CHAPTER ONE
Introduction: Methodologies and Models in the
Study of Executive Function
Patrick Rabbitt Age and Cognitive Performance Research Centre,
University of Manchester, U.K.
A common theme of all contributors to this book is that, because tests of executive function have low test-retest reliability and uncertain validity, it has been hard to draw clear empirical distinctions between âexecutiveâ and ânon-executiveâ tasks. It has also proved difficult to relate particular behaviours which have become regarded as being paradigmatic exemplars of âexecutiveâ function to specific neuroanatomical areas or neurophysiological systems. We argue that some of these difficulties occur because executive behaviour has been discussed, simultaneously, at more than one level of description. Accounts of purposeful and conscious behaviour, in the language of everyday subjective experience, map uneasily onto clinical taxonomies of the modes of dysexecutive function observed in brain-damaged patients. Performance indices empirically measured in laboratory tasks are often treated as being directly equivalent to the hypothetical system performance characteristics specified in functional models of working memory and selective attention. As a result, it has passed unrecognised that hypothetical components of âexecutive behaviour,â such as âinhibition,â âplanning,â âmonitoringâ and âcontrolâ which are, in fact, simply descriptions of task demands, may have very poor construct validity because although these demands appear logically different they can be met by identical production system architectures. Confusions have also arisen because both task performance indices and system performance characteristics have been treated as equivalent to statistical constructs, derived to explain individual differ ences in performance on intelligence tests, such as Spearmanâs (1927) index gf (see review by Rabbitt, 1996).
Individually and collectively, the contributions to this book show why it is unsurprising that all empirical psychologists find it very difficult and some philosophers (Fodor, 1983) think it impossible to give a satisfactory functional account of the central executive system (CE). Classical descriptions of executive impairment associated with prefrontal lesions have immediacy and richness, and are strikingly similar to definitions of âwilledâ âpurposefulâ or âvoluntaryâ behaviour that have preoccupied philosophers and theologians for more than two millennia. For example, contemporary catalogues of the functions of the hypothetical âcentral executiveâ are strikingly similar to the formal criteria for commission of mortal sin given by Roman Catholic theologians: Beasts are soulless automata, and so always blameless. Humans also do not transgress while performing even very complex vile acts, provided that they do so involuntarily and automatically without conscious awareness of themselves as perpetrating agents. The minimal functional processes involved in the commission of a mortal sin are awareness of the self as the intending perpetrator of the act; recognition of the unpleasant implications of the act for others by possession of a theory of mind; recognition of its moral repulsiveness by possession of a theory of the mind of God; an ability simultaneously to represent alternative acts and their possible outcomes in working memory in order efficiently to choose between them; conscious formulation of a well-articulated plan to perform the act successfully; self-initiation and execution of sequences of appropriate actions to consummate this plan during which recognition of personal culpability is maintained by continuous monitoring; recognition of attainment of the vile goal state and an intention to use what has been learned in its pursuit to perform it again if opportunity occurs.
Clearly only the central executive (CE) can sin. Do we yet have functional models that improve on this already comprehensive theological description?
Both theologians and neuropsychologists seem to be describing the same ghost, and it is natural that a class of investigators whom Burgess (this book) labels âultra-cognitive neuropsychologistsâ should try to deduce its properties from the results of selective damage to the machine that it haunts. An impressive, but unlikely, discovery would be that whenever a particular, functionally circumscribed, brain system is damaged the patient becomes capable only of brutish automaticity. An even more interesting finding would be that different lesions disable one, but not others, of a set of logically distinct and tightly specifiable functions such as âplanning,â âsequence initiation,â âmonitoringâ or âinhibition of alternative responses.â This would reveal the âdouble disassociationsâ of effects prized by âultra-cognitivists.â Unfortunately all the chapters in this collection, and a voluminous background literature excellently reviewed by Fuster (1989), Reitan & Wolfson (1994) and others, point out that attempts to relate any of the capabilities that are agreed to be most characteristic of âexecutiveâ behaviour to particular brain systems or even to diffuse changes in volume of brain tissue have been unsuccessful (see especially Foster et al. this book). Failures to offer convincing neurophysiological descriptions of the bases of most complex cognitive skills may, as Burgess (this book) clearly anticipates, sap our faith in the methodology of âultra-cognitive neuropsychology.â It has been suggested that some of these difficulties may be avoided if the problem is restructured as a search for empirical distinctions between âexecutiveâ and âdysexecutiveâ behaviours and the question of where, or what, their neurophysiological bases may be is deliberately, and indefinitely, postponed (e.g. Baddeley & Wilson, 1988). While this ploy frees us to carry out behavioural experiments without worrying about how their results bear on what we know about the brain, it also deprives us of an important source of evidence against which to check the plausibility of our models. In the difficult task of understanding executive function we need all the help that neuropsychology can offer.
Nevertheless, attempts to distinguish behaviourally between âexecutiveâ (E) and ânon-executiveâ (NE) functions at least force us to move beyond the richly allusive language of everyday subjective experience and try to develop a functional taxonomy distinguishing performances, skills, or behaviours that are characteristic of executive function from those that are not. Collectively, our contributors generate an exhaustive list:
First, executive control is necessary to deal with novel tasks that require us to formulate a goal, to plan, and to choose between alternative sequences of behaviour to reach this goal, to compare these plans in respect of their relative probabilities of success and their relative efficiency in attaining the chosen goal, to initiate the plan selected and to carry it through, amending it as necessary, until it is successful or until impending failure is recognised. It is crucial to note that this prospectus does not merely boil down to a statement that executively controlled behaviour is much more complex than NE behaviour. Even choices between very complex behaviour sequences (aka plans) can be carried out âautomaticallyâ under NE control. Examples of this might be skilled drivers negotiating complex traffic while carrying on non-trivial conversations, or chess grand masters rapidly selecting among extremely difficult lines of play by recognising generically familiar positions during âlightning chessâ tournaments. The key distinction seems to be between situations in which a person must, for the first time, recognise, evaluate, and choose among a variety of alternative options and those in which a single effective behaviour sequence, which has been previously identified, and instantiated by practice, is run off without the need to propose and evaluate alternatives.
In contrast to E behaviour, NE behaviour tends to be initiated by, and to continue automatically in response to, changes in environmental input. A possible way to envisage this is that even very complex NE behaviours are externally driven and controlled by sequences of environmental events or by previously learned âplansâ or âprogramsâ held in long-term memory. In contrast, E behaviours can be initiated and controlled independently of environmental input, and can retain the adaptive flexibility to rescue plans even when the environment does not behave as anticipated and no guidance is available from previous experience.
It follows that, should the rules suddenly change, as in the Wisconsin Card-sorting Task or its variants, âexecutiveâ function will be necessary to appreciate from feedback that a change has occurred and to formulate and test new plans or rules until successful. Some authors stress that part of this process is the âinhibitionâ or âsuppressionâ of previously used and so habitual rules in order to project, select, and use alternatives.
Second, a study reported by Burgess (this book) shows that deliberate retrieval of structured information from memory in order to answer a question or to elaborate a partial description may involve the conscious, and reportable, formation and prosecution of well-articulated memory âsearch plans.â Burgess points out that CE control is not only necessary to manage transactions with the external world but also with the âinternal information environmentâ of long-term memory. Thus CE control extends beyond the current internal or external information environment to restructure interpretation of the past as well as to attempt active control of the future. Burgessâs excellent insight is timely because it allows us to distinguish between NE retrieval of information from memory in response to learned associations and environmental cues and active, âstrategicâ or âplannedâ memory search carried out under voluntary control. Closely related suggestions by Phillips (this book) are that individuals with inefficient executive function apparently generate categories of words by simple associations between successive items, or by empirically inscrutable random processes, while the better performance of individuals with intact executive function may be partly attributable to their discovery and use of memory search strategies. These suggestions of a strong link between CE control and efficiency of organised recall of material from memory are particularly interesting in view of Mayesâ and Daumâs (this book) neat demonstrations that lesions in prefrontal areas that are often associated with losses of executive function also impair learning, recall, and recognition. They raise the further question whether prefrontal lesions may impair strategic control of recall and recognition to a greater extent than do the lesions of the mediotemporal lobes and of the hippocampus and limbic system that are more usually associated with the âclassicâ amnesic syndrome. In this case lesions might disassociate strategic control of memory from memory efficiency.
Third, E functions are necessary to initiate new sequences of behaviour and also to interrupt other ongoing sequences of responses in order to do so. A logically related property is that they can suppress, or inhibit and replace, automatic and habitual responses with task-appropriate responses. E functions can also check involuntary perseveration by appropriately âswitchingâ attention to new sources of information. In this sense âinhibitionâ of habitual responses in order to make other responses, or âswitchingâ between one sequence of responses or one aspect of the environment and another, appear to be logically, and operationally, very similar concepts (see Lowe & Rabbit this volume, and Baddeley, Maylor, & Godden, in press). Whether by strategically switching attention from one source to another over time, or by simultaneously and strategically allocating attentional resources to more than one input, E processes control the allocation of attention, particularly in complex tasks in which a variety of different demands must be met simultaneously.
Fourth, E functions are necessary to prevent responses that are inappropriate in context. This rather poorly articulated aspect of dysexecutive function must be logically distinguished from the suppression of habitual or unnecessary responses. It seems first to have been documented by clinical observations of inappropriate social behaviour following frontal lesions. Such lapses of social judgment may perhaps have logical analogues in empirical findings, such as the increases in numbers of bizarre responses that patients with frontal lesions make to questions in the Shallice Cognitive Estimates Test (Shallice & Evans, 1978) and in reports, such as that by Parkin (this book) that older, and so potentially more âexecutively challenged,â individuals make markedly more false-positives in recognition memory experiments than do younger adults. It is hard to suggest any single, comprehensive description of the cause of these and similar lapses but âresponding without taking all of the available information into consideration,â âresponding without considering alternatives,â âresponding on the basis of insufficient informationâ or âfailing to monitor for plausibilityâ may all capture something that they have in common. It is also difficult to disassociate these ideas from findings such as Parkinâs of an increase in false-positives that may, generically, also be described as âloss of strategic control of memory.â Note also that these problems of response production can also be logically, if not functionally, distinguished from failures to monitor responses in order to detect whether they are âcorrectâ or âwrongâ in the sense that they have or have not succeeded in achieving a particular intention. They seem, rather, to be failures to judge when an intention, as well as the response in which it results, is implausible or unacceptable in a particular context.
Fifth, dual-task performance has become a test bed for exploration of functions of planning and control because the simultaneous performance of two tasks may require strategic allocation of attention and synchronisation of responses in order to service both. This may be achieved by rapidly switching from one task to the other and back again as their relative demands fluctuate over time, or by other forms of resource management such as are envisaged in models of âM spaceâ (Case, 1992; Case & Sandieson, 1992) or quantified in terms of the index mu (see Baddeley, della Salla et al. this book). Simultaneous performance of two E tasks is seen as being especially demanding because, in effect, it may involve coping with all of the demands intrinsic to task 1 and task 2 and also, in addition, with the demand of strategically controlling allocation of attention between tasks. As many elegant experiments by Baddeley (1986) have shown, mutual interference between simultaneous tasks is particularly severe if they make simultaneous demands on the same sensory modality or representation system. In contrast, when two tasks are both under NE control, even if both are complex they may be simultaneously performed with less mutual interference and with less need for strategic scheduling of their conflicting demands. However, as recent work by Bourke, Duncan, and Nimmo-Smith (1993) suggests, there may always be some loss of efficiency in dual-task performance.
Sixth, while carrying out plans to cope with novel demands, E functions are necessary to monitor performance in order to detect and correct errors, to alter plans when it becomes clear that they are becoming unlikely to succeed, or to recognise opportunities for new and more desirable goals and to formulate, select among, initiate, and execute new plans to attain them. Even complex sequences of NE behaviours do not seem to be monitored at this level. This distinction requires elaboration because it has long been known that error correction in simple tasks such as serial choice reaction time is very fast and âautomaticâ (Rabbitt, 1968), so that continuous monitoring for discrepancies between intentions and acts does occur, but at speeds (e.g. < 40ms) which are much too fast to be attributable to âconsciousâ, or âexecutiveâ intervention. In contrast, E monitoring is envisaged as a higher level activity that not only detects when actions have not been completed as planned, but can also predict when failure is imminent, or an even more attractive goal may be missed, unless a plan, which is currently being carried out precisely as intended, can be rapidly changed.
Seventh, Manly and Robertson (this book) suggest that another characteristic of E control is that it can enable attention to be sustained continuously over long periods. Among other benefits, this allows prediction of outcomes of long, complex sequences of events. This property seems to have first been suggested by Wilkins and Shallice (1987). Perhaps an example of this property of E behaviour is Duncanâs (1995) clever demonstration that dysexecutive individuals show âgoal neglectâ in a novel task in which they are, at random intervals, signalled either to âswitchâ from one pattern of responding to another or to âstickâ with the response pattern in which they are currently engaged. Another example might be the âliftâ or âelevatorâ monitoring task, designed by Robertson and his associates, in which participants watch a changing display which signals a series of upward and downward movements of a hypothetical conveyor. By monitoring these changes patients have to identify the floor that the conveyor has currently reached. We have already discussed how E control is necessary to formulate plans. Manly and Robertsonâs observations suggest that we should add to this the ability to hold intended sequences of actions in mind over long periods of time so as to carry them out as soon as appropriate circumstances occur. Another general description of such activities might be âr...