15 Chaucer Road, the Unit's home since 1951.

When Mackworth later emigrated to Canada, he was succeeded in 1958 by Donald Broadbent, who shortly afterwards published Perception and Communication, a book that provided the first coherent theoretical account of cognition in terms of the information-processing metaphor previously developed by Craik. It was the forerunner of the revolution whereby cognitive psychology gradually became the dominant theoretical paradigm within psychology. While not supplanting other approaches, cognitive psychology has provided an information processing theoretical framework that allows integration across a wide range of areas, coupled with the potential for developing precise models in more closely specified areas. Broadbent's own work over the succeeding years illustrated this, as does the following Progress Report which represents a blend of approaches reflecting influences as diverse as neurophysiology, animal learning theory, clinical psychology and computer science. During Broadbent's 16 years as Director, he developed and consolidated the Unit's position as an international centre for research in pure and applied cognitive psychology.

During my own period as Director I have tried to continue this tradition. This has involved maintaining core strength in the major areas of cognitive psychology, and at the same time seeking areas of application that are tractable and fundable. Most importantly, such areas should be capable of feeding back results that will challenge and enrich the underlying theory. Psychology has been fortunate over the life of the APU in that it has always been possible to find such areas of application, although it has been necessary to change direction periodically in order to take full advantage of new opportunities.

The Unit was founded in 1944 on the basis of research projects stimulated by the Second World War, and for many years its work continued to be closely related to such military problems as pilot fatigue and the effect of environmental stress. Defence-related research was later largely replaced by problems stimulated by other Government departments such as the Post Office and Ministry of Transport. With the increasing importance of computers, a larger part of our work began to be carried out in association with industry, though sadly we have usually found it easier to establish research links with US, rather than British-based companies. In the present economic climate, however, even North American companies are rarely able to fund anything other than short-term research.

Fortunately, it has in recent years been increasingly clear that health-related problems present an ideal opportunity for applying and enriching cognitive psychology. Our initial ventures were into neuropsychology, where the careful study of single cases with relatively pure cognitive deficits caused by brain injury proved to be extremely fruitful. The field has developed to a point at which our concepts are now able to deal with a much wider range of cases whose cognitive deficits are less specific, and this in turn is stimulating the development of the new and related field of cognitive neuropsychiatry, whereby cognitive psychology is applied to the analysis and understanding of psychiatric disease.

Some 10 years ago, we decided to try to extend the range of cognitive psychology to include studies of the interaction of cognition and emotion, and were able to attract Watts from a senior clinical post at Kings College London. He was later joined by Teasdale from Oxford, and Williams from Newcastle, and in collaboration with Mathews and Macleod from St George's they succeeded in developing an approach that applied the concepts and techniques of cognitive psychology to the study of emotion and its disorders. This is now part of a vigorous area of overlap between cognitive psychology and the problems of clinical practice.

Our most recent attempt to blend cognitive psychology and practical problems is reflected in the neuropsychological rehabilitation group. While the group has not yet reached full size, or indeed gained access to its clinically-based accommodation, I am delighted with the progress that it has made in demonstrating the viability of neuro-rehabilitation as a fruitful applied area of the application of cognitive psychology.

One crucial advantage of our development of neurorehabilitation as a research area is the extent to which it is benefiting from the rapid development of clinical neurosciences in Cambridge. I expect us to continue to benefit from our growing collaborative links, in particular taking advantage of planned expansion of neuroimaging in Cambridge, and of the establishment of the IRC in Brain Repair. At present, the work of the IRC is principally concerned with animal models, but once the feasibility of clinical trials develops, then I would anticipate close collaboration.

It is clear that many of my colleagues see considerable potential for collaboration between cognitive psychology and the neurosciences. I hope my successor will be as enthusiastic about that prospect as I am. At the same time, I trust that he or she will bear in mind that the Unit's success has come from its capacity to maintain considerable strength in the core areas of cognitive psychology, providing a solid scientific base that is capable of capitalising on whatever area of application is currently most productive, while allowing sufficient flexibility to seek other applications when the time is ripe.

Although the Unit's general remit has remained much the same, the form and content of its scientific contribution has evolved under each of its successive directors. Looking back over the period since 1974, I would like to think that research by APU staff has not only kept up with the best national and international work in both pure and applied cognition but also has continued to help shape the core agenda for our principal fields of operation.

In the mid seventies, the domains of coverage for theory and experimentation were largely those of traditional information processing psychology. Fractionation of component operation in simple pure and applied tasks was the dominant mode of scientific enquiry guided by locally constrained mathematical models, "box" models or verbally formulated theory. In the mid nineties, fractionation plays a key role but is now embedded in a much more interdisciplinary outlook. Our typical tasks now range from simple to complex; our theories tend to be more integrative; our empirical methods are of far greater relevance to charting performance in clinical as well as normal settings. A substantial proportion of our modelling effort is now routinely implemented, for example, mathematical models of auditory image creation, connectionist models of language or memory systems, or symbolic representations of mental models or of complex task performance. Not only the form of the theory has developed; its content now extends outwards to form interdisciplinary links with the basic cognitive sciences of linguistics, artificial intelligence and computer sciences through to clinical disciplines and neurosciences. In each of these areas it can be argued that researchers at APU have played their part - reference to core research summaries and teaching texts testifies to this. I hope that the tools, techniques and theories now in place will form a substantial and relevant foundation for future advance in all of these areas.

What follows in this introduction is my overview of the major advances during the last five years of work at the APU, and also of plans for future work, organised into six scientific programmes. Creating an environment for these advances has been a stimulating and thoroughly enjoyable challenge, made possible by the MRC Unit system. Long may it flourish!

Programme 1: Perception and Action

The purpose of this programme is to develop models of perception at a level between sensory physiology and cognition - models that enable us to make quantitative predictions about perception and behaviour in circumstances where they are predominantly governed by the external stimulus. The research falls into three broad areas, one concerned with audition, a second with vision, and a third with the control of movement. All three areas are well established at the Unit, and in the last five years we have been able to capitalise on theoretical advances to produce patents which have been developed commercially.

Roy Patterson's Auditory Image Model, described in our last Progress Report, has been substantially developed over the period, with much of the work being achieved through collaborations with European colleagues through an ESPRIT contract, and subsequently with North American and Japanese groups. Considerable software development has led to a user-friendly version of the model that was made generally available on Internet, resulting in transfer to 125 sites within the past eight months. The model has proved to be an effective "front end" for speech perception and performs somewhat better than the leading FFT preprocessor for both phoneme recognition and speaker identification, despite the fact that the full capacity of the model was not realised in these initial tests.

Theoretical exploration of the model has been primarily concerned with explaining the perception of sound quality or timbre. Traditionally, auditory models predict our perceptions from the power spectrum of the stimulus. Patterson's model differs in assuming asymmetry in the auditory filter system, which predicts that ramped tones, in which the sound pressure increases gradually and terminates suddenly, will give rise to a very different auditory image from damped tones which have the reverse temporal pattern, despite the fact that ramped and damped sounds have identical power spectra. As Patterson's model predicts, marked perceptual differences occur, whether the sound sequences are based on pure tones or noise.

One of the earliest practical products of Patterson's psychoacoustic research was a system for designing auditory warnings that are maximally audible without being startling, highly discriminable, and easy to learn. Warnings designed on these principles are currently used in North Sea helicopters and were used in RAF helicopters in the Gulf War. Subsequent work has developed warnings for civil aircraft and fixed-wing military planes, and the same principles have been applied to developing trackside warnings for British Rail workers and an international standard for warnings in hospital operating theatres and intensive care units.

Future work will continue the development of the auditory image model with a more user-friendly interface and a simulation of the physiological characteristics of the cochlea that should extend its application to the analysis of hearing disorders (in collaboration with colleagues in the Cambridge Psychology Department). In a collaboration with Armstrong Laboratories in the U.S.A. the model will be implemented on an analog VLSI chip as part of an international programme to develop a "silicon cochlea". Utilisation of the model as part of an automatic speech processor will continue in collaboration with Armstrong Laboratories and with ATR in Kyoto.
Future theoretical work will concentrate on two synthetic sounds which produce complex tone and noise perceptions that traditional auditory models fail to explain. The stimuli are matched pairs of damped and ramped sounds, and iterated rippled noise. The perceptions are interpreted in terms of the time-interval patterns the sounds produce in the auditory image. Finally, we will continue to provide expertise in the development of auditory warnings, as and when appropriate.

The psychoacoustics programme has recently been further strengthened by the appointment of Carlyon, who was 1994 recipient of the Acoustical Society of America's Bruce Lindsay Award. He will be studying the physical factors that allow a listener in a noisy room to pick out a single sound source from the complex and intermingled signals at the ear.

In the area of vision, Wilkins has extended his earlier work on photosensitive epilepsy to a wider population, resulting in a general theory of visual stress and discomfort. He has shown that certain frequencies of stripe or flicker are particularly likely to lead to visual distortions, reading problems and headaches in non-epileptic subjects, a response he attributes to the hyperexcitability of neurons in the visual cortex. The theory has relevance for lighting, text design and the development of visual display units.

Recent work has concentrated on the observation that ophthalmically-tinted spectacles can reduce symptoms in susceptible individuals. The optimal tint varies from one subject to another, and so Wilkins has developed an instrument (the Intuitive Colorimeter) that facilitates measurement of the appropriate tine. This has been validated, and is now commercially available. It is being used increasingly widely and it appears to be particularly helpful for certain children with reading difficulties. A simplified assessment procedure using transparent overlays has been devised for screening purposes, and will be further developed over the next few years.

Future plans involve collecting more evidence on the capacity of the model to deal with migraine headaches, and will investigate the therapeutic value of using ophthalmically-tinted lenses. Other applied issues include the study of photophobia in recently head injured patients.

At a more theoretical level, the processes underlying visual discomfort will be further explored using psychophysiological measures including pupil size, eye movements and cortical evoked responses. The optimal method of testing the hyperexcitability theory however, is through PET scanning or functional MRI, both of which should be available locally in the near future.

Wing and colleagues have continued to study the timing and control of movement. His widely used two-level model for repetitive movement has been elaborated into a multi-level, hierarchical model for rhythm. This has been shown to be applicable to situations in which the control of timing is explicit, but not to the timing of patterned responding in tasks such as typing and handwriting in which the rhythm emerges from the task, rather than being explicitly imposed. Yet a third variant on timing occurs when a number of individuals attempt to synchronise their responding, as in a rowing eight, where the externalisation of potential cues for coordinating individual timing provides valuable hints as to the underlying mechanisms.

Research into the variability of timing has been complemented by analyses of force variability. This has developed into a concern with temporal coordination between grip force used to hold an object and forces created in moving the object in the environment. Subtle adjustments in grip force have been demonstrated that anticipate changes in forces acting on the object being gripped. Similar finely tuned relations between natural reaching and hand-shaping also cast light on the subtle link between perceptual factors and the anticipatory control of action. The breakdown of such processes following stroke, cerebellar disease or in utilising an artificial hand, illuminate the underlying mechanisms, and have substantial implications for therapy.

Balance during standing presents another example of the interplay of perception and action. Anticipatory and reactive components of balance have been studied using a recently patented device that provides a simple measure of the subject's weight distribution. Together with a newly developed method for disturbing balance by applying a gentle force to the pelvis, this approach will yield useful tools for analysing and helping remediate balance experienced in neuromuscular disorder.

Perception and timing are both crucial to interception tasks such as hitting or catching a ball, or braking to avoid a car collision. While an influential model exists that accounts for these phenomena in terms of a servo system based on a single optical parameter, Tresilian has demonstrated that other models are also consistent with existing data, and has begun to explore both the theoretical advantages and drawbacks of the various models, and to test them empirically.

Future work on control of force and timing of movement will consider the interesting case of bimanual movement, using procedures that load either common, central or lateralised peripheral components of bilateral movements. This work will be complemented by analyses of ensemble timing as it occurs, for example, in rowing. An important aspect of both approaches will be the development of methods for separating long- and short-term fluctuations in movement measures and the evaluation of the roles of local and global information in maintaining coordination of force and time parameters through an extended sequence of coordinated movement.

When subjects make movements, such as raising an arm, that disturb balance, automatic postural adjustments occur. These appear to have important parallels with anticipatory grip force adjustments. The possibility of a common, underlying neural mechanism will be studied via correlations across a range of experimental conditions in normal subjects, and also via evidence of associated deficits in patients with Parkinson's disease, and patients suffering from cerebellar damage. Standing balance will be investigated in stroke patients with a training study that will include conditions designed to improve anticipatory weight adjustment associated with initiation of walking. This work will be complemented by a study of stroke patients who spontaneously develop a maladaptive response to instability and are particularly prone to falling. Finally, research on balance will continue with further analysis of evidence suggesting that certain concurrent mental tasks may interfere with balance, particularly in elderly subjects.

Programme 2: Attention and Cognitive Control

Attention refers to the process whereby an organism controls the flow of information through its nervous system, selecting and coordinating certain stimuli and actions, and inhibiting others. It has continued to form a major component of the Unit's research programme since Broadbent's classic work in the 1950s rescued the concept of attention from decades of neglect. It typifies the Unit's approach to science in drawing on a very wide range of tasks, from the selection of simple stimuli to abilities as complicated as learning a new word-processing programme, and it attempts to tackle the underlying problems using methods with a wide range of analytic levels, from single unit recording in monkeys to the construction of general models of cognitive processing based on techniques of artificial intelligence. We believe that an adequate theory of attention will require a conceptual understanding at all of these levels.

For purposes of exposition, the work can be split into two major groups; one of these is concerned principally with attention as a means of selecting one component of a complex perceptual array. This line of work tends to use comparatively simple tasks, and is already demonstrating the value of combining cognitive psychology with more neurobiological approaches. The second group of studies is concerned with the selection and coordination of action, and has stronger links with the field of cognitive science, with clear implications for the practical problem of understanding people interacting with complex technological systems. It seems likely that both domains rely on broadly similar underlying mechanisms, in which control is exercised by the inhibition or excitation of neural systems; but it is likely that the underlying neural systems involved will prove to differ.

Duncan contributes to both aspects of attention research, with his integrated competition model of visual attention being a good example of the fruitful interaction of cognitive psychology with neurobiology and neuropsychology. Recent work supports the view that visual attention arises through cooperative activation in multiple brain systems, which converge to work on the same visual object The process develops over several hundred milliseconds, as revealed by extended interference between one object and another. Work with Desimone at NIH has used single unit recording techniques in monkeys to study the neurobiological basis of selective attention. When the animal is required to look for a particular object, neurons in the inferotemporal cortex that are associated with perceiving that object appear to be primed, while those associated with non-targets are suppressed. Work using both lesion and PET scanning methods in human subjects provides further evidence for attention as a process involving integrated competition within multiple brain systems. The process of inhibition in visual attention has been simulated by Houghton using a connectionist computational model that is proving influential in the area of memory as well as attention. In experimental work, Lavie has begun to investigate limits over the voluntary control of visual processing.

Complaints of attentional difficulties occur widely in patients suffering from stroke or head injury, and this has led to active collaboration between cognitive psychologists such as Duncan and Lavie, whose primary concern is with the understanding of normal attention, and neuropsychologists from the rehabilitation group. Goodrich has been studying the phenomenon of extinction, whereby a patient may be able to detect a single stimulus in the visual field contralateral to a lesion, but will fail to do so when a second stimulus is present in the ipsilesional field. The tendency to neglect aspects of personal space is one of considerable theoretical activity both within the Unit and more widely, with theoretical developments beginning to generate promising new methods of treatment, some of which will be described in the rehabilitation section. Preliminary studies indicate that visuospatial neglect may be reduced by input from other modalities, and exacerbated when overall cognitive load is high. We propose to develop and teach coping strategies based on these findings.

Duncan and colleagues hope to increase the extent of interdisciplinary collaboration, combining models and techniques from cognitive psychology with a range of psychophysiological measures, including PET scanning and fMRI measures. The work with Desimone will continue, using single unit recording to investigate the neurophysiological and anatomical basis of attention. Questions to be addressed include the nature of competition both within and between visual systems, and the separability of visual and spatial aspects of attention. A panel of patients with clearly specified lesions is currently being formed, and comparisons will be made between patients with parietal, occipitotemporal and frontal lesions to test hypotheses about the various processes underlying selective attention.

While attention has been regarded as a legitimate topic of investigation within psychology for at least 30 years, the related issue of the nature of consciousness has, until recently, been treated with more scepticism. However, evidence from the laboratory and the neuropsychological clinic from such phenomena as blindsight (the capacity of some cortically blind patients to process visual information in the absence of visual awareness) and implicit learning in amnesia have forced cognitive psychologists to come to terms with the need to develop theories of conscious awareness. Marcel, whose earlier empirical work contributed to the acceptance of consciousness as an important topic within cognitive psychology, has been devoting a good deal of his time to both theoretical and empirical work concerned with the question of whether consciousness should be regarded as unitary. Marcel takes several different lines of his work to suggest a division between phenomenal experience and higher-order reflexive consciousness. This second aspect of consciousness is closely linked to focal attention and self-monitoring, and plays a role in what we can report and what gets consciously remembered. The lack of access between these levels of consciousness can be seen as underlying research on anosognosia, where a person is unaware of a neurologically acquired deficit. Work with Tegnèr in Stockholm has shown that patients with anosognosia are not simply confused, but that their degree of awareness of their deficit may depend crucially on the way in which the relevant information is probed.

A related aspect of conscious awareness is that involved in reporting bodily sensations, a capacity that is very important in medical diagnosis, but little studied. Future work will further investigate the capacity of subjects to localise pain, and will further develop Marcel and Tegnèr's anosognosia scale as a general clinical instrument.
The breakdown in monitoring mechanisms found so strikingly in cases of anosognosia has clear parallels to the need to monitor our own abilities in everyday life. Self-monitoring played an important role in a recent attempt to minimise anaesthetic accidents by encouraging anaesthetists to participate in an anonymous incident-reporting system. Sellen has begun to investigate some 3,000 such reported incidents, developing a categorisation system as part of an attempt to understand the errors and reduce them. The most frequent incidents appear to be simple slips ("the wrong drug problem") associated with haste and distraction. The analysis is continuing and a more detailed observational study is being carried out in association with Papworth Hospital.

Understanding such slips of action requires a theoretical framework that is capable of dealing with attentional control as part of the coordination of action needed to perform complex tasks. Duncan's work in this area stemmed from a study of driving behaviour, but has in recent years concentrated on conceptualising and studying the way in which actions are controlled and scheduled. Again he has used a range of approaches, from the study of goal neglect in laboratory tasks, through studies on patients with frontal lobe lesions, to individual differences in general intelligence. Evidence from PET studies, from lesion studies and from the laboratory all indicate the importance of frontal lobe systems in setting and monitoring goals, and in the processes underlying differences between individuals in general intelligence.
Future research will extend the work on patients with clearly specified lesions, and will utilise dual task methodology to develop a more precise understanding of the nature of executive control. It is planned to combine carefully selected tasks with PET scanning, in order to investigate whether executively demanding tasks place demands on specific areas of the frontal lobes, to what extent they limit performance by common demands on content-specific areas, and to what extent a demanding task is simply one that requires many different functional areas.

The nature of executive control is a particularly important issue in attempting to understand how people perform complex tasks such as those involved in human-computer interaction. This area has been extremely active in recent years, with Barnard playing a particularly central role in ESPRIT programmes, and Richard Young being heavily involved in both the Alvey and Joint Councils Initiative in Human-Computer Interaction. These programmes have brought opportunities, but also incurred substantial logistic costs, from the need to coordinate large international teams, and to operate with other disciplines and with colleagues in industry. At a practical level the programmes have concentrated on the analysis of complex technological tasks such as understanding the utilisation of a new word-processing programme, or analysing the way in which experienced or naïve users might interact with a new computer system. The aim is typically to provide ways in which software designers can be made aware of the limitations of the user, so as to design systems that are user-friendly. Both Barnard and Young have tackled this task by developing high-level models of the way in which people cope with complex tasks.

Barnard has developed a model, Interacting Cognitive Subsystems (ICS), that conceptualises the user in a way that was originally developed for understanding language use, but has subsequently been substantially extended. (See also section on Cognition and Emotion). Within the attention and cognitive control programme the basic extensions have focused on the interpretation of visual scenes. In one of the more practical studies making use of such theoretical extensions, the ICS system was taught to human factors students, who proved to be able to use it effectively, given the appropriate task-description vocabulary. They did not need to understand the underlying theory.

Dissemination is clearly an important issue, and the work has been widely presented and is beginning to appear in text books for software designers. Much of the work in this area is disseminated electronically through Internet, a process that allows dissemination to be monitored. Hence we know that in the last five months there were 1,160 requests for documents from this project. Most (740) were from the various collaborators, but there were some 420 requests from unrelated investigators from a total of 18 different countries. It is becoming clear that this will be an increasingly important method of scientific dissemination, particularly for items such as computational models which are not readily communicated within the current journal system.

Over the next five years, Barnard will be devoting a rather larger amount of his time to applying his model to the study of cognition and emotion, but will continue to operate within the general area of executive control, making the model available in tutorial form, and collaborating in a venture to specify the theory in more formal terms. Future empirical work will concentrate on the task of searching iconic arrays, and on the way in which information is coordinated across sensory modalities.
Young has also used human-computer interaction as a test-bed for a more general conceptualisation of cognitive function, in this case based on the SOAR architecture developed in the U.S. by Newell. The model has a highly constrained architecture, and uses learning and problem solving procedures to tackle a wide range of cognitive tasks. It has been applied by Young and his colleagues in recent years to study the mistakes made by subjects first encountering a Macintosh computer, and to demonstrate the role of mental models in using simpler devices such as calculators. Work with Logica, a major British information technology company, has been concerned to develop methods of assisting software designers in creating usable systems.

Future work will concentrate on using the model to conceptualise the representation of task goals, and to provide a better understanding of the way in which people operate within the visually-based systems that are becoming increasingly dominant within HCI.

At a more theoretical level, Young will be concerned with comparing the way in which different conceptual models within cognitive psychology account for the same phenomena. For example, although the concept of working memory plays an important role in SOAR, different assumptions about working memory are made by SOAR, by Barnard's ICS, and by Baddeley and Hitch's working memory model. It would be valuable to know which of these differences are largely notational, and which reflect incompatible but testable basic assumptions.

The final component of this section concerns the Unit's work on driving, a topic that we have studied for many years, but which has now lapsed with the retirement of Brown and the departure of Groeger. The group has been very active and well supported over recent years, when they have been paying particular attention to the process of learning to drive. One important finding from their careful analysis is the huge difference in amount of practice that occurs across different subcomponents of driving skill. Typically subjects have a great deal of practice on control skills, but much less on skills of decision, particularly about situations in which they need to assess what other road users will do. This work was accompanied by another programme concerned with the judgement of risk, demonstrating that video traffic scenes provide a valid means of assessing the capacity for risk judgement. There are clear implications for ways in which driver training should be changed, and of methods whereby decision processes may be practised and subsequently assessed. Groeger has now moved to the University of Leeds where, I am happy to note, the programme will be continuing.

Though our commitment to the analysis of complex, real-life tasks remains strong, with the termination of work in driver behaviour, and Barnard's increasing commitment to cognition and emotion, overall effort in this area is somewhat diminished. At the same time, the neurobiological aspects of the attention programme are increasingly promising, and complement developments in cognitive neuroscience in Cambridge with establishment of the MRC Brain Repair Centre, the Wolfson Functional Brain Imaging Centre, and the Innes Centre for primate neuropsychology. Through collaborations with these other Cambridge groups, and elsewhere, we expect lively developments in this aspect of the attention programme.

Programme 3: Memory

The capacity to learn and remember plays a crucial role in human cognition, and has always formed an important area of research at the APU. Five senior scientists contribute to the area, although only Bekerian and Baddeley would probably regard themselves as primarily working on memory. Consequently, the boundaries between this and other Unit programmes are often indistinct. For example, much of the research on working memory could be categorised as part of the attentional project, while research on prospective memory and on test development is highly relevant to similar concerns within the rehabilitation group, as is Maylor's work on memory and ageing, and Murre's modelling of amnesia, which ties in directly with theoretical work on rehabilitation by Robertson. Links with research on language are equally obvious; a good deal of Bishop's work is related to short-term and working memory, while Patterson's work on semantic memory forms part of her general interest in language. Finally, Bekerian's work again ties in very clearly with the theme of cognition and emotion.

Human memory can be regarded as an alliance of separate but interacting systems that have in common the function of storing information and subsequently making it available when needed. It can be broadly divided into short-term or working memory, involved in the temporary maintenance of information being used to perform other cognitive tasks, and long-term memory. Long-term memory itself can be subdivided into episodic memory, the capacity to recollect specific experiences, and semantic memory which can broadly be regarded as knowledge of the world. Finally, in recent years it has become increasingly clear that we need to distinguish between the previously-described systems which provide information that is explicitly available, and implicit learning and memory, involving the learning and retention of skills and access to information under conditions where the subject may demonstrate learning by performing a task more efficiently, without necessarily having any recollection of the learning experience. Work at the Unit has addressed all of these areas.

Baddeley and Andrade have been concerned with developing a model of working memory originally proposed in 1974 by Baddeley and Hitch. Work has concentrated on two components of the model, namely the phonological loop, a system for temporarily maintaining speech-based information, and the central executive, a limited capacity attentional control system.

Much of the work on the phonological loop has been designed to test the hypothesis that it has evolved as a mechanism for language acquisition. Work carried out in collaboration with Papagno and Vallar in Milan, and Gathercole in Bristol, supports this view by showing that: (1) Adults with a neurologically acquired phonological loop deficit and children with a developmental deficit both have difficulty in acquiring new vocabulary. (2) Normal children show a clear association between a task involving immediate phonological repetition and vocabulary development. (3) Variables that are known to interfere with the operation of the phonological loop also interfere with the acquisition of novel vocabulary, but not with the capacity to learn to associate pairs of meaningful words. We are currently testing the hypothesis that the same system is necessary for the acquisition of grammar.

The central executive is the term applied to the system that is assumed to control working memory. It is a complex and almost certainly multi-component system. We are investigating it by splitting off individual executive processes, which we then analyse with a range of methods involving both neuropsychological patients and normal subjects.

One postulated function of the executive is to coordinate concurrent tasks. This has been studied in collaboration with Spinnler, Della Sala and Logie in Milan and Aberdeen, principally using studies of patients suffering from Alzheimer's disease (AD). The method involves titrating the level of difficulty of concurrent visuo-spatial and verbal tasks to a point at which patients and controls are performing at an equivalent level on each task alone. When combined performance is required, AD patients show a disproportionate decrement, whereas the normal elderly show no such effect. As the disease progresses, dual task performance deteriorates substantially more rapidly than single task. We have also modified our original methodology so as to provide a simple paper-and-pencil version, and have replicated our original results. We are now beginning to explore the generality of this deficit and find that it is also present in a particular subsample of patients with frontal lobe damage.

Other executive processes already investigated include random generation, in which the subject is required to produce an unpredictable stream of responses, a task that appears to load heavily on the executive. More recently we have been investigating the processes involved in responding to stimuli while ignoring irrelevant items presented within either the same or a different modality. Preliminary results suggest that the capacity to resist within-modality interference is particularly susceptible to the effects of ageing.

In the area of long-term memory, much of Baddeley's work has been concerned with collaborative research on rehabilitation, and with the development of improved memory tests. These have included a test of semantic memory, based on both speed of comprehension and capacity of vocabulary, while a second test measures episodic memory performance. Visual and verbal recall and recognition are tested using naturalistic material, comprising coloured photographs of doors and sets of people's names. Preliminary results suggest that the test is highly sensitive, is good at differentiating visual and verbal memory deficits, and furthermore, is enjoyed by the patients. Future research by Baddeley on working memory and long-term memory will form the substance of a separate proposal for work to be carried out elsewhere.

Research by Maylor has been concerned with analysing the cognitive changes that accompany normal ageing. A theme underlying much of her work has been the extent to which ageing represents a general overall decrement, and to what extent certain cognitive functions may be disproportionately impaired or preserved. One series of studies required subjects to make judgements about pictures of objects or people, and to produce the appropriate name. Older subjects uniformly tended to respond more slowly, but this deficit was disproportionately large when naming was required. This effect seems to represent the more frequent occurrence of "blocking" in the elderly, where the name was on "the tip of the tongue". Such a deficit was not found in a selected group of young and elderly experts. Participants in "Mastermind" appeared to show no effect of age on response time, suggesting that up to a point at least, expertise can counteract the slowing in retrieval.

Another area in which the elderly seem to be disproportionately impaired is in experimental studies of prospective memory; remembering to do things. Even when they have demonstrated that they remember the relevant instruction, elderly subjects are still more likely than the young to forget to carry it out when tested under laboratory conditions. Under real-life circumstances however, the elderly tend to be rather better than the young since they make more systematic and effective use of reminders and memory aids.

Maylor's future work will further investigate the area of prospective memory, in particular focusing on factors that may make this function particularly susceptible to the effects of age, such as the need for self-initiation of action, the demands of concurrent activity, and the extent to which environmental cues can prompt recall. Other research will be concerned with further analysis of the effect of age on retrieval from long-term memory. Finally in collaboration with Wing, she will study the interaction between extent and type of concurrent cognitive activity and postural stability, an important practical problem in the elderly.

Prospective memory also featured in a study carried out by Sellen at Rank Xerox, on memory problems in the workplace, where prospective memory lapses proved to be the major source of reported memory problems. This study led to the development of a new computer-based reminding system that is currently under development by Rank Xerox. The technological capacities of Rank Xerox also allowed Sellen to carry out a more theoretically-based study on factors influencing the prospective memory task of remembering to press a button every two hours, a simulation of pill-taking. Subjects wore an "active badge" that monitored their location and their responses. The results suggest two separate components, one concerned with self-prompting and a second dependent upon environmental cues. Further work will use this technology to study the effect of a range of variables including the extent to which the subject "worries" about performing the task. If successful, the device may well prove useful in studies of cognition and emotion.

The classic topic of memory research is episodic memory, the capacity to recollect a specific experience. This is the focus of the research by Bekerian, who is particularly interested in the study of memory under the naturalistic circumstances that occur when a witness attempts to recall a dramatic incident such as an attempted rape, or a series of incidents such as repeated physical or sexual abuse. Evidential testimony represents a situation in which it is essential to assess the accuracy of recollection, and to avoid the danger of distorting memory through the process of interrogation. Bekerian has investigated aspects of testimony that influence the degree of perceived authenticity, including consistency, spontaneity, and the presence of reported perceptual detail. These features were modulated by the method of reporting; for example, certain interrogation methods encourage the subject to try to visualise the scene, and there is evidence to suggest that this leads to more perceptual detail, some of it erroneous. Such detail may then give rise to an unwarranted increase in the confidence of the witness and the preparedness of a juror to accept the report. Other factors influencing accuracy include whether recall is spoken or written: written recall seems to lead to a reduction in correctly recalled information, possibly because the more formal response mode encourages a more "polished" product.

Bekerian has worked extensively with the Home Office, the Police, the County Statutory and Caring Agencies and the Regional Health Authority, all of whom are concerned to improve their quality of interviewing, particularly in areas such as that of child abuse. Interviewing for therapeutic or evidential purposes demands very different styles, with the Home Office stipulating that therapeutic-style interviewing should not be used to obtain evidence. Bekerian's work on the analysis of interviewer behaviour has led to many requests for advice in developing better training of interviewers. Currently, the Crown Prosecution Service are forced to abandon a large number of cases on the grounds of inadequate interview procedure. Such statistics should provide evidence of the success of our attempts to improve interviewing skills.
Investigation into methods of evaluating the accuracy of testimony, and of avoiding distortions during the interview process, will continue. The commonly used "cognitive interview technique", which among other things encourages the use of imagery, is based on theoretical memory research, and does appear to be an advance on previous techniques, but is almost certainly capable of improvement.

New lines of development include the reviewing of current practice in investigating serial crime, including the technique known as offender profiling. Bekerian has been requested to compare approaches in Britain with those in the USA, Continental Europe and Australia, and it seems likely that this may well give rise to further empirical work. Finally, a joint interest with Dalgleish (who is about to join the Cognition and Emotion group) concerns the influence of emotion on the memory of patients suffering from post-traumatic stress disorder, in which persistent vivid memories of the incident often form one of the more distressing symptoms.

While episodic memory may be the most striking and characteristic aspect of human memory, semantic memory which comprises the repository for our knowledge of the world is equally important. Karalyn Patterson has been studying semantic memory in collaboration with Hodges of the Neurology Department at Cambridge. They have concentrated on three issues: the anatomical localisation of the system responsible for semantic memory; whether semantic and episodic memory represent genuinely separable systems; and the functional organisation of semantic memory. The work is based on longitudinal studies of the decline of semantic memory in two types of patient: Alzheimer's disease (AD) patients, in whom a prominent episodic memory deficit is usually associated with a subsequent decline in semantic processing; and patients with a recently labelled syndrome of semantic dementia, in whom the semantic deficit occurs in a relatively isolated form. Brain imaging studies suggest that the temporal neocortex is particularly important for semantic memory, while differences in the specific location disrupted in the two patient groups hint at the possibility of anatomically separable semantic and episodic systems. Further investigation of this issue will depend upon a more detailed analysis of episodic memory in semantic dementia cases; the issue is not straightforward since, even if semantic and episodic memory reflect separate systems, there is no doubt that they interact.
A study of the breakdown of semantic memory in patients with semantic dementia suggests a consistent pattern, with specific information (such as the fact that dogs bark) being lost before more general information (e.g. dogs are animals). This is compatible with a widely-held hierarchical concept of the structure of semantic memory, but is also open to other explanations. Differences in number of relevant features in a distributed network could make the higher level concept of animal more redundant and hence more robust to partial disruption by brain damage.

Future plans include serial in vivo imaging of semantic dementia patients, followed by neuropathological analysis at post-mortem. Such a strategy will also be pursued with AD patients, with the additional aim of relating the more varied pattern of deficits in these patients to the localisation of neurodegeneration. The issue of whether the decline in semantic memory represents loss of basic semantic representations, or impaired access to relatively preserved representations, will be tackled using measures of consistency across successive tests, an approach that raises some challenging statistical issues. Finally, in addition to detailed exploration of semantic and autobiographical memory in patients during the early stages of decline, it is proposed to develop methods that will allow semantic processing to be assessed during the later stages of dementia. Work by Jean Mandler in San Diego on pre-linguistic babies employs a measure of the time that a baby will spend looking at an object as an indication of its degree of perceived novelty. When four items from a similar general category, for example living things, are presented in succession, presentation of a fifth living thing leads to shorter inspection than presentation of a non-living object. Using this technique, Mandler has shown that babies acquire the living/non-living distinction at a very early age. Patterson and Hodges predict that this distinction will be robust in dementia patients.

Andrew Young, who recently joined the Unit, is also concerned with semantic memory, but is particularly interested in the way in which people are perceived and remembered. Having previously concentrated on face perception, he is interested in the extent to which memory for known faces uses the same underlying cognitive and neural structures. One way of studying this is through an analysis of prosopagnosia, a deficit in face recognition that occasionally results from brain damage. His work has identified at least two types of patient, both of whom experience great difficulty in recognising familiar faces, but who differ in their capacity to generate images of faces. One type of patient is able to form accurate images of faces which are no longer recognised, and hence can answer questions about the appearances of pre-morbidly familiar people, whereas the other has no capacity for imaging faces. This pattern of results suggests a multi-stage system for recognising faces, as do other results obtained before his appointment to the Unit by Young and his co-workers demonstrating that some prosopagnosic patients show implicit recognition of famous faces that they do not report as familiar. This can be demonstrated by showing that presentation of a non-recognised politician's face will influence the subsequent recognition or categorisation of his name.

Future work will develop new techniques for testing the theory both with normal subjects and with patients. Prosopagnosic patients are rare, but it is commonly the case that patients suffering from anterograde or retrograde amnesia have difficulty in face recognition. A detailed examination of such patients will first of all identify any for whom face recognition problems are particularly marked, and will then check whether the deficit is one of recalling the person, which would be reflected in a parallel difficulty in recognising a voice or providing information about a named person, or is specifically visual.

Further studies will exploit the fact that people represent an interesting class of semantic concept, since their visual appearance changes as they age, and new facts about them are being constantly learnt. This allows a test of one of the classic issues of concept formation, that of whether concepts are based on a large number of stored instances, or represent an abstraction from the individual experiences. The former might suggest that it would be easier to associate a picture of a person's face when young with information from that time, (e.g. that Cliff Richard sang "Living Doll") than it would be to associate an early photograph with a more recent piece of information.

A prominent feature of the Unit's work on learning and memory over recent years has been an increasing interest in computational, and in particular connectionist, modelling techniques. This powerful new modelling approach, combined with the judicious use of empirical data, allows much more detailed and ambitious models than were possible using a purely descriptive conceptualisation. Interestingly, there are close affinities between recently developed connectionist models of human learning and earlier mathematical models based on conditioning in animals. Work by Shanks has shown that an associative conditioning model of this type gives a very good account of whether human subjects perceive pairs of events as causally related. Shanks has also explored connectionist models of category learning, demonstrating major problems with a standard learning algorithm, back-propagation, and proposing an alternative connectionist model that avoids these difficulties.

Murre has been developing a general connectionist model of long-term memory and forgetting. He begins with the problem of massive interference that occurs in some basic back-propagation models, spelling out a number of ways in which this can be avoided. He has been particularly intent on ensuring that his model is biologically plausible, with the result that it has formed the basis for a collaboration with Robertson to generate the framework for a general model of rehabilitation. Concern with neuropsychological data has led to the first connectionist model of learning that also gives a good account of both anterograde and retrograde amnesia. Research on forgetting has been comparatively neglected over the last 20 years, and Murre's model should provide an excellent basis for a revival of interest in this important topic.
One final area in which connectionism has influenced the Unit's work is in attempts to model the phonological loop component of working memory. Norris, Page and Baddeley, funded by the Joint Council's Initiative in Cognitive Science, are capitalising on the fact that modelling the loop involves two of the fundamentally important issues of connectionist modelling: (i) how to represent serial order; (ii) how to capture the process of chunking whereby individual units are aggregated to form larger units, which many models assume to be the basic process of long-term learning. Extensive existing data place valuable constraints on possible models. In addition to developing a connectionist model that gives a good account of the initial data set we have also developed a more abstract mathematical characteristic of the serial recall process which not only gives a precise quantitative account of the data but also helps specify a set of properties which must be possessed by any adequate model of serial recall.

The next issue is that of obtaining a deeper understanding of simulations by mathematical modelling. Houghton's competitive queuing model, initiated at the Unit and subsequently developed at University College London, has been influential in modelling development. Future work will elaborate the mathematical analysis of the connectionist model, and will concentrate on the question of how items are represented in memory, and how the serial recall process avoids the inappropriate repetition of earlier items while at the same time being able to cope with sequences in which the same item occurs more than once.

Programme 4: Language and Communication

This area which is central to the study of cognitive psychology continues to form a major component of the Unit's programme. Our approach is characterised by the utilisation of a wide range of techniques and subject groups, and by the attempt to blend pure and applied research.

Cutler, who has now left to become a director of the Max Planck Institute for Psycholinguistic Research, and Norris are concerned with the problem of how we segment the continuous stream of a spoken utterance into its constituent words. They tackle the problem using standard experimental paradigms, computational modelling and cross-linguistic studies. A basic hypothesis, the Metrical Segmentation Strategy, proposes that prosodic information is used to segment the speech stream, with the details of the strategy depending on the specific language. English rhythm is based on the foot, French on the syllable, and Japanese on the mora (typically a consonant-vowel unit). Stressed vowels have been shown to be longer, louder, and greater in pitch movement than unstressed syllables, while a detailed analysis of the stress pattern of English indicates that a high proportion of words have stress on the initial syllable, making it a valuable cue to segmentation in English.

The Metrical Segmentation Strategy is utilised by Norris's computational model SHORTLIST, which represents the first new model of speech perception for over a decade. Unlike its major competitor, the TRACE model of McClelland and Elman, Norris's model does not require the duplication of processing networks, and does not assume an interaction between processing levels. His computational model is simpler in assuming only a billionth as many inhibitory connections, and is able to cope with a vocabulary of over 20,000 words, as opposed to 1,000 words for the TRACE model. The model has been shown to account for existing data well, and has been tested empirically in a series of studies in which subjects were instructed to detect words embedded in other words, for example the word CAM in CAMEL, where the results are compatible with the assumption of competition between word representations as proposed by the SHORTLIST model.

Collaborative work between Cutler and Jusczyk has shown that young infants from English-speaking families prefer to listen to words with stress on the initial syllable, suggesting that lexical segmentation is being acquired at a very early age. Future work will evaluate this further, using statistical simulation of the learning process, and utilising the model to investigate the nature of segmentation in more detail. There is, for example, evidence that subjects also make use of statistical characteristics of syllable and word endings. Two basic assumptions of the SHORTLIST model, namely the modularity of its component processes, and its bottom-up rather than interactive architecture, will be subjected to more rigorous empirical test. Finally, studies of the role of syllabic information in lexical representation will be extended to native speakers of French and Dutch in order to assess their cross-linguistic generality.

Karalyn Patterson also uses connectionist modelling and cross-linguistic methods to study word recognition and production, but relies more heavily on neuropsychological evidence than do Norris and Cutler. She is particularly interested in the interaction, in various language tasks such as reading aloud, between representations of a word's orthography (spelling), phonology (sound or pronunciation) and meaning. A major component of her work involves study of the breakdown of language in patients suffering from progressive neurological disease, a programme that is carried out jointly with Hodges from the Department of Neurology in Cambridge. Patterson and Hodges have been particularly interested in "semantic" dementia, a neurodegenerative condition involving focal atrophy of temporal neocortex. Because this condition produces relatively selective loss of semantic memory, it offers the opportunity to study the impact of loss of word meaning on other aspects of language processing. For example, the fact that patients with semantic dementia develop surface dyslexia (a reading disorder producing regularised pronunciations of words with irregular spelling-sound correspondences, e.g. reading pint to rhyme with "mint") suggests an important interaction between meaning and phonology in reading.

Work on normal subjects has also been concerned with the interaction between orthography, phonology and meaning. Recent work by Patterson and Strain has demonstrated that the speed and accuracy with which normal adults read written words aloud is affected not only by the word's frequency and typicality of spelling-sound correspondence, but also by a semantic variable, imageability (low imageability words are abstract, high imageability words refer to concepts with many sensory/motor properties). This result again implicates word meaning in the transcoding from orthography to phonology, a task which does not, on the face of it, demand access to meaning. The nature of this transcoding has also been studied in native readers of Japanese kanji, a writing system completely different from alphabetic English, to yield insights into aspects of language processing which are universal and those which are tuned to language-specific characteristics.
Results from normal and impaired English readers play a vital role in evaluating computational models of the transcoding from orthography to phonology. Whereas Norris' connectionist reading model is principally concerned with simulating normal readers' response times to name words under laboratory conditions, Patterson's collaborative work (with McClelland and Plaut in Pittsburgh and Seidenberg in Los Angeles) focuses on the capacity to simulate a range of neuropsychological reading deficits. The model does not yet have a semantic component but is capable, after extensive training, of correctly pronouncing virtually all monosyllabic words, even low-frequency irregular words. When "lesioned", however, it does not successfully simulate the pattern of reading performance typically observed in semantic dementia -- preserved naming of regular words and nonwords, impaired and frequency-modulated naming of irregular words. If the training of the model is stopped at an earlier point, however, a better match to surface dyslexia in semantic dementia is obtained. Future work will explore the hypothesis that degree of skill in spelling-sound transcoding is determined in part by interaction with word meaning.

A third approach in Patterson's work on language has involved PET scanning, in collaboration with Wise and Price at the MRC Cyclotron Unit and Howard at Birkbeck College, London. Studies of both spoken and written word recognition in normal subjects have revealed maximal rCBF increases in regions at or very near those implicated by lesion studies. A novel finding is the effect of rate of spoken word presentation: although most areas of bilateral temporal cortex show monotonic increases in rCBF with increasing word rate, Wernicke's area (known from lesion studies to be specialised for spoken word recognition) is maximally activated even at slow rates. Future functional brain imaging work will take advantage of the fact that techniques are now capable of dealing with single-subject data, making neuropsychological case studies feasible. The role of the right hemisphere in language processing will be investigated with single-subject data from normal subjects and patients with a variety of sites/sizes of left-hemisphere lesions.

Further work on semantic dementia will concentrate on the role of meaning, not only in language tasks (such as word and sentence repetition, and lexical decision), but also in non-language abilities such as object recognition. The progressive decline of language and other cognitive abilities in Alzheimer's disease also provides a valuable source of information about the interaction between different domains of representation. The completion of a three-year longitudinal study carried out jointly with Hodges, with approximately 30 Alzheimer's patients tested at six-monthly intervals on a wide range of tasks, will yield a rich database for hypothesis testing on the structure of language and its decline. Work with Tyler (at Birkbeck College, London), using on-line measures of spoken language processing, is beginning to provide significant advances in the study of both the nature of the meaning loss in semantic dementia, and the separability of semantic and syntactic components of language.
New work will concern aspects of speech production in patients with another form of neurodegenerative disease -- progressive nonfluent aphasia -- involving deterioration of phonology and syntax but relatively preserved single-word meaning. Similar longitudinal analyses will be made of patients with progressive anomia, whose spontaneous speech is fluent but "empty" of specific content words. Our current studies constitute the first systematic experimental analysis of progressive nonfluent aphasia, and both types of patients are yielding results germane to models of speech production. Collaborations with neuropsychological colleagues in Japan will once again enable evaluation of the generality or language specificity of some of these findings. For example, the reading performance of Japanese patients with semantic dementia will be assessed for parallels with surface dyslexia in English.

While Patterson is using the breakdown of language in neurodegenerative disease as a way of understanding its normal operation, Bishop is learning about language processing by studying failures of acquisition in children with specific language impairment (SLI). Although SLI is a clinical problem of some magnitude, it is still poorly understood, being associated with a relatively wide and varied pattern of symptoms. Bishop has used a twin study to investigate the aetiology of SLI, the existence of sub-types and their cognitive basis. Some 90 pairs of same-sex twins were selected on the basis that at least one of the pair has significant language problems, performing at or below the 10th centile on one of four carefully-chosen language tests. The importance of an inherited factor was indicated by the higher concordance (54%) in monozygotic than in dizygotic twins (30%). There was no clear association with other medical problems, other than toxaemia, a poorly-understood disease of pregnancy which may possibly be serving as a marker for other immunological abnormalities.

Future work in this area will use the twin study method to try to obtain a better definition of the SLI phenotype, in particular attempting to decide whether it represents a qualitatively distinct disorder, or a quantitative departure from normality. The issue of heritability of language skills in the normal range will be investigated by testing a normal sample of monozygotic and dizygotic same-sex twins. Data from the SLI twin sample will be further analysed to study the link between language, literacy, motor development and handedness. This sample will also be used to try to identify the best marker for SLI, in particular testing Tallal's proposal of a deficit in rapid serial processing, Gopnik's suggestion that syntactic problems are fundamental, and the proposal by Gathercole and Baddeley, and by Bishop herself, that a deficit in working memory might be critical.

One problem in dealing with SLI is that of identifying some of the subtler components of language deficit. In "semantic-pragmatic disorder" the child is fluent but verbose, showing poor comprehension and a tendency to interpret language literally, together with difficulties in the more social aspects of conversation such as turn-taking. Bishop and Adams in Manchester are developing a method for conversational analysis that aims to pinpoint such problems, using precise and reliable coding. Future work will investigate the extent to which a deficit in social interaction might be fundamental, using tests of nonverbal behaviour and "theory of mind" tasks taken from the study of autism. Preliminary evidence is not encouraging for these hypotheses, however, suggesting the possibility of some form of more general attentional deficit. It is proposed to modify some of Duncan's tasks for use with children so as to investigate the attentional deficit hypothesis.

Preliminary work on the grammatical deficit in SLI suggests that it does not represent a systematic and specific deficit, since the same grammatical rules will be followed on one occasion and broken on another. The fact that the likelihood of error increases with the length of utterance suggests that the difficulty may be a limitation of general processing resources. Future work will test this in a number of ways. SLI children will be required to repeat sentences varying in length, with and without inserted errors, and will be required to judge their grammatical correctness. This will allow a more precise test of the hypothesis that errors occur when processing load is high. Another way of testing this hypothesis is to combine sentence repetition with a demanding non-verbal task; again the prediction is that grammatical errors should increase in SLI children. Finally, it is proposed to use the methods of language analysis developed from SLI children to study language and communication in a wider range of conditions. For example, work with Skuse at the Institute of Child Health in London will apply the methods of conversational analysis to children suffering from Turner's syndrome, who typically show a mature interactional style but have difficulty in making friends, suggesting a subtle deficit in social interaction.
The more complex communicative aspects of language are also the concern of Wright, although in her case the focus is on written communications rather than with face-to-face spoken interactions. When people use written materials to accomplish a task they need to find the relevant information, understand and remember it, then act on that information. Changes in document design can be used to explore how people integrate these cognitive activities. One component of Wright's programme, concerned with issues of understanding, has examined the way in which readers and writers use such graphic supports as sketch maps and diagrams. She has found two kinds of graphic-text relation that differ in their effects on comprehension if readers study the illustrations while reading the text. Graphics explaining textual details can hinder readers trying to follow the main thread of the text; graphics that offer an organising schema for the text repay study during the course of reading. However, readers do not distinguish these graphic roles and adopt a single strategy of studying the graphics either before or after reading. Document design can encourage readers to change their strategies, allowing the possibility that the design can be tailored to create the optimal reading strategy. However, it was found that writers also adopted inappropriate strategies when giving directions about a route to be taken. They did not spontaneously provide aids such as sketch maps, although they recognised that these would facilitate communication. These studies point to the complexity of the communication skills needed by adults working with written information and graphics.

One situation in which it is often necessary to combine verbal and visual forms of communication is in finding one's way in an unfamiliar environment. Here the focus is on the application of the knowledge gained from reading. In a study concerned with route-finding by patients within a large hospital, provision of a map led to more satisfaction and less re-tracing of the path than relying on the hospital signposting. This occurred despite the fact that use of the map did not lead to more rapid arrival at the destination, because people with the map chose to spend time planning their route and considering alternatives. Thus speed may not always be an appropriate indicator of good information design. Instead, good design needs to meet all the purposes for which people will want to use it. An incidental finding during this research was that one source of problems for hospital signposting is the way in which the same location can be described in different ways, Room C120, the Eye Clinic and Dr Smith's Clinic for example. Emphasising one of these in the appointment letter and using this label on hospital signboards would simplify the problem.
Another component of Wright's programme has examined readers' strategies for finding information within electronic documents where readers can move within the texts in new ways, jumping immediately to related information. Readers compared information from different parts of a text in order to reach a decision. It was found that their choices about how to move within the document were determined by perceptual factors, such as the spatial relation between text windows on the computer screen, even though this sometimes resulted in a search pattern that made it more difficult to remember the information found. Similarly, the inclusion of verbal information in pop-up windows changed its status within the discourse structure of the text and so enhanced subsequent recall. These studies of search strategies show the interplay between information design and readers' allocation of cognitive resources.

Work on creating usable written communications has wide potential applicability, but presents the problem of how such information should itself be communicated to the potential user. Wright devotes a good deal of her time to this issue, contributing to seven British Standards for documentation, and delivering invited talks and keynote addresses to a wide range of professional groups outside psychology. Future research will continue to investigate the roles of graphics in text, extending the work to instructional materials, because the need to create plans for action changes the nature of the dominant cognitive demands on readers, demands which may be met or hampered by animated graphics. When people follow complex multi-step instructions, segmentation into appropriate constituents will help both comprehension and memory. Wright will investigate whether adults do this spontaneously, and if not whether appropriate design can encourage them to do so. A different way of giving instructions is to provide a model of the performance required and ask people to follow it. This technique is often used to encourage people to write in a certain style, although the model given is of the product rather than the performance. This kind of instruction-giving will be explored in order to determine whether people simply imitate the surface features of the model or whether they can abstract the underlying discourse structure.

Work will continue on how people search within electronic documents, but the search tasks will be broadened to include multi-featured targets. Electronic documents allow a rich array of search strategies to be investigated using both verbal and iconic materials, and also a variety of document structures including tabular arrays. This work will contribute to existing theoretical accounts of readers' search activities which have so far considered only printed, linear documents such as student textbooks.
Green is also concerned with the role of communication, and in particular in the analysis of non-linguistic artificial communication systems such as diagrams, tables and programming languages, and their use in the increasingly large array of "information artifacts", such as personal calculators and word processors. Green attempts to identify what he terms "structural features" that are applicable to any well-defined information artifact, from timetables to music scores. The success of his attempt is measured by its capacity to deal effectively with a wide range of very different systems.

For example one desirable aspect of any notational system is that it should be "role-expressive", that is it should allow one to break the system into an appropriate hierarchy of subsystems. Applying this characteristic to the programming languages Prolog and Pascal, Green predicted that Prolog was less role-expressive, and hence should give greater difficulties even to experienced programmers. He tested this by giving programmers specific programs with components snipped out. Their task was to restore the appropriate pieces. As predicted, the Prolog programmers took longer and were more slowed down by increased complexity.

Another study tested the conventional wisdom that graphics communicate more effectively than text. A "box and wire" notation from a widely used graphical programming language (LabVIEW) was compared with a verbal equivalent. It was harder to comprehend and led to more errors, probably because it overloaded the subject's working memory to a greater extent. Another concept from Green's analysis is that of "viscosity", the extent to which a system resists local change. For example, a word processing system might have the characteristics that changing section numbers had extensive knock-on effects, creating problems in the index references etc. Using an existing computer science technique termed "entity-relationship modelling", Green has developed a measure of viscosity. Future work will develop this analysis and apply it to further types of information display, including various forms of graph. Possible ways of dealing with viscosity problems, for example by introducing intermediate levels of abstraction, will be investigated.

The essence of Green's approach is that it should be applicable across a wide range of domains, and future work will involve extending it to new applications. It is proposed to collaborate with colleagues at the University Department of Architecture on computer-aided design (CAD) systems for creative architectural design. A range of small demonstration devices will also be constructed, for example bibliographic systems that illustrate the effects of low and high viscosity. Finally, attention will continue to be paid to the issue of communicating to the wide and varied range of potential users. In particular, it is proposed to develop a teaching package that will illustrate the methods that have been developed, evaluating the package using Open University students.

A final area of research in communication concerns the development by Rank Xerox of videoconferencing systems, and their evaluation. Sellen has compared face-to-face communication with three different conferencing systems, each communicating different aspects of the participants, from voice-only to systems which use multiple cameras. The only major difference observed was between face-to-face conversations and the rest. Conversations mediated by technology all induced a more formal style of interaction, with fewer interruptions and a more studied handover of the floor from one speaker to the next. A related project concerned collaborative working, either face-to-face, or on remote stations linked by up to four video cameras. Subjects sometimes found the additional camera information useful, but had difficulty in reaching a joint frame of reference, underlining the importance of "common ground" in efficient communication. Future work will extend this research to field studies of video links between Welwyn and Venray in Holland for the distributed design and manufacture of photocopiers. The study will be principally observational, and will investigate issues like adaptation to the system over time. It will, however, also be possible to carry out experiments, for example allowing subjects to select amongst different views of the remote site.

Programme 5: Neuropsychological Rehabilitation

This is the newest research area at the APU. It is developing along the lines proposed in our last Progress Report, and approved by the Visiting Subcommittee who recommended additional posts and funding. The recommendation came at a time of considerable financial pressure, and the Strategy Committee decided that the initial development should be funded out of the Unit's existing budget, with an encouragement to return for further funding once the programme was established. We believe that we have reached that point, and will be requesting the additional support proposed.

The programme has made an excellent start scientifically, despite substantial logistic problems. At the time of our last Progress Report, Addenbrooke's Hospital had offered to provide space, as part of an ambitious development of their somewhat limited rehabilitation service. The promised space was, however, reassigned as part of the pre-election campaign to cut NHS waiting lists, and only now after several years of postponement does it seem that we shall finally be moving the Rehabilitation group into adequate accommodation at Addenbrooke's. The proposed expansion of clinical rehabilitation has still not occurred, but fortunately this was less critical than seemed likely because of the substantial increase in strength of the Clinical and Neuroscience Departments at the University Medical School. We have excellent and productive research links with Neurology, Neurosurgery, Psychiatry and Anaesthetics, while the development of the Interdisciplinary Research Centre in Brain Repair and the upgrading of Neuroimaging facilities are likely to strengthen our links with Clinical Neurosciences. Finally, the Lifespan Trust, which is responsible for health in the community in the Cambridge area, plans to develop a model Neurorehabilitation Unit for post-acute patients, with Wilson serving as Scientific Director. In the meantime, Wilson and Robertson have been able to compensate for the comparative paucity of local rehabilitation facilities by collaborating with colleagues elsewhere.
The Rehabilitation group have focused their attention on the assessment and treatment of brain injured people, attempting to analyse the role of neural plasticity and re-learning in their recovery. We have chosen to focus on cognitive deficits in memory and attention since these are both pervasive symptoms of brain damage which are not only intrinsically undesirable, but in addition are likely to interfere directly with any attempt at rehabilitation. They also represent topics that are of central theoretical concern to the Unit, with the result that there is very active collaboration between the Rehabilitation section and colleagues whose work is principally reported elsewhere. Consequently, the question of whether a particular piece of work is reported as part of the work on rehabilitation, or as research on attention or memory, is to some extent arbitrary.

Wilson's work can be divided into three related themes. The first of these concerns the detailed observation of the natural history of recovery from brain injury, which forms a basis for the second theme, namely that of developing better methods of assessment, which in turn feed into the third aspect, improving methods of treatment. Although we know a great deal about the sequelae of various types of brain damage, much previous work is based on cross-sectional studies, with very little attempt to monitor the same patient over a substantial period of time. Wilson has been carefully following up patients that were seen and studied in depth during her period at Rivermead Rehabilitation Centre, Oxford (1979-85). These studies have generated important baseline data regarding natural recovery of function in different cognitive domains, and such data will be crucial for the theoretical understanding of the processes underlying recovery and rehabilitation, and for the practical goal of optimising existing rehabilitation regimes.

One example of this detailed observation of natural history comes in a joint study with McLelland and Shiel at Southampton who used the careful observation of patients recovering from severe head injury as a basis for developing a standardised assessment usable from a state of coma through to the point at which other more conventional neuropsychological tests can be deployed. Particularly during the early stages, signs of recovery can be quite subtle, often resulting in the erroneous conclusion that the patient is not improving. Extensive longitudinal data on 88 patients have been collected, and are currently being analysed with a view to standardising scales that should prove helpful in patient monitoring and in prognosis. Two other lines of research have stemmed from this study. The first concerns the issue of whether learning can occur in coma; evidence suggests that it can, and that it may have implications for training methods that may minimise undesirable features such as contractures. The second has concerned a more detailed analysis of the confusional state known as post-traumatic amnesia (PTA). Length of PTA is commonly used as a predictor of probable outcome, but its accurate measurement is acknowledged to be problematic. Studies at Southampton and subsequently with Boismeir in the Department of Neurosurgery in Addenbrooke's are providing a much more detailed account of PTA, and it is hoped will lead to the development of better methods of assessment.

The developments in understanding of memory function at the APU and elsewhere have furthermore led to the construction of tests which incorporate facets of memory hitherto largely ignored in clinical memory testing, such as, for instance, prospective memory (remembering to do things at the right time). It is perhaps this combination of the use of everyday materials together with theoretical sophistication, which has led to the strongest correlations yet found between a clinical test (in this case the Rivermead Behavioural Memory Test, developed by Wilson et al) and real-life functional memory performance and adjustment.
Wilson and colleagues are pursuing work on this and other facets of the memory system in developing a specific test of prospective memory based on a paper-and-pencil text-based procedure, where the reader must obey instructions at some later point in the text. Preliminary studies in normal ageing on this test are promising. Wilson is also collaborating with Burgess in London and Alderman in Northampton in developing the BADS test (Behavioural Assessment of Dysexecutive Syndrome), influenced by Shallice's work on the frontal lobes and Baddeley's concept of working memory. Preliminary data are encouraging, and norms are now being developed.

In general, this approach to clinical testing which Wilson and colleagues have developed has been very successful and influential worldwide. For instance, the RBMT is now used in some 17 different countries in 11 languages, and versions for children and Down's syndrome have been developed.

In the area of treatment, much of Wilson's work has focused on a technique known as "errorless learning" in which the learning task is structured so as to avoid allowing the subject to make mistakes. This approach was developed in the animal learning field many years ago and has had some application to learning in mentally handicapped people. Its possible value in helping patients with memory deficits was suggested by Baddeley and Wilson, who argue that one of the major functions of explicit episodic memory is that it allows the organism to avoid errors. Amnesic patients who lack this memory system will therefore tend to be captured by their own erroneous responses, with each incorrect response making it more likely that the same error will be made again. An experiment designed to test this prediction indicated that amnesic patients were indeed particularly susceptible to disruption by earlier errors. A series of single case studies have demonstrated that errorless learning can be used with advantage across a range of patients and a range of tasks, from learning names to acquisition of the skill needed to programme a memory aid. It is equally clear, however, that the method does not always work in a simple and straightforward way, and it is proposed over the next few years to explore the theoretical basis of the errorless learning phenomenon at the same time as assessing its generality across patient groups and materials.

A particularly promising potential development over the next few years is the proposal that Wilson acts as Scientific Director to a new Rehabilitation Unit to be created by the Lifespan Trust. The Unit will deal with some 15 post-acute patients during the period between leaving hospital and returning to work. The Unit will be modelled on similar units in the USA and Denmark, and its success will be evaluated in collaboration with Stilwell of Warwick who is currently responsible for the assessment component of the Department of Health National Brain Injury Study. In addition to providing an ideal environment for developing and evaluating rehabilitation practice, the proposed centre offers a promising novel method of interaction between the Council and the NHS.

Robertson's principal concern is with attention and its disorders. His programme of work ties in closely with that of Duncan's group reported in the Attention section, and to a lesser extent with the attentional control components of the work on cognition and emotion. Robertson takes as his theoretical basis evidence from a range of sources, including PET scanning, that suggests that, in addition to the competitive processes underlying attention proposed by Duncan, there is a need to propose the existence of at least three supramodal attentional systems, for selection, sustained attention and spatial orientation respectively. One must also assume that deficits in these attentional systems strongly determine recovery of function following brain damage, given that much recovery of function depends on learning, which itself is heavily influenced by available attentional resources. Robertson has developed a clinical test of these different attentional processes, using the same principles as espoused by Wilson, namely of combining everyday materials with theoretical structure in clinical test development. The Test of Everyday Attention (TEA), using such materials as maps and telephone directories, has a factorial structure (based on a standardisation sample of 154 normals) which indeed supports the PET and other evidence about the existence of separable attentional control systems. Furthermore, in a population of stroke patients, performance on certain subtests of the TEA at two months post-stroke significantly predicted recovery of functional independence and everyday life abilities at eight months post-stroke. This result provides the first evidence of a theoretically-predicted link between attention and recovery of key physical functions following brain damage. The test is about to be published.
In the past, rehabilitation has all too often been based on the ad hoc application of pragmatic techniques aimed at relieving symptoms. Robertson, in collaboration with Murre, has proposed a theoretical framework for understanding the process of neuropsychological recovery, a framework that was strongly influenced by Duncan's work, and by Murre's interest in connectionist models of learning and memory. They suggest four ways in which performance may be disrupted, together with ways of alleviating this disruption.

(1) Overall level of sustained attention or arousal may be too low. They have developed metacognitive, self-instructional methods to train patients to increase sustained attention.

(2) A neural circuit may be malfunctioning, but may be supplemented by the activation of related compatible neural circuits.

(3) The operation of a malfunctional neural circuit may be exacerbated by competition from intact circuits. This inhibitory competition can be reduced by inducing even minimal responses from the impaired circuits.

(4) Subjects may have dysfunctional automatic control of action, but this can be supplemented by the development of conscious strategies.

Evidence for each of these sources of disruption and methods of treatment have been investigated and are being further explored and developed.

The right hemisphere-based sustained attention system may have particularly strong connections with the right hemisphere spatial orientation system which malfunctions in unilateral neglect. To demonstrate improvements in neglect in response to non-spatial sustained attention training would therefore provide a theoretically important piece of evidence as well as a clinically novel and useful strategy for training both unilateral neglect and sustained attention. In a study carried out jointly with colleagues in Stockholm and Southampton, patients were taught a self-instructional procedure that brings phasic arousal under verbal control. After five one-hour treatment sessions, subjects showed reduced neglect, and enhanced performance on the TEA. Another study using an attentional control procedure identical to one used by Teasdale with recovered depressed patients, also appears promising in producing performance enhancement in head injury patients together with self-reported improvement in everyday functioning. Both these lines of research will be followed up using more extensive group studies.

Robertson's work fits closely with Duncan's view of the role of competition and integration of neural circuits in attentional control. One example of this approach draws on the evidence that there are three separate but interconnected spatial circuits for body (personal) space, near-body (reaching) space and far (locomotor) space. Patients who show left neglect following a right hemisphere stroke can be taught to reduce the extent of the neglect by active movements of the left hand in left hemispace. Neglect is not diminished when the movement is passive, or when the left hand moves in right body space, or the right hand is active in either right or left body space. The nature of the action is also critical, with reaching and grasping causing more enhancement than simply pointing. Future work will examine these findings in more detail, in particular separating out neglect of stimuli versus neglect of output. A clinical trial will be carried out in collaboration with McMillan at the Wolfson Rehabilitation Centre, testing a Limb Activation Device which is located in the neglected field and emits a signal at irregular intervals to which the subject is required to respond with the neglected limb. Preliminary results are encouraging, but the planned more extensive study will give clearer evidence of the clinical viability of the method.

Work on reducing competition effects is closely related to the research of Duncan and Goodrich who are studying the phenomenon of extinction, whereby the detection of an object in the contralesional field can be prevented when a competing ipsilesional stimulus appears. Robertson has shown that the extinction phenomenon applies also in the motor domain, and that the neglect-reducing effects of left hand movements are abolished when the right hand is simultaneously moved, as the competition hypothesis would predict. Robertson has also shown that left neglect patients veer to the right when walking, but that this can be corrected by left hand movements while walking. This result can also be interpreted in terms of the competition hypothesis, namely that the cortical activation underlying the left hand movement, reduces competition from the undamaged left hemisphere of the brain. Robertson proposes to investigate the therapeutic implications of competition-reduction in other areas of perceptual deficit, as well as in non-spatial attentional problems.

In collaboration with Duncan, Robertson will study the problems of attentional control in patients suffering from frontal lobe damage. Duncan's concept of goal neglect will be investigated, and the general framework of competitive circuits used to generate and test methods of attentional control.
Work in this area has so far relied principally upon behavioural measures. The integrated competition model of rehabilitation proposed by Robertson and Murre does, however, make very specific predictions as to the underlying neural consequences of damage and treatment. Robertson and Duncan plan to take advantage of the planned development in PET scanning and fMRI in Cambridge to test the model more directly. For example, it is predicted that left-hand grasping in left visual neglect patients will have effects on bloodflow that extend substantially beyond those of simple motor activation, and that these effects will be abolished when the grasping is bilateral.

Finally, if our research in the area of rehabilitation is to have any impact, then it is essential that it is communicated to therapists who are in day-to-day contact with patients. Many of these will be occupational, speech or physiotherapists who would not be expected to read the psychological literature in which our results are published. For that reason it is essential that Wilson and Robertson devote at least some of their time and effort to dissemination. Both have great strengths in this area and are in constant demand for delivering workshops and addressing groups of clinicians and therapists, as well as receiving many requests for advice on either clinical or research issues. The fact that they are able and willing to provide this vital link with the clinical community bodes well for the future influence of this component of the Unit's work.

Programme 6: Cognition and Emotion

Over the last decade, the Unit has played an active role in linking cognitive and clinical psychology. We have been concerned with the way in which the emotions influence attention and memory, and how they are themselves modulated by such cognitive processes, in the case of both normal and clinical populations. Such a development had two aims, to provide a broader and more complete understanding of cognition, and to help guide the development of better methods of treating emotional disorders. We have made good progress in the first of these aims, which is already beginning to feed through to the difficult, but enormously important second goal.

Mathews and his colleagues in Cambridge and elsewhere have been concerned to understand the nature of the symptoms associated with anxiety, and have studied both patients suffering from anxiety disorders, and also members of the general public who volunteer to participate in studies of "worriers". Anxious patients rate the subjective likelihood of negative events as higher than non-anxious subjects, and are more likely to interpret ambiguous events in a negative way. They are also more distracted by such stimuli; for example when required to name the colour of the background on which a word is printed, anxious subjects are slowed down by the presence of a negative word such as cancer, an effect that occurs even when the word is presented so briefly that it cannot be consciously identified. When the anxiety disorder is treated, then the difference between patients and controls disappears.

Worriers resemble anxious patients in rating the likelihood of bad things happening to them as higher than do control subjects. They are also poorer at giving reasons why such calamities might be unlikely, but when they are successful in doing so, this does influence their estimate of future probability of such negative events, suggesting a possible line of treatment for excessive worry.

Although anxiety has marked and consistent effects on the likelihood of attending to negative events, anxious patients do not consistently show any greater tendency to recall negative items. Nor do they take any longer than controls to decide whether a given word such as cancer is pleasant or unpleasant. This pattern of results suggests an overall hypothesis. It is necessary for survival for an organism to be vigilant for potential sources of threat, and for the level of this vigilance to be increased at times of danger. Patients suffering from generalised anxiety demonstrate a chronically vigilant style in which there is a bias towards the automatic selection of stimuli of a potentially threatening nature, which in turn is likely to increase further the level of anxiety and maintain the bias. Although negative stimuli are preferentially encoded, processing such stimuli in depth may be inhibited, with the result that the bias does not show up in memory measures or in the results of conscious evaluation of words.

Future work will examine in much more detail the nature of the process of detection, contrasting judgements of the pleasantness of a briefly presented word with its direct identification, and studying the effects of a secondary task on the processing of pleasant and unpleasant words by anxious and control subjects. If the bias has its principal effect at a relatively automatic preattentive level, then it should show more strongly in judgements of pleasantness than in word naming; in contrast, a later locus would predict that such a bias would be more obvious when the subject's executive capacity is reduced by the requirement to perform a concurrent task. Other work will be concerned with methods of reducing the biasing effect.

Another line of work, particularly involving Dalgleish who is joining the Unit from the Institute of Psychiatry, will be concerned with the relationship between memory and anxiety. Failure to observe that anxious events are better recalled is less straightforward than might at first appear, and in post-traumatic stress disorder vivid memories of an extremely stressful event such as a rape or an accident can often be one of the more distressing symptoms. As will be clear from the section below, patients may develop a style of retrieval from memory that reduces the level of anxiety in the short term, possibly by consciously inhibiting the memory, but at the expense of longer term adaptation to the stressful event.

Work by Williams identified the tendency in parasuicide patients to have difficulty in retrieving specific autobiographical memories. When asked for specific recollections, such patients typically can only come up with rather general memories, such as "being with my girlfriend". This pattern is associated with hopelessness, with poor performance on a task requiring the production of effective solutions to practical problems, and with poor prognosis; such patients tended to be still depressed some seven months later. Overgeneral memory has now been observed in other centres elsewhere in the UK and abroad. In addition to research on depression and parasuicide, work by Brewin at the Institute of Psychiatry for example, has shown that overgeneral memory is found in women who have suffered sexual and physical abuse. Work by McNally at Harvard has found similar deficits in Vietnam veterans who have post-traumatic stress disorder. The pattern of data emerging from these studies suggests that overgeneral encoding and retrieval even of neutral events represents a long-term cognitive style which may arise early in the development of response to traumatic events. Such a style could be mimicked, however, by reduced working memory capacity at the time events are being retrieved. The combination of a long-term overgeneral memory style and reduced capacity at retrieval is particularly disabling. Williams' work is likely to be important since many cognitive treatments depend on the capacity to access specific memories in order to restructure the self-image, and as part of strategies for training and solving social problems. He has moved to UCNW BANGOR, but is continuing this line of research and continuing to collaborate with colleagues at the APU (see below).

Depression is a major disease, both in terms of the suffering it causes, and because of its economic consequences, in terms of absence from work, and impaired productivity while working. It forms the focus of Teasdale's work which has been strongly influenced by Barnard's Interacting Cognitive Subsystems (ICS) model. Collaborative work between Teasdale and Barnard has led to an important new development which incorporates the influence of emotion on the operation of the model. This has been published as a monograph which is already attracting a considerable amount of interest because of the way in which it provides a fruitful interface between cognitive psychology and issues of clinical practice.

Prior to ICS, attempts to relate cognition and emotion had been dominated by two views, one from a cognitive psychologist, Bower, and the other from a clinician, Beck. Bower conceptualised emotions as additional nodes within an associative net, while Beck linked emotion into a unitary propositional model of meaning. Teasdale and Barnard's model recognises two kinds of meaning, a "propositional" level that allows the manipulation of semantic representations and concepts, and a second more holistic "implicational" level. Affect is directly linked only to the more general level of representation. This implicational level is directly accessed by body state and sensory information, and is responsible for emotionally-held beliefs, which can of course be quite different from what one rationally knows. As a result of experience, we build up complex schemata or models of ourselves, and in the case of depression, such models may well be highly dysfunctional - "myself as worthless" for example. Convincing the subject rationally that this is not true may influence propositional representations, but only experience will create a similar change within more holistic implicational representations. The process of cognitive therapy can be seen as a means of achieving this deeper change.

One line of research has attempted to contrast this view that affect is linked to a schematic level of representation with the earlier associative network model of emotion. Subjects were required to complete statements about the world that are assumed to reflect dysfunctional models, where the dysfunctional completion word is not one that would be likely to be associated directly with a negative emotion. For example, depressed and control subjects were invited to complete sentences such as "Always to put others' interests before your own is a recipe for --". Depressed subjects tend to respond with a term like "happiness", whereas controls are more likely to produce "disaster" as a completion. This argues against a simple associative model of emotion, which would predict a negative word-completion response from depressed patients, and supports a model based on more complex schemata.

Another approach to teasing out the dysfunctional schemata associated with depression was produced by Barnard, in collaboration with Murray at the Winnicott Unit in a study concerned with the interaction of depressive mothers with their young children. The study analysed the spoken discourse of children playing with dolls in a family situation, using a psycholinguistic technique known as case grammar analysis. The study concentrated on one sub-unit of analysis, animate nouns, which can fulfil a number of case roles including agent, object, experiencer and dative. Children of depressed mothers were consistently less likely to refer to themselves in the agent role than were control children, and more of their self-references were phrased with negative form. Barnard is continuing this work as part of a programme grant to Murray studying the stability of the effect over time and its association with later dysfunction. He also plans to use the method to study depressive adults in different mood states, in particular contrasting the depressive, neutral and manic phases of manic depressive patients. The technique would appear to have considerable promise as an indirect measure of psychological dysfunction.

A major factor in the maintenance of depression appears to be the stream of ruminative negative thoughts that patients often find difficult to break out of. Teasdale has been concerned to understand this process, using both a working memory model, and subsequently ICS. He began by studying the intrusion of irrelevant thoughts in normal subjects observing that they were reduced by a concurrent memory load and appear to be associated with the operation of the central executive, rather than with the visual or verbal slave systems. Consistent with the notion that thought production and tasks making high demands on central executive functions compete for the same limited resources, the occurrence of such thoughts was associated with a reduction in randomness in a random generation task. Equally, the capacity of a task to inhibit them decreased as practice made the task more automatic. This pattern of results links the model with a range of attentional training techniques that have been found to be helpful in teaching patients to cope with pain and stress, and which form part of a study concerned with prevention of relapse in patients treated for depression.

While treatment of depressed patients with appropriate drugs can be very effective, residual symptoms may occur, and when they do so, as many as 78% of patients may relapse within the next nine months. As part of the MRC Neurosciences Approach to Human Health Initiative, Teasdale is collaborating with Paykel in Cambridge and Scott in Newcastle in a clinical trial of depressed patients with residual symptoms. This study compares long- and short-term outcome for patients treated either with drugs or with drugs supplemented by cognitive therapy. Both practical and theoretical aspects will be studied. While there is some evidence to suggest that cognitive therapy leads to a lower relapse rate than antidepressant drugs, it is of course expensive in terms of therapists' time. Teasdale, in collaboration with Williams and with Segal in Toronto, is concerned with attempting to develop and evaluate methods of attentional control which can be taught to the patient relatively easily and subsequently used with a minimum of further supervision. Similar measures are being evaluated by Robertson as a means of helping head injured and stroke patients to manage their attentional and emotional problems more effectively.

Such work on prevention of relapse will continue over the next five years, and will be supplemented by studies based on the ICS model. Teasdale will test the hypothesis that cognitive treatments of depression are essentially concerned with normalising dysfunctional thought patterns, and that to do so requires the patient to develop an alternative store of less depressogenic thought structures. A range of new techniques for measuring such structures and their change are currently under development.

Two new lines of research in this area will result from the appointment of new staff members. Dalgleish will be continuing to study the link between emotional factors and long-term memory, as mentioned above, but will in addition be investigating the effects of emotion on reasoning, beginning with simple syllogism problems which have been shown in normal subjects to be influenced by semantic as well as logical factors. He will begin by exploring whether syllogisms based on factors such as self-worth are particularly vulnerable to distortion in depressed patients. If so this should provide a tool for measuring both dysfunction and recovery, at the same time as giving a method for exploring the nature of the distortion.

Andrew Young who has recently joined the Unit from Durham has worked extensively on the perception and recognition of faces. He has become interested in the question of whether recognition of emotion involves a separate system from that required for processing personal identity. Evidence for such a view comes from the study of an amygdalotomy patient with a specific deficit in the capacity to identify emotional expression. Future work will follow up this and related cases, investigating whether failure to identify emotion from the face will be accompanied by a similar failure to make voice-based emotional judgements. Other studies will be concerned with the suggestion that patients with frontal lobe damage may have difficulty in judgements of facial emotion, and with the development of methods and material for the analysis of emotional judgements in normal subjects. Young suggests that a mismatch between the systems responsible for registering facial identity and facial emotion may be reflected in certain syndromes of delusional misidentification. Patients suffering from the Capgras syndrome believe that those surrounding them have been replaced by impostors, while patients suffering from Cotard delusion believe that they themselves are dead. One hypothesis is that, while recognising the identify of surrounding people, emotional reactions based on the face are distorted, a state of affairs which may induce a suspicious patient to suspect that others have been changed, and a depressed patient to attribute the change to himself. Collaborative work is underway to investigate this hypothesis.

It is clearly the case that facial expression and body posture represent an important channel for communicating socially and emotionally significant information. Research on human-computer interaction using simulated faces and voices has begun to demonstrate the extent to which a person's attitude and behaviour can be influenced by cues that are sufficiently subtle that the subject fails to detect them. One aspect of Barnard's work will be exploring this, and relating it to the work of Young on the perception of emotion by normal subjects and by neuropsychological and neuropsychiatric patients. This area of research is still in its infancy, but has considerable promise for understanding the problems of certain patients in interpreting their social environment, and in transmitting the social signals that, although implicit, are crucial for successful social behaviour.

Role of the APU in the Council's Scientific Strategy

According to the Council's Corporate Plan (p. 72), the Council's Strategy is as follows:

1. To encourage cognitive and developmental psychology, including modelling, to study the development and impairment of perception, memory, cognition and language.

In general, we would regard the area specified as being the heartland of cognitive psychology, and would regard the Unit as continuing to function as a centre of international excellence in the field. Our work is, of course, principally concerned with adult cognition, although it is becoming increasingly clear that work on children can have important implications for general cognitive theories.

2. To develop models of cognitive processes using computational and connectionist approaches that are consistent with clinical and experimental data, and which have a secure theoretical basis.

While computational modelling continues to be a strength in psychophysics (R Patterson) and motor control (Nimmo-Smith, Tresilian, Wing), computational and connectionist approaches to memory are so new that we had difficulty recruiting experienced and trained scientists. Instead we opted for a policy of collaboration (Shallice, Patterson) with existing centres of excellence in the US and developing our own strengths (Norris, Houghton, Shanks). While the Unit has undoubtedly suffered from the move of Shallice, Houghton and Shanks to UCL, we have been able to attract some excellent young scientists to non-established posts (Murre, Page), and to attract some first-rate research students with skills in this area. Our particular strength is in combining theoretical and computational skills with close attention to experimental and neuropsychological evidence. The area is flourishing at the Unit, but in my opinion would benefit from a further senior appointment to ensure its future stability.

3. To expand work on the cognitive rehabilitation of neurological patients building on an understanding of the way the brain functions and on opportunities for relearning.
Our newly developing Rehabilitation research group was set up with precisely this aim in mind. Neuropsychological rehabilitation does not have a strong existing research tradition, either nationally or internationally, but we have been fortunate in attracting two very strong rehabilitation research scientists to form the core of the group.

The turmoil resulting from the re-organisation of the NHS has limited the promised clinical local development in rehabilitation, but fortunately links with colleagues in rehabilitation groups elsewhere have allowed the programme to develop rapidly. A major local bonus has been the recent development of clinical neurosciences within the University, providing a much more hospitable and stimulating clinical environment than the Unit has experienced in the past. The Rehabilitation group has already established an international network of contacts and collaborations, and is receiving far more requests for working visits than can currently be accepted. New accommodation at Addenbrooke's is promised by the autumn, and with the exciting possibility of Wilson's involvement in the development of a model post-acute Rehabilitation Unit, the future looks extremely promising.

4. To develop suitable outcome measures for disorders involving loss of physical, mental and social functions, and use these to inform rehabilitation strategies.
Throughout the whole 50 years of its existence, the Unit has had an interest in using state-of-the-art techniques and concepts from cognitive psychology for the practical purpose of measuring human performance. In the early years, we were principally concerned with the assessment of the influence of environmental stress on military performance, whereas more recently our concerns have been with neuropsychological assessment, and with measuring the impact of emotional factors on cognitive functioning. As predicted in our last Progress Report, our interests are now shifting from assessment to treatment of both emotional and neuropsychological problems. We continue to develop new measurement techniques, publishing them in the standard journals, but are increasingly concerned with the development of standardised tests, for use within a clinical context.

Our clinical work has begun to extend from diagnostic measures to a concern with developing and assessing new treatment methods. In the area of rehabilitation, most of the work is still at the level of single cases or small groups. We are, however, increasingly being asked for advice on larger multi-centre trials, and look forward to the stage at which our new therapeutic techniques are ready for evaluation on this scale. We also anticipate close cooperation with the IRC in Brain Repair, once their work has reached the point at which clinical trials are feasible.

The Cognition and Emotion group have already reached this stage, with their involvement in a multi-centre trial concerned with the influence of pharmacological treatment and cognitive therapy on relapse in depressive patients, a study funded by the Council's Neurosciences Approach to Health programme. Mental health represents an area of enormous importance both in terms of human suffering and economic cost through absenteeism. The Unit is internationally recognised as a centre for excellence for research in this area, as reflected by the involvement of Teasdale in a US-funded project to develop methods of cognitive therapy that are less therapist-intensive.

Finally, there are two areas of the Council's Scientific Strategy in which the Unit is less involved than we would like, namely:

5. To bring together research on animal behaviour, in particular primates, with studies of normal and disordered brain function in humans.
We clearly do not plan to carry out animal work at the APU, but it is already clear from Duncan's continuing collaboration with Desimone at NIH on attention in monkeys, that the concepts and techniques of cognitive psychology are applicable to understanding the data obtained from neurophysiological studies of non-human primates. Duncan would welcome the opportunity of pursuing collaborative work more locally, and has already begun preliminary discussions with Rosalind Ridley whose Council-funded research on primates has recently moved to Cambridge.

6. To apply techniques of in vivo brain imaging to combine studies of cognitive function with anatomical mapping.

Brain imaging combined with good cognitive psychology can be a very powerful tool for investigating the neurobiology and psychology of cognitive function. Patterson and Duncan are already involved in some collaborative work involving PET scanning. However the potential for involvement of the Unit in collaborative work of this kind has hardly been touched, simply because of the lack of access to suitable imaging equipment and expertise. We are therefore delighted to note that such a facility is likely to develop locally in the near future, and expect to play a substantial role in its scientific programme. This does not feature prominently in our current proposals, simply because the development is not sufficiently far advanced for us to be able to present a fully articulated programme. Furthermore, since this would represent a relatively substantial development for the Unit, it is perhaps more appropriate that such detailed proposals should await the appointment of my successor.

The APU and Council Strategy

In writing their introduction, Directors are asked to relate the work of their Unit to the Council's Corporate Plan and Scientific Strategy, and to place it within a national and international context. I have chosen to use the Corporate Plan and Strategy documents as a framework for responding to this request, considering each of the objectives specified and highlighting the way in which the Unit's work contributes, commenting where appropriate on the national and international relevance of our work.

Corporate Plan Objectives

1. To strengthen the research base: We regard our principal role as that of contributing to the continued development of cognitive psychology. We believe that we have maintained the excellent international reputation that was reflected in comments on our last Progress Report, and trust that our peers will support this view on reading the present Progress Report.

2. To address the needs of users, in particular the Health Departments: Over the years, our applications have been progressively more health related, and this process has been accelerated by the development of the Rehabilitation Section. We believe that the attempt to apply cognitive psychology to clinical problems is proving effective in neuropsychology, neurorehabilitation and in our work on cognition and emotion. We believe that our work is innovative, sound and internationally respected. Meanwhile, we continue to interact with a range of other users, including the Home Office, the Ministry of Defence, and industry, where our work on information technology is of particular significance both nationally and internationally.

3. To create a good research environment: We are fortunate in having had the generous and long-term commitment of the Council which has allowed us to create an excellent environment for research in cognitive psychology. Evidence for this is provided by our capacity to recruit first-rate scientists, research scientists and support staff, by the number of visiting scientists who come to work at the Unit (an average of 10 per year) and their subsequent comments.

4. To provide good training and research development: We typically take two MRC students per year and one funded from other sources. Since the Unit began taking students, we have had over 50, of whom only three have failed to complete a PhD. By far the majority have moved on to academic or research posts, with information technology being the next most frequent career.

Even more important than our research student training is the post-doctoral opportunities afforded by the Council's three- to five-year short-term contracts. We are able to attract first-rate candidates, and to offer them the opportunity of broadening their research capability within an active and stimulating research environment, providing a strong foundation for a subsequent academic or research career. Of the 28 holders of short-term scientific posts at the Unit over the last decade, 16 are currently in tenured academic posts, 3 in research posts, 2 in industry and 2 in clinical posts.

5. To work with other research sponsors: While I assume that this is principally aimed at formal Council links, I can report that the APU has been making its contribution by attracting funding from a very wide range of sponsors, ranging from international programmes such as the Human Frontier Program, through European programmes such as Esprit, to more national sources of funding such as the Ministry of Defence, the Home Office and medical charities.

6. To disseminate and apply research: Much of our applied work leads not to a specific product, but rather to increased understanding, which will allow existing activities to be pursued more effectively. Hence our work in information technology depends for its effectiveness on its being conveyed to software engineers, while our rehabilitation research will only be effective if it is conveyed to therapists. In both these areas we have a high international profile, and intend to continue to devote a substantial proportion of our efforts to ensuring that our work is applied.

More locally, members of the Unit have been involved in teaching in a range of Cambridge departments, including Experimental Psychology, Anatomy, Physiology and Engineering, with occasional courses taught much further afield.

7. International collaboration: The extent of this is obvious from our publication list, which reflects many publications in collaboration with colleagues from a wide range of countries.

8. To increase the public awareness of science: We are fortunate in that psychology is a topic that is of considerable intrinsic interest to the general public. We encourage this growing interest by a consistent willingness to present our work through the media, including television (some 16 programmes and 4 news items over the last five years), radio (12 programmes) and the press, both general (35) and more specialised (47), in addition to Robertson's regular column in the BMJ, and frequent articles in The Times. The press coverage tends to be national, with television being rather more international with the Unit featuring prominently in science programmes made by Dutch, Canadian and Japanese, as well as British companies.

One index of the Unit's commitment to the dissemination of science is the continual stream of requests for information, advice and assistance of varying forms, from the media, from Government, public and private institutions, and from fellow scientists in various disciplines, in this country and around the world. Members of staff are asked to record these outside contacts systematically and a memo, reporting all known contacts is distributed to staff weekly. A summary of these contacts over the last five years is given below.

1989 1990 1991 1992 1993

1. Requests for information/advice: 238 268 216 297 262
2. Invitations to visit: 14 38 18 29 22
3. Requests to visit APU: 49 43 36 57 70
4. Invitations to make a presentation: 178 131 126 226 210
5. Invitations to author/edit books, etc: 23 13 13 25 33
6. Requests for extended commitment of effort: 35 36 41 52 46
Totals 537 529 450 686 643

The reporting of contacts is certainly incomplete, and the above summary statistics therefore merely index the minimum demand on Unit resources. However, meeting even this level of demand represents a considerable expenditure of time and effort by Unit staff, over and above their normal commitments to research and scientific publication. We do, however, regard it as an important and potentially very productive part of our work.

9. To use resources effectively: The previous Visiting Subcommittee, and in particular its foreign members, commented very positively on our efficient use of resources. Over the last five years, we believe that our flexible management structure, making greater use of two assistant directors, has allowed us to be even more cost-effective. A rough indication is provided by the fact that our publication rate has risen by about 14%, with no increase in number of established posts. A simple publication count does not, of course provide an adequate assessment of quality. For this we rely on our referees.

Concluding Statement

I would like to thank Council for its sustained support for the APU during the fifty years of its life, and for entrusting to me its directorship over the last two decades. I trust that our referees will agree that I am returning it in good heart, ready to begin tackling the challenges and opportunities of the next fifty.

Other sections in the 1990-1994 report

1. SUMMARY

2. PERCEPT-ACTION

3. ATTENT-COG CONTROL

4. MEMORY

5. LANGUAGE

6. REHAB

7. COGNITION-EMOTION