Dissociating executive mechanisms of task control ... - Semantic Scholar

Brain (1998), 121, 815–842

Dissociating executive mechanisms of task control following frontal lobe damage and Parkinson’s disease R. D. Rogers,1 B. J. Sahakian,2 J. R. Hodges,3 C. E. Polkey,4 C. Kennard5 and T. W. Robbins1 Departments of 1Experimental Psychology, 2Psychiatry and 3Neurology, University of Cambridge, and 4King’s College Neuroscience Centre and 5Division of Neuroscience and Psychological Medicine, Imperial College of Science, Technology and Medicine, London, UK

Correspondence to: Robert Rogers, Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, UK

Summary Twelve patients with focal damage of the frontal cortex and 12 patients with mild, medicated, early stage Parkinson’s disease switched between letter- and digitnaming tasks on every second trial of a task-switching paradigm. Compared with age- and IQ-matched control performance, patients with left-sided, but not right-sided, frontal damage exhibited markedly increased time costs associated with these predictable switches only when there was a general incidence of interference or ‘crosstalk’ between the tasks, and particularly so when the available task cues were relatively weak and arbitrary. The same patients also showed evidence of an increased sensitivity to the facilitatory and inhibitory effects of previous processing, when required to switch between tasks. Both groups of patients (with left- or right-sided frontal damage) exhibited slow, disorganized performance early in practice. In contrast to these frontal effects, the

Parkinson’s disease patients showed little indication of larger time costs of task switches but they did show progressive increases in the error costs, while age- and IQ-matched control subjects showed reductions. We propose that while both left and right frontal cortical areas are involved in the organization of cognitive and motor processes in situations involving novel task demands, only the left frontal cortex is involved in the dynamic reconfiguring between already-established tasksets, and specifically, that it is the site of an executive mechanism responsible for the modulation of exogenous task-set activity. Finally, dopaminergic transmission, along the nigrostriatal pathway, may be implicated in sustaining various cognitive and motor processes over prolonged periods, including the operation of those executive control mechanisms that accomplish reconfiguring between task-sets.

Keywords: frontal cortex; Parkinson’s disease; task-set; executive control processes Abbreviations: ANOVA 5 analysis of variance; LF 5 left-sided frontal damage; RF 5 right-sided frontal damage; RT 5 reaction time; WCST 5 Wisconsin Card Sorting Test

Introduction Several lines of evidence indicate that the moment-bymoment control of behaviour depends on the integrity of the frontal cortex. Evidence from the earliest studies with experimental monkeys (Bianchi, 1922) to common clinical experience with human patients suggests that damage to the frontal cortex results in a general disruption of the organization and monitoring of goal-directed action (Luria, 1966; for review, see Duncan, 1986; Schwartz et al., 1991). More specific manifestations of a loss of control of behaviour, all widely cited features of the so-called ‘frontal syndrome’, include a tendency to perseverative behaviour (Luria et al., 1964; Nelson, 1976; Sandson and Albert, 1987), difficulties with response suppression (Perret, 1974; Drewe, 1975; © Oxford University Press 1998

Burgess and Shallice, 1996), increased distractibility (Knight, 1984) and an inability to plan and co-ordinate sequences of actions for the satisfaction of goals that are not immediately attainable (Shallice, 1982; Owen et al., 1990). Phenomena such as these have suggested to researchers that the frontal cortex must be the site of a controlling or ‘executive’ system (e.g. Shallice, 1988; Fuster, 1989) responsible for the highest levels of behavioural organization. However, the architecture of this system tends to be specified in terms of the functional requirements of control systems in general (e.g. being able to handle ‘error correcting’ or ‘trouble-shooting’ situations) rather than in terms of elucidated cognitive mechanisms. Perhaps the phenomenon that is most central to questions

816

R. D. Rogers et al.

about how control mechanisms in the frontal cortex achieve behavioural control, and one that has received a good deal of experimental attention, is the inflexibility of thought and behaviour that often results following damage to the frontal cortex. In the clinic, this is most often demonstrated using set- or concept-formation tasks such as the Wisconsin Card Sorting Test (WCST) (Grant and Berg, 1948). In this task, the patient is shown a series of cards bearing patterns that differ in their colour, number and shape, and asked to sort the cards according to a rule, known only to the experimenter, that must be inferred from trial-by-trial feedback given after each response. Many studies have demonstrated that when the experimenter suddenly changes the rule, patients with frontal lesions are not able to shift to a new rule as efficiently as control subjects (Milner, 1963, 1964; Drew, 1974; Robinson et al., 1980). Furthermore, following subtle modifications to the task itself, it has been shown that this deficit is probably associated with perseverative behaviour whereby patients with frontal lesions continue to sort the cards by the previously relevant but now-irrelevant rule (Nelson, 1976). Thus, the deficits exhibited by these patients in shifting from one rule to another in the context of the WCST seems to indicate some sort of failure of the controlling, presumably neocortical, system to reorganize the patients’ behaviour so as to match more appropriately current behavioural objectives. Whilst tasks like the WCST and others (e.g. Cicerone et al., 1983; Delis et al., 1992) undoubtedly reveal something important about the role of the frontal cortex in sustaining the normal flexibility of behaviour, several difficulties of interpretation remain. One problem is that the specificity of the WCST for frontal damage is unclear. Several studies have failed to show impairments in patients with frontal lobe damage (Eslinger and Damasio, 1985; Heck and Bryer, 1986; Grafman et al., 1990; Anderson et al., 1991) while other studies have shown deficits in neurological patients with known non-frontal damage (Teuber et al., 1951; Hermann et al., 1988; Anderson et al., 1991). A second problem is that the psychological character of the frontal deficit in the WCST remains controversial. While it is tempting to interpret the frontal deficit as reflecting the failure of the controlling system to reorganize some component of the cognitive system, e.g. to override an attentional bias towards a previously reinforced attribute of the stimuli (Owen et al., 1991; Dias et al., 1996), the complexity of the task means that other interpretations remain possible, including a failure to keep task-relevant information available, ‘on-line’, in working memory (see Goldman-Rakic, 1987, 1991; Berman et al., 1995). However, the third and most serious reason why deficits on these tasks may not arise purely from frontal problems with dynamic behavioural organization is that they are all tasks of acquisition in which the patient is initially unaware of the next behaviour to adopt. Specifically, successful performance on the WCST depends not only on the capacity to shift from one rule to another when the current sorting rule changes, but also on the identification,

through trial and error, of which rule is now in effect. Thus, these tasks may be telling us as much about the problemsolving capacities of patients with frontal lesions as about the neural basis of specific executive mechanisms responsible for behavioural control. This brief survey of the psychological nature of conceptformation tasks highlights a distinction between those executive functions that are involved in the induction of strategic and problem-solving processes and that are activated by situations requiring the organization of new and untried behavioural tacks, and other executive functions that are required to organize a subject’s cognitive and motor resources to give expression to different behaviours according to environmental demands. This latter phenomenon is the notion of task-set control which has been receiving revived attention in the field of cognitive psychology (Allport et al., 1994; Rogers and Monsell, 1995; Rubinstein et al., 1998).

Task-set The adoption of appropriate task-sets underlies effective performance of a whole range of activities, ranging from simple reactive tasks such as naming the ink colour of a Stroop stimulus in the cognitive laboratory to more ‘everyday’, composite, tasks such as making a cup of tea, or purchasing a series of items listed on a shopping list. [These latter tasks can be thought of consisting of chains or hierarchies of simpler tasks such as ‘putting teabag into teapot’ or ‘inspecting next item on shopping list’; see Monsell (1996) for a discussion of ‘single-step’ versus ‘multi-step’ tasks.] However, to perform any task at all, the brain must somehow be able to assemble and link together a set of perceptual, cognitive and response processes that, once arranged in a suitable configuration, will accomplish the behavioural requirements of the moment (Rogers and Monsell, 1995). Thus, processes of sensory analysis appropriate for the currently receiving sensory modality need to be linked to various stimulusidentification and -recognition processes, that in turn need to be linked with other processes responsible for accomplishing relevant decisions about the incoming stimulus. Finally, these decisional processes have to be connected to various response selection and output processes that enable a response to be emitted by the appropriate effector system. In everyday life, our brains constantly reconfigure between such task-sets as ever changing, sometimes new and unpredicted, behavioural demands are placed upon us. It is this capacity to reconfigure between different task-sets moment-by-moment, our ‘task control’, that confers flexibility on both cognition and its behavioural expression. Currently, the functional characteristics of the executive control processes that achieve dynamic reconfigurations of task-set remain almost entirely unknown. However, a few studies addressing the relationship between the frontal syndrome and the neuropsychology of basal ganglia disorders suggest that the neural basis of task control may go beyond the suspected involvement of frontal cortical systems and

Frontostriatal mechanisms of task control involve subcortical structures neurotransmitter modulation.

and

their

associated

Task-set and frontostriatal loops The recent progress made in understanding the functions of prefrontal cortex has taken account of its co-ordinated communication with inter-connected brain structures (e.g. Goldman-Rakic, 1987). In particular, it appears that it is the interchange of information through a number of ‘functional loops’ linking frontal cortex, basal ganglia structures and motor cortex (Alexander et al., 1986; Joel and Weiner, 1994; Parent and Hazrati, 1995) that mediate performance on a range of tasks that tap so-called ‘executive function’. For example, experiments using PET technology have shown that acquisition of sequences of motor responses activates both frontal and basal ganglia structures (Jenkins et al., 1994). Consistent with this frontostriatal hypothesis of executive function, corresponding deficits on concept formation tests have now been demonstrated in patients suffering from various forms of striatal dysfunction. Both Parkinson’s disease patients and Huntington’s disease patients show deficits on the WCST (Bowen et al., 1975; Josiassen et al., 1983; Lees and Smith, 1983; Cools et al., 1984) and a range of other concept-formation tasks (see Taylor et al., 1986; Downes et al., 1989; Channon et al., 1993; Owen et al., 1993; Lawrence et al., 1996). Thus, to the extent that these tasks do exercise the control of task-set in requiring shifts of a rule, albeit following some sort of problem-solving or hypothesis-testing process, we should expect difficulties in task-set reconfiguration for patients with some form of striatal pathology. In fact, a difficulty with the executive control of task-set has been widely cited as one of the central cognitive changes in Parkinson’s disease. For example, Cools et al. (1984) have claimed, on the basis of impaired performance on the WCST and assorted motor sequencing tasks, that Parkinson’s disease causes a generalized deficit in ‘shifting-aptitude’. However, this position faces serious difficulties. Several studies, involving explicit experimenter-cued shifts between tasks as diverse as spatial discriminations over stimuli with and without mental rotation (Brown and Marsden, 1988a), the application of the Odd-Man-Out rule over letters and shapes (Flowers and Robertson, 1985) and various verbal fluency categories (Downes et al., 1993) have all failed to provide evidence that Parkinson’s disease patients are impaired at accomplishing these shifts compared with control subjects. In response to such findings, Flowers and Robertson (1985) and Robertson and Flowers (1990) have argued that the deficit lies, not so much in switching between cognitive and motor sets, but in maintaining or protecting them in the face of stimulus interference. Thus, one possibility is that dopaminergic dysfunction in the striatum results in the failure of executive control processes to maintain effectively the links, or ‘bindings’, between the various cognitive modules that together constitute a task-set. However, once again,

817

problems with set maintenance are not always found in Parkinson’s disease (Brown and Marsden, 1988b; Richards et al., 1993). Under what conditions, then, does Parkinson’s disease impair the control of task-set? Brown and Marsden (1988b) have argued that Parkinson’s disease patients are impaired in switches between tasks only in the absence of cues indicating the now-relevant stimulus attribute, and have suggested that Parkinson’s disease produces a difficulty in responding on the basis of ‘internal’ as opposed to ‘external’ cues. Taylor et al. (1986) expressed a similar idea in suggesting that Parkinson’s disease produces deficits in any task requiring a degree of ‘self-directed planning’. The problem with characterizing the deficit in this way is that the distinction between internal and external cues cannot be separated from the more natural continuum of cue strength (Robbins and Brown, 1990; Brown et al., 1993). The point here is that any task cue presented in close temporal and spatial proximity to a stimulus will naturally have a greater impact on the subject’s actual performance than the inevitably weaker task cues that might be held internally, say, in working memory. In summary, the consequences of Parkinson’s disease for task-control mechanisms remain unresolved. Some of the inconsistencies in the literature probably stem from uncontrolled variability in patient selection in terms of medication status and progress of the disease. However, it is important to note that experimental procedures that measure task control necessarily involve control processes beyond those required for accomplishing some sort of ‘shift’ or ‘switch’. To perform these procedures, subjects need to be able to organize a range of, sometimes incompatible, cognitive and motor processes, and to co-ordinate their activity in ongoing performance. These co-ordinating processes are naturally subsumed under the heading of ‘executive function’ and should be expected to be disrupted by frontal and subcortical dysfunction. Thus, it is not surprising that situations involving simultaneous activities of one kind or another, and which presumably place the cognitive system’s organizational capabilities under particular stress, have been shown to be sensitive to Parkinson’s disease. Examples include concurrent choice reaction time (RT) tasks over different sensory modalities (Malapani et al., 1994), divided attention between different attributes of a visual stimulus (Brown and Marsden, 1991; Filoteo et al., 1994) and the simultaneous programming of motor actions (Talland and Schwab, 1964; Benecke et al., 1986).

Measuring task-set reconfiguration: the notion of ‘switch cost’ In many ways, the experimental procedures used to examine task-set in Parkinson’s disease patients are more appropriate for examining dynamic task control than the conceptformation tasks, such as the WCST, originally developed for use with patients with frontal lesions. The typical procedure

818

R. D. Rogers et al.

involves asking subjects to shift repeatedly between tasks on successive blocks of trials with the measures of interest being the increase in RT and error rates caused by these shifts over the first few trials of each block, and the rate at which performance improves over the ensuing trials (e.g. Brown and Marsden, 1988b). Although this approach can be informative, a more direct method is to compare performance on trials requiring a switch of task with that on trials requiring no such switch. The difference in speed and accuracy of responding on the switch trials compared with the non-switch trials can be thought of as a ‘switch cost’, reflecting some extra processing burden that subjects must bear when switching from one task to another. Several experimental procedures have been devised to capture this basic comparison (Jersild, 1927; Allport et al., 1994; Rogers and Monsell, 1995), and investigating the properties of these switch costs is now an important preoccupation in cognitive psychology. The experiment reported in this paper uses the ‘alternatingruns’ procedure presented in Rogers and Monsell (1995). This procedure requires a switch of task on every second trial of a block, so that the sequence of tasks is as follows: Task A, Task A, Task B, Task B, Task A, Task A, Task B, etc. The crucial comparison is then between trials requiring a switch (Task A preceded by Task B, Task B preceded by Task A) and trials not requiring a switch (Task A preceded by Task A, Task B preceded by Task B). This within-block comparison avoids two difficulties inherent in the more traditional approach (Jersild, 1927), in which performance during ‘mixed’ blocks of trials requiring immediate alternations of task (Task A, Task B, Task A, Task B, etc.) is compared with performance during ‘pure’ blocks of trials requiring just one task by itself (Task A, Task A, Task A, etc. or Task B, Task B, Task B, etc.). First, mixed blocks of trials are experienced by subjects as more difficult than pure blocks so that a comparison between them may be confounded by between-block differences in subject effort or arousal. Secondly, subjects performing the mixed blocks have to keep two task-sets active and available, as well as actually switch between them. Since merely keeping two tasks available may by itself increase RTs in the mixed blocks, switch costs calculated by this method may reflect something more than just the extra processing incurred when subjects switch from one task to another. The study reported in this paper examined the efficiency of executive control processes in both Parkinson’s disease patients and patients with frontal lesions, when required to switch repeatedly between two simple reactive tasks: letterand digit-naming. Since age appears to exert demonstrable effects on a range of executive functions (Moscovitch and Winocur, 1995), two separate control groups were used: one matched for IQ and age against the sample of patients with frontal lesions, and another matched for IQ and age against the sample of Parkinson’s disease patients. Three background neuropyschological tests confirmed the behavioural profiles of the two groups of patients: those with frontal lesions and

those with Parkinson’s disease. The study examined several issues simultaneously, as follows.

Task-specific interference Recent results have suggested that switching from one task to another is harder (i.e. the time cost incurred by a task switch is increased) in the presence of a stimulus that activates the currently inappropriate task (Rogers and Monsell, 1995). Given that an important aspect of our task control appears to be that everyday stimuli (occurring in appropriate contexts) can trigger entire action sequences associated with them (Reason, 1979, 1990), and that a pathological failure to modulate this stimulus-driven task activation is an occasional consequence of frontal lobe damage in so-called ‘utilization behaviour’ (Lhermitte, 1983; Shallice et al., 1989), it seemed natural to examine the efficiency of task switching in conditions with and without such stimuli.

Task cues One common observation in studies of task switching with normal, non-brain damaged individuals is that switch costs can be reduced, and sometimes even reversed, when taskrelevant information is presented to act as an effective memory cue for specifying the next task (Jersild, 1927; Spector and Biederman, 1976). Given that frontal damage is widely reported to lead to inattentiveness (Stuss and Benson, 1986), and that several experiments with Parkinson’s disease patients suggest an amelioration of performance deficits where strong task cues are available (Brown and Marsden, 1988b; Lee and Brown, 1992), we investigated task switching in the context of both weak and strong task cues.

Inhibition Although very little is currently known about the executive control processes that together mediate dynamic shifts between task-sets, it can be presumed that processes of ‘selective attention’ that allow subjects to process preferentially certain stimuli at the expense of others will be involved where, for example, a change of task requires a shift of attention from one stimulus to another, or from one stimulus attribute to another. Plenty of evidence now suggests that the active inhibition of irrelevant information is one important contributory process in successful selective attention (e.g. Allport et al., 1985; Tipper and Cranston, 1985). The paradigm used in this study permits a preliminary examination of the role of these inhibitory processes in the context of task switching.

Practice Most conceptions of frontal cortex function incorporate the notion that its involvement is greatest when the behavioural demands of the moment are novel (Shallice and Burgess,

Frontostriatal mechanisms of task control 1996). It follows from this that increased practice should reduce the involvement of frontal cortical systems whilst enhancing the involvement of other neural systems, such as the basal ganglia. In this study, we addressed the issue of practice on two levels. First, we examined performance of each experimental condition in terms of the number of encounters (i.e. the number of blocks of trials) the subject had with that condition to investigate whether greater experience or practice ameliorated or exacerbated any deficits seen in the patient groups. Secondly, we examined performance over successive intervals of trials within each block to see whether continued switching between tasks improved the efficiency of switching task or impaired it. In particular, there are many clinical reports of the effect of ‘fatigue’ in patients with Parkinson’s disease in the performance of various motor tasks (e.g. Schwab, 1960), suggesting that one consequence of dopaminergic dysfunction is impaired persistence of motor set. Previous work has suggested that switch costs can be progressively increased over successive trials (Rogers, 1993) and we wished to discover whether such effects could be shown in either frontal or Parkinson’s disease patients.

Methods This study was approved by the Cambridge Psychology Research Ethics Committee (University of Cambridge), and all subjects gave informed consent.

Subjects Patients with frontal lesions Twelve patients with damage to the frontal lobes participated in the study. Eight of these patients had undergone frontal excisions. Of the remaining four patients, there was one case of cerebral infarct, one case of subarachnoid haemorrhage, and two cases of unoperated tumour. Of the excision patients, there were three cases of left-sided lobectomy, two cases of right-sided resection, one case of a right-sided arteriovenous malformation removal and two cases of right-sided astrocytoma removal. All lobectomy and resection patients had undergone surgery for intractable epilepsy. Overall, the cortical damage suffered by these patients is confined to the frontal lobes, the only exceptions being the one case of left frontal lobectomy whose medial excision extends posteriorly to include the extreme rostrum of the corpus callosum (see Patient 5 below), and the one case of right-sided astrocytoma removal, whose MRI scan shows a mass lesion in the corpus callosum (Patient 11). It should be noted that the pattern of results to be reported here, and also their statistical reliability, are not changed when these patients are excluded from the data analyses. Descriptions of the leftand right-sided damage seen in the patients are set out below, while four examples of representative excisions are shown

819

in Fig. 1. Other demographic and psychometric details are set out in Table 1. Patient 1 had damaged areas in left anterior dorsolateral cortex (2.5 cm) associated with the presence of a calcified meningioma (2.5 cm). Some thickening and sclerosis of underlying tissue and slight surrounding oedema are also present. Patient 2 had a lesion in the area of the lower-mid convexity of left frontal cortex (2 cm) caused by calcified glioma. There is some compression of the sylvian fissure and some ventricular enlargement. Patient 3 had loss of cortex from the left anterior pole, including the anterior half of the gyrus rectus on the medial surface and extending back to the anterior operculum on the lateral surface. Patient 4 had a lesion of the left dorsolateral frontal cortex produced by an ischaemic event (left middle cerebral artery). Patient 5 had extensive cortical loss incorporating most of the anterior and medial part of the frontal cortex, including the gyrus rectus, medial orbital gyri and cortex as far back as the frontal horn of the lateral ventricle. The extreme rostrum of the corpus callosum had also been damaged. Patient 6 had limited damage to middle and inferior left frontal cortex, incorporating some orbital gyri in their most lateral aspects. The inferior medial surface is entirely spared. Patient 7 had cortical loss incorporating most of the right frontal pole, extending back to middle frontal gyri laterally, but just sparing the cingulate gyrus medially. Patient 8 had extensive loss of cortex in right superior lateral and medial regions of frontal cortex, sparing the pole anteriorly but including the superior operculum posteriorly. Damage to white matter extended deep to the right frontal horn. Patient 9 had cortical loss in inferior medial regions of the right frontal pole. The anterior horn of the ventricle was slightly deformed with some displacement of the head of the caudate nucleus. Patient 10 had a lesion limited to right inferior lateral and medial frontal cortex during surgery for the clipping of an aneurysm. Patient 11 had extensive cortical loss, incorporating the superior medial right frontal cortex but sparing the cingulate cortex. Tissue loss extended to the lateral ventricle but spared the inferior and lateral cortex. There was evidence of a mass lesion in the corpus callosum. Patient 12 had cortical loss involving the superior and middle right frontal gyri across the medial surface. However, the cingulate gyrus was spared.

Parkinson’s disease patients Thirteen patients participated in the study. All patients presented to a general neurology clinic and were initially diagnosed (by J.R.H.) as having idiopathic Parkinson’s disease. All have been followed-up at regular intervals. The data of one patient were excluded from the analyses because

820

R. D. Rogers et al.

Frontostriatal mechanisms of task control

821

Fig. 1 A series of single-slice MRIs to show the extent of excision in two representative cases each of left- and right-sided frontal lobe lesions (the positions of sections a, b and c are indicated schematically in each case): (A) left-sided inferior frontal excision; (B) leftsided unilateral lobectomy; (C) right-sided superior frontal excision; (D) right-sided unilateral resection.

822

R. D. Rogers et al.

Table 1 Characteristics of the patients with frontal lesions, the patients with Parkinson’s disease, and the two corresponding groups of control subjects Group (n)

Sex M/F

Age (years)

Verbal IQ (NART)

Side of damage (L/R/Bi)

Hoehn and Yahr

Years since onset

Frontal lesions Patients (12) Controls (14)

5/7 7/7

37.3 6 2.3 37.2 6 2.4

117 6 1.9 118 6 1.7

6/6/0 –

– –

7.1 6 1.4* –

Parkinson’s disease Patients (12) 6/6 Controls (12) 5/7

59.2 6 1.8 58.8 6 1.8

111.0 6 1.9 113.5 6 2.8

– –

1.2 6 0.1 –

2.4 6 0.3 –

M/F 5 male/female; L/R/Bi 5 left/right/bilateral. *Excludes one case of unoperated glioma and one case of unoperated meningioma.

of a lack of response to L-dopa medication suggesting that the diagnosis of Parkinson’s disease was uncertain. All 12 remaining patients were in the early stages of their disease with 10 patients showing a Hoehn and Yahr (1967) rating at stage I, and two patients showing a rating at stage II. All were receiving L-dopa medication with good effect. One patient was also receiving anticholinergic medication. The mean duration (6 SEM) of the group’s disease was 2.4 60.3 years. Other details are summarized in Table 1.

Control subjects Two groups of normal, healthy volunteers were recruited to participate as control subjects in the study. One group of 14 subjects was selected as the control group for the patients with frontal lesions, and another group of 12 was selected as the Parkinson’s disease control group. Each group of control subjects was selected in order to match their particular patient group closely in terms of both age and premorbid verbal IQ, as assessed using the National Adult Reading Test (Nelson, 1982). Table 1 summarizes the characteristics of all four groups of subjects. One-way analyses of variance (ANOVA) showed that there were no significant differences between the patients with frontal lesions and their control subjects in terms of age (F, 1) or estimated premorbid verbal IQ (F , 1). Similarly, there were no significant differences between the Parkinson’s disease patients and their control subjects in terms of either measure (F , 1 for both). Finally, all subjects had normal or corrected vision; in particular, their colour vision was tested before the experiment began, using Ishihara plates (Kanehara, 1992) to check for colour-blindness.

Background neuropsychological assessments Verbal fluency This task was administered according to the method of Benton (1968). Each subject was given 60 s in which to generate as many words as possible beginning with first F, then A, and finally S. Instructions were to avoid item repetitions, proper names and words sharing the same stem. Additionally, each subject was given 90 s in which to generate as many animals as possible. Two scores were calculated:

the sum of words generated for F, A and S, and the total number of animals generated.

Spatial working memory Subjects were required to ‘search through’ an array of boxes presented on the screen by touching each one so that it ‘opened up’, revealing what was inside; see Owen et al. (1990) for full description. The object of the task was to collect ‘blue tokens’ hidden inside the boxes, and once found, to use them to fill an empty column at the side of the screen. At any one time, there was only a single token hidden inside one of the boxes, and the subjects were required to search until they found it, at which point the next token would be hidden. The key instruction was that once a blue token had been found within a particular box, that box would never be used again to hide a token. Two types of search error were possible. First, a subject may have returned to open a box in which a blue counter had already been found (a ‘betweensearch’ error). Secondly, a subject may have returned to a box already opened and shown to be empty earlier in the same search sequence (a ‘within-search’ error). Subjects could search the boxes in any order, but for control purposes, the number of empty boxes visited (excluding errors) before a token was found was determined by the computer. There were four test trials with each of four, six and eight boxes. The task was scored according to the number of betweenand within-search errors at each level of difficulty. An efficient strategy for completing this task was to follow a predetermined search sequence, beginning with a particular box and then returning to start each new search with that same box as soon as a token had been found. The extent to which this repetitive searching pattern was used as a strategy for approaching the task was estimated from the total number of sequences started with the same box, within each of the more difficult levels of six- and eight-box arrays. The total of these scores provided a single measure of strategy for each subject, with a high score (i.e. many sequences beginning with a different box) representing low use of the efficient strategy and a low score (i.e. many sequences starting with the same box) representing more extensive use.


823

Recognition memory: patterns See Sahakian et al. (1988) for a full description. Subjects were shown a series of 12 coloured patterns appearing one at a time inside a white box located in the centre of the screen (presentation) phase. Each of these ‘target’ patterns was presented for 3 s, the screen was then cleared and the next pattern appeared. In the second (recognition) phase, 12 pairs of coloured patterns appeared on the screen (one pair at a time) and the subject was required to respond to each pair by touching the pattern they had already seen during the presentation phase. Each of the target patterns was presented in reverse order and paired with a distractor pattern that differed in form but not in colour from the target. Each response was accompanied by an auditory tone, and visual feedback was provided by the computer in the form of green ticks and red crosses. This procedure was repeated with 12 new patterns (maximum score 5 24).

Recognition memory: spatial See Sahakian et al. (1988). This task was also presented in two phases. In the first (presentation) phase, subjects were shown a series of five unfilled white squares, appearing one at a time, at different locations on the screen. Each square was presented for 3 s before the screen was cleared and the next square presented. In the (recognition) phase, two squares appeared simultaneously on the screen and the subject had to select which location had been used before in the presentation phase. The target squares were presented in reverse order and paired with distractor squares which appeared in novel locations never used as target locations. Again, each response was accompanied by an auditory tone, and visual feedback was provided in the form of green ticks and red crosses. This procedure was repeated three more times using new target and distractor locations (maximum % score 5 20).

Task switching The tasks The tasks used were letter- and digit-naming. Following Rogers and Monsell (1995), each stimulus consisted of a pair of characters presented side by side. In the letter-naming task, one (and only one) of the characters was a letter, and the subject was simply required to name that letter as quickly as possible without making a mistake. Similarly, in the digitnaming task, one (and only one) character was a digit, and the subject had to name it as fast as possible, without making a mistake. The relevant character was randomly presented on the left or right of the pair and the two characters were closely adjacent. In the ‘no-crosstalk’ manipulation, the irrelevant character was always drawn from a set of neutral, non-alphanumeric characters. Since none of these characters are nameable as letters or digits, every character pair in this manipulation afforded a response only with respect to the currently

Fig. 2 (A) Example stimuli from the no-crosstalk and crosstalk manipulations (‘G%’ and ‘G7’). In the no-crosstalk manipulation, 100% of the stimuli afforded a response by the current task only while, in the crosstalk manipulation, this was true of only 33% of the stimuli. The remaining 66% afforded a response by the currently appropriate task and inappropriate task; (B and C) Typical sequences of trials as seen with the two kinds of cue used in the experiment. In the colour cue manipulation (B), the current task was indicated by the colour of the stimulus window while in the word cue manipulation (C), the current task was indicated by the name of the current stimulus category printed at the top of the stimulus window.

appropriate task. In the ‘crosstalk’ manipulation, the irrelevant character was drawn from the neutral set on one third of the trials. However, on the other two thirds of trials, the character pair contained both a letter and a digit together. Thus, in this condition, the irrelevant character was sometimes associated with a response by the currently inappropriate as well as the appropriate task, and sometimes not (see Fig. 2A). In the colour cue manipulation, the character pair appeared centred within either a red or a green stimulus window presented on a computer screen (Fig. 2B). A subject might

824

R. D. Rogers et al.

be told to name the letter and ignore the digit (if a digit were present) when the stimulus window was green but to name the digit and ignore the letter (if a letter were present) when it was red. We assumed that the link between the colour of the stimulus window and the adoption of a particular taskset such as letter or digit-naming would be essentially arbitrary, and therefore, that the colour cue manipulation constituted a situation in which the current task was signalled by an initially weak cue. The assignment of colour (green or red) to task (letter or digit-naming) was balanced across each subject group. In the word cue manipulation, the stimulus window was white but with the word ‘LETTER’ or the word ‘NUMBER’ printed at the top in large bold letters (Fig. 2C). The subject was instructed to name the letter and ignore the digit (if a digit were present) when the word cue was ‘LETTER’ and to name the digit and ignore the letter (if one were present) when the word cue was ‘NUMBER’. Arguably, the links between these written cues and their respective task-sets are more immediate, or less arbitrary, than those of the colour cue condition. Thus, we assumed that the word cue manipulation provided a situation in which the current task was signalled by an initially stronger task cue. As described above, successive trials followed an alternating-runs sequence so that two letter-naming trials were followed predictably by two digit-naming trials. An example sequence is shown in Fig. 2B and C with both colour cues and word cues. Notice that the correct responses are indicated by either the colour of the stimulus window, or by the named category written in bold black letters. However, in all conditions, the subject was required to switch task on every second trial of the sequence.

Procedure The subjects sat with their eyes approximately 60 cm from the screen of a Taxan SV-775EV touch-sensitive computer screen controlled by a DELL 486 Dimension personal computer. All the stimuli were presented within the stimulus window that was 12.5 cm wide and 9.2 cm high. Each character pair was displayed in an upper-case Helvetica font, subtending 1.2° horizontally and 0.6° vertically, and remained on the screen until the subject responded by naming one of the characters. There then followed a response–stimulus (R– S) interval of 1000 ms before the next character pair was presented. Responses were monitored using a purpose-built voice key constructed at the Department of Experimental Psychology at the University of Cambridge, and a small throat microphone (RS Components 250–479). RTs were measured in milliseconds. Letters were sampled randomly from the set {G, K, M, P, R, A, E, U} and digits from the set {2, 3, 4, 5, 6, 7, 8, 9}. Neutral characters were drawn from the set {?, *, %, #}. Fresh pseudo-random stimuli sequences were constructed for every subject with the restriction that the same character did not appear on two successive trials.

Between blocks, the word ‘Ready’ was displayed until the experimenter presented the display by pressing the space-bar of the computer keyboard. The first character pair of the next block appeared 2000 ms later. At the end of each block, a feedback display indicated to the subject the mean RT for that block. The experimenter indicated the number of errors. The subject was instructed to try to minimize RT while avoiding errors. A rest break was allowed every four blocks.

Design Testing consisted of four experimental conditions in which the cue and crosstalk manipulations were crossed. Thus, the experimental conditions were as follows: (i) colour cue and no crosstalk; (ii) colour cue and crosstalk; (iii) word cue and no crosstalk; and (iv) word cue and crosstalk. The order of presentation of these combinations was counterbalanced within each subject group. Each condition consisted of four blocks of trials with each block consisting of 40 trials, the first three of which were discarded as ‘warm-up’ trials and excluded from the data analyses. Also, before each of these conditions, the subject was given two blocks of practice with that condition. These practice blocks were examined in later analyses. During the colour cue conditions, a reminder of the colourtask assignments was placed beneath the computer screen. This consisted of red and green colour samples placed next to cards with the words ‘LETTER’ and ‘NUMBER’ printed on them in large black letters.

Pre-experimental training Before completing the experimental conditions, preliminary training was given with the naming tasks. Character pairs were displayed one at a time in the centre of a white stimulus window. The irrelevant character was always neutral and the R–S interval was 1000 ms. Each subject alternated twice between 24-trial blocks of letter-naming and digit-naming. Performance feedback was given as above.

Results Background neuropsychology Verbal fluency Table 2 shows the performance of all four subject groups on the three background neuropsychological tasks. In the verbal fluency test, the patients with frontal lesions generated significantly fewer items than their age- and IQ-matched control subjects [F(1, 24) 5 7.2, P , 0.05], with a strong tendency for this reduction in fluency to be larger in the F, A and S letter section of the test than in the animal section [F(1, 24) 5 4.1, P 5 0.06]. There was also a tendency for the patients with left-sided damage (LF patients) to generate fewer items than the patients with right-sided damage (RF patients) and the control group (see Table 2) [F(2, 23) 5


825

Table 2 Performance of patients and matched control subjects on neuropsychological tests Frontal lesions

Parkinson’s disease

Left-sided patients

Right-sided patients

All patients

Control subjects

Patients

Control subjects

36.2 6 3.6 23.0 6 3.4

37.3 6 5.2 26.7 6 2.2

36.8 6 3.1 24.8 6 2.0

48.9 6 2.9 30.1 6 2.2

35.4 6 2.3 21.8 6 2.4

39.8 6 3.1 27.3 6 1.4

Spatial working memory Six boxes 10.5 6 1.7 Eight boxes 31.3 6 7.3 Strategy 35.3 6 2.1

12.0 6 3.4 29.5 6 3.6 37.2 6 1.3

11.3 6 1.8 30.4 6 4.0 36.3 6 1.2

4.3 6 1.3 12.7 6 2.6 30.3 6 1.5

12.0 6 4.9 27.0 6 4.5 30.1 6 3.1

6.2 6 1.7 20.0 6 4.2 30.9 6 1.8

Recognition memory Pattern 18.3 6 2.7 Spatial 15.5 6 1.0

20.3 6 1.1 15.7 6 0.7

19.3 6 1.4 15.6 6 0.6

21.9 6 0.5 16.8 6 0.3

20.4 6 0.5 14.3 6 0.7

21.3 6 0.7 16.1 6 0.6

Verbal fluency F, A and S Animals

2.8, P 5 0.08]. The Parkinson’s disease patients also generated fewer items than their control subjects, although not quite significantly so [F(1, 22) 5 3.1, P 5 0.09]. The interaction between group and section was not significant (F , 1).

terms of the laterality of damage failed to show any marked differences for either pattern- or spatial-recognition memory. The Parkinson’s disease patients did not make significantly more errors compared with their control subjects for the pattern recognition task (F , 1), but they did for the spatial recognition task [F(1, 22) 5 5.1, P , 0.05].

Spatial working memory Consistent with earlier data (Owen et al., 1990), the patients with frontal lesions showed severe deficits on this selfordered search task. First, the number of between-search errors was increased relative to control performance [F(1, 24) 5 19.1, P , 0.001], with the increase being significantly greater when the patients were required to search through eight boxes compared with only six boxes [F(1, 24) 5 5.7, P , 0.05]. Secondly, the greater number of between-search errors of patients with frontal lesions was also associated with the use of less efficient search strategies than their control subjects [F(1, 24) 5 9.2, P , 0.01]. Analysing the data in terms of laterality of damage revealed that both LF and RF patients showed significantly increased betweensearch errors compared with the control subjects (Newman– Keuls; P , 0.01). However, it was the RF patients who showed the significantly impaired strategy scores compared with control levels (Newman–Keuls: P , 0.05), a finding consistent with results recently reported by Miotto et al. (1996) using a formally equivalent task. In contrast to the patients with frontal lesions, the Parkinson’s disease patients did not make significantly more between-search errors compared with their control subjects [F(1, 22) 5 1.3]. Their strategy scores were also not consistently worse than control subjects (F , 1).

Recognition memory: pattern and spatial Compared with their control subjects, the patients with frontal lesions showed near-significant increases in the number of errors in both the pattern-recognition and the spatialrecognition memory tasks [F(1, 24) 5 3.1, P 5 0.09 and F(1, 24) 5 3.3, P 5 0.08, respectively]. Further analysis in

Summary The patterns of deficit shown by both the frontal and Parkinson’s disease patient groups matched those seen in previous studies using the same tasks. Thus, the patients with frontal lesions showed significant impairments in the verbal fluency task and the self-ordered search task, both of which have previously been found to be sensitive to frontal lobe damage (Milner, 1964; Benton et al., 1968; Owen et al., 1990, 1996a). Meanwhile the medicated, early stage Parkinson’s disease patients showed deficits on the spatial recognition memory task that has been demonstrated previously to be sensitive to Parkinson’s disease at various stages, as well as to frontal lobe pathology (Sahakian et al., 1988; Owen et al., 1995). Therefore, both the sample of patients with frontal lesions and the sample of Parkinson’s disease patients showed cognitive deficits that are typical of their patient populations as a whole.

Task switching Analysis The following trials were excluded from all data analyses: trials with naming latencies (mean RTs) of ,200 ms; trials with naming latencies of .5000 ms; any trial which the experimenter judged to be being unreliable (e.g. if the voice key was triggered by extraneous noise or unintended lippops); any three trials immediately following an error or any other excluded trial. Mean RTs and error rates from the test conditions were initially subjected to repeated measures ANOVAs having the between-subject factor group (control subjects versus patients

826

R. D. Rogers et al.

Table 3 Switch costs for patients with frontal lesions and their matched control subjects in the four experimental conditions Colour cues

Word cues

No crosstalk

Crosstalk

No crosstalk

RT (ms)

Error (%)

RT (ms)

Error (%)

RT (ms)

584.7 570.8 13.9

0.6 0.7 –0.1

783.8 701.8 82.0

3.9 2.1 1.8

486.8 475.6 11.2

Their matched control subjects Switch trials 452.9 Non-switch trials 437.0 Switch cost 15.9

0.2 0.2 0.0

640.1 606.9 33.2

3.4 2.3 1.1

442.4 435.0 7.4

Patients with frontal lesions Switch trials Non-switch trials Switch cost

Crosstalk Error (%)

RT (ms)

Error (%)

0.9 0.7 0.2

723.5 713.7 9.8

4.9 2.6 2.3

0.1 0.4 –0.3

634.8 644.0 –9.2

3.9 2.6 1.3

Switch costs are calculated by subtracting the mean RTs and error rates for the non-switch trials from those for the switch trials.

with frontal lesions or Parkinson’s disease patients) and within-subject factors cue condition (colour cue versus word cue), crosstalk condition (no crosstalk versus crosstalk), task (letter-naming versus digit-naming) and trial type (switch versus nonswitch). Although no transformations were deemed necessary for the RT data, the error rates were arcsintransformed as is appropriate whenever the variance is necessarily proportional to the mean (see Howell, 1987). However, the error data shown in those figures represent untransformed values. Simple interaction effects were tested only when all of the higher-order interaction terms were also significant. Moreover, marginally significant main effects and interactions involving repeated measures were further tested with a Greenhouse–Geisser correction in those cases in which the assumption of covariance was violated (Howell, 1987). Follow-up analyses with different between- and withinsubject factors are described in the text. A summary of the main findings is presented at the end of the Results section.

Effects of task cue Patients with frontal lesions versus control subjects. Switch costs were calculated by subtracting the mean RTs and error rates for the non-switch trials from the mean RTs and error rates of the switch trials. Irrespective of subject group, mean RTs on the switch trials were significantly slower than on the non-switch trials [F(1, 24) 5 9.5, P , 0.01]. Error rates were also significantly increased [F(1, 24) 5 6.7, P , 0.05]. Thus, switching between the letterand the digit-naming tasks yielded significant time and error costs. Furthermore, the time costs, but not the error costs, were reduced when the task cue was a word printed at the top of the stimulus window (as in the word cue conditions) as opposed to the colour of the stimulus window itself (as in the colour cue conditions) [F(1, 24) 5 9.5, P , 0.01]. Table 3 shows the mean RTs, error rates and switch costs separately for the patients with frontal lesions and their ageand IQ-matched group of control subjects, for each of the conditions under study. Both groups showed similar effects of cue condition on the time cost of a task switch. Thus,

while the control subjects were able to reduce the time cost from 25 ms in the colour cue conditions to –1 ms in the word cue conditions, the patients with frontal lesions showed a corresponding reduction of 48 to 11 ms. Consequently, the three-way interaction between group, cue condition and trial type was not significant for either RTs or error rates (all Fs , 1). In summary, there was no evidence of a frontal impairment in the ability to improve the efficiency of switching between two naming tasks in the presence of stronger task cues as opposed to weaker task cues.

Parkinson’s disease patients versus control subjects. Averaging over subject group, mean RTs on the switch trials were significantly longer than those on the nonswitch trials when the task cue was the colour of the stimulus window [F(1, 22) 5 7.4, P , 0.05], but not when the task cue was a printed word, the two-way interaction between cue condition and trial type being significant [F(1, 22) 5 22.7, P , 0.001]. Error rates were significantly higher on switch trials compared with non-switch trials regardless of the type of task cue available to the subjects, yielding a significant error cost of 1.4% [F(1, 22) 5 19.1, P , 0.001]. Table 4 shows the mean RTs, error rates and switch costs separately for the Parkinson’s disease patients and their control subjects. Comparing the Parkinson’s disease patients with the control subjects reveals that the reduction in the time cost when switching between tasks cued by words rather than colours was considerably enhanced in the case of the control subjects relative to the Parkinson’s disease patients [F(1, 22) 5 5.3, P , 0.05]. Moreover, as Table 4 indicates, this effect was itself modulated by the existence, or otherwise, of crosstalk between the two tasks. In fact, the time costs for the two groups of subjects were quite similar in all of the experimental conditions except for the word-cue and crosstalk condition. Here the control subjects showed a very large switch ‘benefit’ of 75 ms compared with the small switch cost of 2 ms shown by the Parkinson’s disease patients (see Table 4). This was confirmed by a significant four-way interaction between group, cue condition, trial type and crosstalk condition for the RTs [F(1, 22) 5 6.4, P , 0.05].


827

Table 4 Switch costs for patients with Parkinson’s disease and their matched control subjects in four experimental conditions Colour cues

Word cues

No crosstalk

Crosstalk

No crosstalk

Crosstalk

RT (ms)

Error (%)

RT (ms)

Error (%)

RT (ms)

Error (%)

RT (ms)

Error (%)

530.3 510.9 19.4

0.8 1.2 –0.4

733.7 708.6 25.1

4.4 1.7 2.7

517.4 509.8 7.6

0.6 0.3 0.3

691.8 689.7 2.1

3.8 1.5 2.3

Their matched control subjects Switch trials 485.2 Non-switch trials 471.2 Switch cost 14.0

0.2 0.1 0.1

649.3 631.9 17.4

5.7 1.7 4.0

477.0 471.1 5.9

0.7 0.2 0.5

647.7 722.4 –74.7

4.5 2.5 2.0

Parkinson’s disease patients Switch trials Non-switch trials Switch cost

Superficially, it appears as if the Parkinson’s disease patients were not as effective as the control subjects in using the word cues to reduce the time cost of task switches further. However, the sizeable switch benefit for the control subjects was not in fact caused by dramatically reduced RTs on the switch trials so much as increased RTs on the non-switch trials (see Table 4). On the other hand, it is clear that the Parkinson’s disease patients were at least able to abolish the time cost of task switches when the task cues were words rather than colours.

Effects of crosstalk (between-block comparison) Patients with frontal lesions versus control subjects. Averaging over subject group, interference between the two tasks (in the crosstalk conditions as opposed to the nocrosstalk conditions) significantly increased overall RTs [F(1, 24) 5 95.6, P , 0.0001] and overall error rates [F(1, 24) 5 44.7, P , 0.0001], and also increased the time cost of a task switch with colour cues but not with the word cues [F(1, 24) 5 8.1, P , 0.01]. The error cost of a task switch was increased with both colour and word cues [F(1, 24) 5 7.0, P , 0.05]. Regardless of task cue, crosstalk did not produce significantly longer RTs or higher error rates for the patients with frontal lesions when compared with the control subjects (F , 1 for both). Nor did crosstalk appear to lead to significantly larger time or error costs of a task switch for the patients compared with the control subjects [F(1, 24) 5 1.3 and F , 1], respectively. However, Table 3 clearly indicates that the patients with frontal lesions showed well over twice the time cost of the control subjects in the colourcue and crosstalk condition. To investigate this difference further, the mean time cost of the control subjects was calculated for this condition, together with 95% confidence intervals placed around this mean (see Fig. 3A). The individual time costs of the patients with frontal lesions were then plotted and compared with this mean. Figure 3A shows that LF patients generally showed larger time costs than the RF patients and the control subjects. This

finding was confirmed in a separate analysis of the colourcue and crosstalk condition by itself with group (control subjects versus LF patients versus RF patients) as the between-subject factor. The two-way interaction between group and trial type was highly significant (see Fig. 3B) [F(2, 23) 5 6.3, P , 0.01], with analysis of simple effects showing significant differences in RTs between the groups when switching task [F(2, 25) 5 4.6, P , 0.05], but not when performing the same task [F(2, 25) 5 1.5]. Importantly, an additional, direct comparison of the two sets of patients with frontal lesions indicated significantly greater time costs for a task switch after left-sided damage compared with right-sided damage [F(1, 10) 5 9.0, P , 0.005]. In summary, the LF patients were significantly slower to switch task when there was interference or crosstalk between the two tasks. Finally, a separate analysis was undertaken to examine whether the increased time costs shown by the LF patients with the arbitrary task cues of the colour-cue and crosstalk condition were differentially reduced compared with those of the RF patients and control subjects by the stronger, more immediate, task cues of the word-cue and crosstalk condition. In fact, the reductions were from 33 to –9 ms for the control subjects, from 28 to –25 ms for the RF patients, and from 136 to 45 ms for the LF patients. Thus, although all three subject groups showed substantial improvements in the speed of switching tasks [F(1, 23) 5 10.1, P , 0.005], there was little indication that stronger task cues produced markedly greater improvements for the LF patients. Consequently, the three-way interaction between cue condition, trial type and group was not significant (F , 1). The very large LF deficit with arbitrary task cues of the colour-cue and crosstalk condition was reduced, but not abolished, with the stronger cues of the word-cue and crosstalk condition.

Parkinson’s disease patients versus control subjects. Overall RTs and error rates were increased for both the Parkinson’s disease patients and the control subjects in the crosstalk compared with the no-crosstalk conditions [F(1, 22) 5 115.8, P , 0.0001 and F(1, 22) 5 84.3, P , 0.0001, respectively]. However, switching task in the

828

R. D. Rogers et al.

Fig. 3 (A) Time costs of a task switch for individual patients plotted against the mean time cost of the control group (centred within 95% confidence intervals) in the colour-cue and crosstalk condition. The patients are split into two groups; LF patients are those with left-sided damage and RF patients those with right-sided damage. (B) Mean RTs and error rates of the LF patients, the RF patients and control subjects on the switch and non-switch trials separately. Bars represent one standard error of the difference between the means (SED) taken from the betweengroup component of the group3trial-type interaction term of the ANOVA. This is an appropriate index of variation for computing post hoc tests between the mean values of the groups and is calculated according to the formulae provided by Cochran and Cox (1957).

crosstalk conditions produced improvements in RTs on the switch trials relative to the non-switch trials when the task cues were words but not when they were the colours [F(1, 22) 5 10.9, P , 0.01]. As noted above (see Effects of task cue), the size of these time ‘benefits’ when switching task in the word-cue and crosstalk condition was significantly smaller for the Parkinson’s disease patients than for the control subjects [F(1, 22) 5 6.4, P , 0.05].

Effects of crosstalk (within-block comparisons) To investigate the differences between the two groups of patients with frontal lesions (LF and RF patients) and the control subjects in the colour-cue and crosstalk condition further, a separate comparison was made between those trials on which the irrelevant character of the presented character pair was associated with a response by the now-inappropriate task (66% of all the trials) and those trials on which the irrelevant character was neutral (33%). Thus, the RT and error data were subjected to repeated measures ANOVAs with the between-subject factors of group and within-subject factors of value of the irrelevant character (task-associated or neutral), task and trial type. Table 5 shows the mean RTs, error rates and switch costs for each of these different types of trials for both groups of patients with frontal lesions and their control subjects, as well as for the Parkinson’s disease patients and their group of control subjects.

Patients with frontal lesions versus control subjects. Regardless of subject group, overall mean RTs and error rates were significantly increased when the irrelevant character was associated with a response by the now-inappropriate task rather than not so associated [F(1, 23) 5 124.9, P , 0.0001 and F(1, 23) 5 43.7, P , 0.0001, respectively]. Although the time cost was only slightly larger on task-associated trials (71 versus 60 ms) (F , 1), the error cost was significantly increased, from zero when the irrelevant character was neutral to 2.4% when it was task-associated [F(1, 23) 5 4.4, P , 0.05]. There was a strong tendency for the LF patients to respond more slowly than the RF patients and the control subjects (797, 632 and 608 ms, respectively) [F(2, 23) 5 2.7, P 5 0.08]; however, the LF patients also made fewer errors (1.3, 2.9 and 2.0%) [F(2, 23) 5 1.5]. Task-associated irrelevant characters did not significantly increase mean RTs and error rates more for either group of patients with frontal lesions compared with the control subjects [F(2, 23) 5 2.1 and F(2, 23) 5 1.5]. Although Fig. 4A shows that the LF patients produced the largest time cost of switching task on trials when the irrelevant character was associated with a response by the now-inappropriate task (146, 27 and 38 ms for the LF patients, RF patients and control subjects, respectively) this was also true when the irrelevant character was neutral (105, 39 and 36 ms), indicating that their deficit extended to trials on which crosstalk was expected but, in


829

Table 5 Switch costs according to presence of crosstalk and type of irrelevant character LF patients

RF patients

LF 1 RF controls

PD patients

PD controls

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

Task-associated (colour cue and crosstalk) Switch trials 967.0 4.4 Non-switch trials 820.8 1.2 Switch cost 146.2 3.2

716.1 688.6 27.5

6.7 5.0 1.7

691.7 653.4 38.3

4.9 3.0 1.9

786.6 756.0 30.6

5.8 2.3 3.5

691.3 669.4 21.9

7.4 2.7 4.7

Neutral (colour cue and crosstalk) Switch trials 753.0 Non-switch trials 647.3 Switch cost 105.7

0.0 0.0 0.0

582.4 543.1 39.3

0.0 0.0 0.0

560.8 524.7 36.1

0.0 0.0 0.0

653.6 628.2 25.4

0.8 0.4 0.4

587.1 565.1 22.0

0.0 0.0 0.0

No crosstalk (colour cue; all trials) Switch trials 697.4 Non-switch trials 676.0 Switch cost 21.4

0.6 1.2 –0.6

491.8 476.4 15.4

0.8 0.0 0.8

454.2 436.8 17.4

0.2 0.2 0.0

530.3 510.9 19.4

0.8 1.2 –0.4

485.2 471.2 14.0

0.2 0.1 0.1

Irrelevant character

PD 5 Parkinson’s disease.

fact, absent. Consequently, although the two-way interaction between group and trial type was highly significant for the RTs [F(2, 23) 5 6.1, P , 0.01], the interaction between these two factors and the value of the irrelevant character was not (F , 1). Finally, in a supplementary analysis, performance on the neutral trials of the colour-cue and crosstalk condition was compared with performance in the colour-cue and nocrosstalk condition (in which all trials were neutral) in order to assess directly whether the general incidence of crosstalk differentially increased the time costs of the LF patients on trials that were identical in not affording a response by the now-inappropriate task. Overall RTs were increased for all subjects in the crosstalk condition compared with the nocrosstalk condition (601 versus 538 ms) [F(1, 23) 5 5.1, P , 0.05]. Error rates were slightly, but significantly, reduced (0.0 versus 0.5%) [F(1, 23) 5 14.9, P , 0.001]. All subject groups showed increases in the time cost in the crosstalk relative to the no-crosstalk condition [F(1, 23) 5 16.1, P , 0.001]. However, this increase was significantly greater for the LF patients (see Fig. 4A) [F(2, 23) 5 4.0, P , 0.05], confirming that the deficit of the LF patients in switching tasks in the colour-cue and crosstalk condition was not just associated with particular trials requiring the control of current response interference but generalized to trials in which interference was expected but, in fact, absent.

Parkinson’s disease patients versus control subjects. Overall naming latencies and error rates in the colour-cue and crosstalk condition were significantly increased when the irrelevant character was task-associated rather than neutral [F(1, 22) 5 74.5, P , 0.0001 and F(1, 22) 5 55.2, P , 0.0001, respectively]. The error cost was also significantly increased (4.1 versus 0.2%) [F(1, 22) 5 12.3, P , 0.01]. The impact of such interference on overall RTs and error rates was not significantly different for the Parkinson’s disease patients from the control subjects (F , 1

for both). Furthermore, even though the increase in the time cost produced by task-associated irrelevant characters was slightly larger for the Parkinson’s disease patients (6 ms) compared with the control subjects (0 ms), the increase in the error cost was smaller (3.5 versus 4.7%). Neither difference approached statistical significance (see Table 5) (F , 1 for both). Thus, in contrast to the patients with frontal lesions, there is no evidence that task-relevant interference produced longer time costs in medicated, early stage Parkinson’s disease patients, when they are required to make a rapid switch from one task to another.

Effects of inhibition of competing categories The data from the colour-cue and crosstalk condition also permit an examination of the extent to which processing on the previous trial influences performance on trials requiring the subjects to switch between tasks. For example, one might suppose that having been confronted by a character pair such as ‘2P’ on a non-switch trial that required a number-naming response (see the colour cue condition of Fig. 2B), the processes of selecting first the correct stimulus category (numbers) and then the appropriate response for output might involve the temporary suppression of the ignored stimulus category (letters) to which the irrelevant character belongs. If this were the case then switching to name a member of that category on the following switch trial would be slowed relative to trials on which the previous trial required no such inhibitory operation, with character pairs such as ‘2#’ (see the word cue condition shown in Fig. 2B). Thus, the difference between naming latencies on switch trials preceded by nonswitch trials with task-associated irrelevant characters (e.g. ‘G8’ preceded by ‘2P’) and latencies on switch trials preceded by non-switch trials with neutral irrelevant character (‘G8’ preceded by ‘2#’) provides a measure of how much inhibition of the previously-ignored, but now-relevant, stimulus

830

R. D. Rogers et al.

Fig. 4 (A) The time and error costs for task-associated and neutral trials of the colour-cue and crosstalk condition, and for all trials of the colour-cue and no-crosstalk condition, for patients with left-sided frontal damage (LF patients), right-sided frontal damage (RF patients), and their control subjects. (B) The inhibitory and facilitatory effects of processing on preceding nonswitch for trials on the mean RTs on switch trials. Columns represent the differences between mean RTs for switch trials on which the irrelevant character on the previous non-switch trial was neutral and mean RTs for switch trials on which the irrelevant character on the preceding nonswitch trial was associated with the other task. Negative values indicate slowed responding (inhibitory effects), positive values indicate faster responding (facilitatory effects). Effects are shown separately for trials on which the irrelevant character on the current switch trial was taskassociated and neutral.

category has to be overcome when subjects switch from one task to another.

Patients with frontal lesions versus control subjects. Figure 4B shows such differences as functions of whether or not the irrelevant character on the current switch trial was itself task-associated or neutral. It is clear that when the irrelevant character on the switch trial was task-associated, and therefore itself a source of competition for a naming response, ignoring a task-associated irrelevant character on the previous non-switch trial slowed RT more for the LF patients than for the RF patients and control subjects (60, 16 and 13 ms). However, surprisingly, this effect was reversed on those switch trials on which the currently irrelevant character was actually neutral. On these trials, a taskassociated but non-irrelevant character on the previous nonswitch trial actually facilitated RT more for LF patients than for either of the other subject groups (74 ms versus 29 ms versus –5 ms). The increased sensitivity of the LF patients to these opposing inhibitory and facilitatory effects of preceding processing on performance on switch trials was confirmed first by the presence of a significant two-way interaction between value on the irrelevant character on the previous trial and value of the irrelevant character on the current trial [F(1, 23) 5 13.5, P , 0.01], and then by a significant three-

way interaction between these two factors and that of group [F(2, 23) 5 4.8, P , 0.05].

Parkinson’s disease patients versus control subjects. Both Parkinson’s disease patients and older control subjects showed similar effects of prior processing on switch trials. Specifically, when the irrelevant character on the current switch was itself task-associated, ignoring a taskassociated irrelevant character on the previous non-switch trial facilitated naming latencies relative to a neutral one for the Parkinson’s disease patients and their control subjects (44 and 14 ms, respectively). On the other hand, when the irrelevant character on the current switch trial was neutral, a task-associated irrelevant character on the previous nonswitch trial slightly inhibited naming latencies for both groups (21 and 11 ms). These effects combined to produce a nearsignificant two-way interaction between value of the irrelevant character on the previous trial and value of the irrelevant character on the current trial [F(1, 22) 5 3.7, P 5 0.06]. However, the three-way interaction between these factors and group was not significant (F , 1), suggesting that, in general, older subjects may show different effects of prior processing relative to younger control subjects, and that there is little effect of Parkinson’s disease in this regard.


831

Fig. 5 The time and error costs of a task switch as a function of trial interval in the crosstalk conditions only. (A) LF patients versus RF patients versus control subjects; (B) Parkinson’s disease patients versus control subjects.

Effects of practice (between-trials) To examine the effects of practice within a block of trials, the 40 trials of each block were broken up into five intervals and used to construct an extra within-subject factor of ‘trial interval’ with five levels (trials 3–8, 9–16, 17–24, 25–32 and 33–40). Mean RTs and error rates from the test conditions were then subjected to repeated measures ANOVAs having the between-subject factor of group and within-subject factors of trial interval, cue condition, crosstalk condition, task and trial type.

Patients with frontal lesions versus control subjects. Irrespective of group, overall mean RTs increased significantly over the course of 40 trials [F(4, 92) 5 12.4, P , 0.0001]. However, error rates did not increase or decrease [F(4, 92) 5 1.3]. The increase of RTs was slightly greater when there was crosstalk between the tasks than when there was none, but not significantly so [F(4, 92) 5 1.4]. Neither the time cost nor the error cost of a task switch showed significant changes over 40 trials [F(4, 92) 5 1.1 and F(4, 92) 5 1.5, respectively], and there were no reliable interactions involving trial interval, trial type and subject group. A separate analysis was conducted on the two crosstalk conditions considered by themselves (see Table 6, Fig. 5 and the section on Parkinson’s disease patients versus control subjects below). This confirmed that there were no marked changes in the time and error costs over the course of 40 trials for either the patient groups or control subjects so that neither the twoway interactions between trial interval and trial type, nor the three-way interactions between these two factors and that of group, were significant for the RTs or error rates (all F , 1.5).

Parkinson’s disease patients versus control subjects. Averaging over both subject groups, overall RTs increased significantly over the course of 40 trials [F(4, 88) 5 22.1, P , 0.0001]. Error rates, on the other hand, showed no significant increase or decrease (F , 1). Mean RTs also increased significantly more when there was crosstalk or interference between the tasks than when there was not [F(4, 88) 5 11.6, P , 0.0001]. Both the mean RTs and time costs of the Parkinson’s disease patients tended to increase more over 40 trials than those of the control subjects [F(4, 88) 5 2.2, P 5 0.07 and F(3, 88) 5 2.6, P 5 0.06, respectively]. However, similar but significant effects were evident for the error cost in the two crosstalk conditions of the experiment (see Fig. 5) as seen by the breakdown of a highly reliable four-way interaction between group, trial type, trial interval and crosstalk condition into its simple interaction effects [F(4, 88) 5 4.3, P , 0.01]. Thus, the control subjects showed significant error costs in the crosstalk conditions for trials 4–8 [F(1, 11) 5 9.3, P , 0.05], for trials 9–16 [F(1, 11) 5 8.9, P , 0.05], for trials 17–24 [F(1, 11) 5 14.5, P , 0.01], but not for trials 25–32 or trials 33–40 (F , 1 for both). In contrast, the Parkinson’s disease patients show a near-significant error cost for trials 4–8 [F(1, 11) 5 3.2, P 5 0.09], no significant costs for trials 9–16, trials 17–24, trials 25–32, but then a very large and reliable error cost of 6.0% for trials 33–40 [F(1, 11) 5 9.7, P , 0.05]. Furthermore, direct analysis of the error costs themselves revealed a significant decrease over successive trials for the older control subjects [F(4, 88) 5 3.9, P , 0.01], but a significant increase for the Parkinson’s disease patients [F(4, 88) 5 2.9, P , 0.05]. To

832

R. D. Rogers et al.

Table 6 Switch costs from the crosstalk conditions as a function of trial interval LF patients

RF patients

LF 1 RF controls

PD patients

PD controls

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

RT (ms) Error (%)

Trials 3–8 Switch trials Non-switch trials Switch cost

834.3 712.1 122.2

3.5 0.7 2.8

622.7 593.3 29.4

7.3 3.1 4.2

595.4 595.1 0.3

2.8 3.8 –1.0

651.7 661.3 –9.6

4.3 0.8 3.5

618.2 648.2 –30.0

8.0 1.2 6.8


842.0 781.3 60.7

2.9 0.6 2.3

630.0 649.3 –19.3

6.7 3.9 2.8

620.4 615.2 5.2

3.9 1.3 2.6

702.7 699.9 2.8

2.4 1.3 1.1

647.3 676.6 –29.3

4.5 1.7 2.8


829.3 757.9 71.4

0.6 1.0 –0.4

668.0 655.3 12.7

7.2 5.2 2.0

635.8 606.1 29.7

3.3 2.7 0.6

732.9 693.1 39.8

1.9 2.6 –0.7

625.1 684.4 –59.3

5.6 1.0 4.6


882.2 787.8 94.4

2.4 1.3 1.1

678.0 689.8 –11.8

5.0 1.7 3.3

655.0 660.9 –5.9

2.9 2.7 0.2

722.6 706.8 15.8

4.9 2.3 2.6

674.2 686.1 –11.9

3.3 4.3 –1.0


882.9 780.3 102.6

3.7 2.1 1.6

666.8 670.4 –3.6

4.5 3.8 0.7

680.7 650.2 30.5

5.4 1.7 3.7

754.1 734.5 19.6

6.9 0.9 6.0

677.9 690.4 –12.5

4.1 2.3 1.8

Irrelevant character

PD 5 Parkinson’s disease.

summarize, the Parkinson’s disease patients, but not patients with frontal lesions, exhibited increases in the error cost of task switching across successive trials while age- and IQmatched control subjects showed decreases.

Effects of practice (between-blocks) In order to assess the effects of practice in the experimental conditions across blocks of trials, an extra within-subject factor was constructed that combined the two practice blocks and the four test blocks of each condition. The resulting factor of ‘block’ had six levels: (blocks 1, 2, 3, 4, 5 and 6). The data were subjected to repeated measures ANOVAs having the between-subject factor of group and within-subject factors of block, cue condition, crosstalk condition, task and trial type.

Patients with frontal lesions versus control subjects. Although the overall RTs of all three subject groups reduced significantly over the course of six blocks [F(5, 115) 5 20.9, P , 0.0001], the LF patients showed significantly greater effects of block on the time cost of a task switch in the colour-cue and crosstalk condition, only as confirmed by the breakdown of a significant five-way interaction between group, block, cue condition, crosstalk condition and trial type [F(10, 115) 5 2.1, P , 0.05]. The data from this condition are shown in Fig. 6. Several features of the data are striking. First, the overall mean RTs of both the LF and RF patients were noticeably increased over blocks 1 and 2 than over blocks 4–6 compared

with the control subjects, reflected in significant two-way interactions between group and block [F(5, 90) 5 2.8, P , 0.05 and F(3, 90) 5 3.2, P , 0.05, respectively]. Error rates also tended to be increased for the LF patients compared with the RF patients and control subjects [F(10, 115) 5 1.8, P 5 0.07]. Secondly, the performance of both the LF and the RF patients showed greater between-subject variability early in the practice blocks compared with later. Thus, while the between-subjects variability of the control subjects remained unchanged between block 1 and block 6 (t 5 0.3, P . 0.25) (see Howell, 1987), the LF and the RF patients showed near-significant (t 5 1.9, 0.05 , P , 0.10) and significant reductions in variability (t 5 2.9, P , 0.05), respectively. Finally, separate analyses of the time costs, excluding the RF patients, showed that the LF patients exhibited a significantly greater time cost compared with control subjects on block 1 [F(1, 97) 5 13.2, P , 0.0001], and then increasing deficits over blocks 2 through to 6, as the time costs of the control subjects gradually decreased while those of the LF patients gradually increased [F(4, 72) 5 3.9, P , 0.01]. Thus, it appears that the effect of practice over successive blocks of trials involving task interference only highlighted the increased time cost of the LF patients compared with the other two subject groups.

Parkinson’s disease patients versus control subjects. Mean RTs were significantly reduced over six blocks for both groups of subjects [F(5, 110) 5 50.2, P , 0.0001]. Although the Parkinson’s disease patients showed a slightly greater reduction in both RTs (119 versus


833

Fig. 6 Mean RTs and percentage errors as a function of block in the colour-cue and crosstalk condition only. (A) Frontal lesion control subjects; (B) LF patients; (C) RF patients; (D) Parkinson’s disease control subjects; (E) Parkinson’s disease patients. Blocks 1 and 2 were the practice blocks, blocks 3, 4, 5 and 6 were the test blocks. Open squares represent switch trials; open diamonds represent non-switch trials.

82 ms) and error rates (0.6 versus –0.3%), neither difference was significant [F(5, 110) 5 1.4 and F(5, 110) 5 1.1, respectively]. Furthermore, there was no evidence of greater or smaller reductions in the time or error costs for the Parkinson’s disease patients relative to the control subjects (F , 1). A separate analysis of the colour-cue and crosstalk condition (see Fig. 6 and the section on patients with frontal lesions versus control subjects) likewise failed to reveal any between-group differences in the effect of block on overall performance or switch costs (F , 1 for all interactions). Thus, in contrast to the patients with frontal lesions, the Parkinson’s disease patients showed little evidence of markedly disorganized performance early in practice, and in contrast to the LF patients specifically, little evidence of increasing deficits in task switching across later blocks of trials.

showed evidence of greater sensitivity than control subjects to the inhibitory and facilitatory effects of previous processing when switching between tasks (see Fig. 4B). (iv) Both LF and RF patients showed slow, disorganized performance early in practice with the colour-cue and crosstalk condition. However, only LF patients showed increasing time costs relative to control subjects over the four succeeding blocks of trials (spanning some 240 trials) (see Fig. 6B). This was not the case for RF patients or the Parkinson’s disease patients. (v) Parkinson’s disease patients showed progressive and significant increases in the error costs of task switching while matched control subjects showed significant decreases (see Fig. 5B). No such changes were seen in patients with either left or right frontal cortical damage. (vi) Both the patients with focal damage of the frontal cortex and those with Parkinson’s disease were able to reduce the time cost of a task switch with strong cues compared with weak, arbitrary cues.

Summary To summarize, the principal results are as follows. (i) LF patients, but not RF patients or medicated, early stage Parkinson’s disease patients, exhibited significantly larger time costs of a task switch than age- and IQ-matched control subjects when they switched between tasks requiring the control of interference or crosstalk (see Fig. 3A). (ii) The time costs to the LF patients were significantly increased relative to control subjects, not only when there was crosstalk at the time of a task switch, but also when crosstalk was expected but, in fact, absent (see Fig. 4A). (iii) LF patients, but not RF patients or patients with Parkinson’s disease,

Discussion The present results demonstrate several important deficits in the task-switching performance of patients sustaining focal damage to the left frontal cortex (i.e. the LF patients) as well as indications of very much milder deficits in patients with early, medicated Parkinson’s disease. Specifically, the LF patients exhibited larger time costs associated with switching from one task to another under certain conditions of the experiment whereas Parkinson’s disease patients showed only progressive increases in the error costs of such switches as

834

R. D. Rogers et al.

a function of time on task. Both the LF patients and the Parkinson’s disease patients were typical in their patterns of deficit on three background neuropsychological tasks. These results support the general concept of a role for frontal cortex in the dynamic organization of behaviour (see Fuster, 1989), particularly in achieving the moment-by-moment reorganization of the cognitive system, or reconfiguring of task-set, that allows performance of different tasks in response to changing behavioural objectives. We first discuss the results in terms of the precise conditions under which taskset shifting deficits were observed, and then consider the novel theoretical implications of these findings for understanding the functional interactions of the frontal cortex with subcortical systems, specifically the striatum. The experimental paradigm used here to measure the performance costs of predictable switches between the simple reactive tasks of letter- and digit-naming was validated internally by replicating two principal properties of such costs previously demonstrated in normal subjects, namely, that the time cost was increased when there was crosstalk between the tasks (Rogers and Monsell, 1995), but largely abolished when the tasks were signalled by strong, as opposed to weak, task cues (Jersild, 1927; Spector and Biederman, 1976). Therefore, the paradigm used in this experiment was well-suited to exploring possible deficits in the task control of neurologically damaged patients.

Laterality effects of frontal lesions in task-set control The most important result of this experiment is the finding that the LF patients, but not the RF patients, showed a selective deficit when required to switch from one task to another, in that they exhibited significantly larger time costs associated with such switches compared with age- and IQmatched control subjects. However, the appearance of this deficit depended on the general incidence of interference or crosstalk between the two tasks, and was particularly evident when the discriminative cue for the next task was essentially arbitrary, such as the colour of the stimulus window, as opposed to a stronger, and arguably more direct cue, such as a word naming the currently relevant stimulus category. When these two conditions were satisfied (as they were in the colour-cue and crosstalk condition), the LF patients took markedly longer to execute a rapid switch from one task to another, suggesting that the left frontal cortical region mediates some of the mechanisms that control moment-bymoment reorganizations between task-sets. Furthermore, the LF, but not the RF, patients also showed evidence of increased sensitivity to the facilitatory and inhibitory effects of previous processing when switching tasks. Finally, both the LF and the RF patients showed markedly slower and more disorganized performance than the control subjects when first introduced to the colour-cue and crosstalk condition which, in the case of the RF patients at least, improved after only limited amounts of practice (,250 trials).

Before exploring the significance of these findings, we briefly consider a few issues concerning interpretation. The first can be dismissed easily. This is the possibility that the increased time costs of the LF patients were simply a consequence of generally increased overall response times in this patient group. In this case, the LF patients’ RTs on the switch trials might have been increased relative to those of the control subjects to the same extent as the corresponding increase in RTs on the non-switch trials. In fact, the proportionate increase in non-switch RT for the LF patients over the control subjects was only 25% while the increase in switch RT was 40%. Thus, the deficit of the LF patients is not a scaling artefact. Moreover, although the generally increased RTs of the LF patients suggest that left-sided damage may lead to general impairments on RT tasks that involve verbal material, like the one used here, the detailed performance analysis provided here has revealed several highly specific deficits that relate directly to the executive control of dynamic switches between simple tasks. A second issue concerns the specificity of the deficit in terms of the experimental conditions. These data demonstrate the largest increases in the time costs of a task switch for the LF patients in the colour-cue and crosstalk condition. However, it is unclear that it was the precise combination of crosstalk between the tasks plus weak, arbitrary task cues that particularly required the involvement of left-sided frontal cortical function in task-set reconfiguring processes. In this regard, it is important to note that the increased time costs of the LF patients in the colour-cue and crosstalk condition were not differentially reduced compared with those of the RF patients and the control subjects in the word-cue and crosstalk condition, i.e. when the task cues were arguably stronger and more immediately effective. Thus, while it is the case that the task control of the LF patients was improved by stronger task cues (as was the task control of the other two groups) there was little indication that such cues abolished the deficit. A related issue concerns the nature of the cortical damage present in our patients. In all cases, the lesions were partial, and therefore it is possible that additional deficits in the task control of both the LF and RF patients were concealed in some of the conditions of the experiment by the use of appropriate compensatory strategies mediated either by the intact regions of the frontal cortex or by other neural systems. The majority of the patients included had sustained damage to their frontal lobes several years prior to participation in the study and had, no doubt, developed a range of such strategies for coping with the cognitive deficits consequent to that damage. However, notwithstanding this general reservation, it is clear that, in the colour-cue and crosstalk condition, the requirement to switch repeatedly between the simple reactive tasks of letter- and digit-naming went beyond whatever compensatory strategies the LF patients had to offer, and revealed a marked deficit in task switching. Finally, with respect to the lateralized character of the frontal impairment, it is important to note that there were no

Frontostriatal mechanisms of task control gross differences in the extent and aetiology of the cortical damage in the LF compared with the RF patients (see Methods, subjects section and Fig. 1). Moreover, there was little evidence that the LF patients showed greater cognitive deficits on the background neuropsychological tasks. In fact, it was the RF patients, and not the LF patients, who showed the significantly impaired strategy scores on the self-ordered search task (Miotto et al., 1996), a measure thought to capture an important ‘executive’ characteristic (Owen et al., 1990). Thus, there is no obvious feature of the clinical and cognitive profiles of the two frontal groups that explains the left-sided deficit in task control. Previous reports of laterality effects in the behavioural deficits of patients with frontal lesions have tended to be sporadic and inconsistent. With regard to motor performance, the apraxic difficulties of neurological patients are most often associated with left- rather than right-sided brain damage (Liepmann and Maas, 1907; Kimura and Archibald, 1974). However, studies with motor sequencing procedures designed to assess apraxic deficits have shown that specific difficulties with copying oral movement seem to follow damage to left frontal cortex while difficulties with copying limb movements follow damage to left parietal cortex (Kolb and Milner, 1981; Kimura, 1982). With regard to more explicitly cognitive function, perhaps the best evidence of specifically left-sided frontal involvement has been performance on verbal fluency tasks for which several studies have suggested greater deficits in patients with left-sided frontal damage than those with right-sided damage (Milner, 1964; Benton, 1968; Perret, 1974). On the other hand, while Shallice (1982) found greater impairments in planning cognition after left-sided anterior damage compared with right-sided, Owen et al. (1990) found roughly equivalent deficits. More recently, the use of brain imaging techniques has also shown specifically left-sided activations in normal volunteers using tasks originally designed for use with frontal lobe patients, including both verbal fluency (Frith et al., 1991) and the Tower of London Task (Owen et al., 1996b). Furthermore, investigations of the neural basis of working and episodic memory suggest differential left and right frontal cortical involvement in some of the underlying cognitive processes. For example, the rehearsal of verbal material in short-term verbal memory appears to activate left frontal cortex, particularly around Broca’s area (Paulesu et al., 1993; Petrides et al., 1995), while encoding and retrieval processes in episodic memory have been claimed to activate selectively left and right frontal cortex respectively (Kapur et al., 1994; Shallice et al., 1994; Tulving et al., 1994). Thus, there is some reason to believe that the left frontal cortex is preferentially activated by some clinically used neuropsychological tasks, including those characterized as ‘executive’ in character. However, other executive tasks seem to activate right frontal cortex preferentially, including one task analogous to the self-ordered search task used here as a background neuropsychological test (Owen et al., 1996c). The finding that some of the cognitive processes underlying

835

short-term verbal memory appear to depend on left frontal cortex highlights the possible connection between successful maintenance of task-relevant information ‘on-line’, in a highly available state (see Goldman-Rakic, 1987, 1994, 1996), and the effective performance of task-switching procedures. In this context, it is intriguing that correlations run over the forward and backward digit spans available for eight of the control subjects and their time cost in the colourcue and crosstalk condition failed to show a significant correlation with forward span (r 5 –0.11), but did show a significant negative correlation with backward-span (r 5 –0.74, P , 0.05), suggesting that smaller backward spans go together with larger time costs. Most versions of the working memory framework include a distinction between the rehearsal and maintenance of verbal material (through an ‘articulatory loop’) in a short-term storage device (or ‘phonological buffer’), and the selection and manipulation of this information by a separate executive system (e.g. Hitch, 1978; Baddeley, 1986). The lack of a relationship between forward span and the time cost of repeated task switches suggests that the control processes responsible for task control do not involve the mere maintenance of information in this phonological buffer. By contrast, the negative relationship between backward span and time cost suggests that some of the processes underlying successful backward digit span may also contribute to effective task control by, for example, co-ordinating rehearsal processes with the sequencing of responses required for output. Other candidate processes that have been related to the mnemonic function of frontal cortex include the integration of task and stimulus-relevant information within working memory (Kimberg and Farah, 1993), the successful maintenance over time of essentially task-context information (for review see Cohen et al., 1996), or other information relating to future events and actions (‘prospective’ memory; Cockburn, 1995). Clearly, the relationship between short-term memory and task control is an important area for research in executive function. With respect to the concept-formation tasks, the evidence for preferential left frontal cortical involvement is inconclusive. While some clinical studies with the WCST have suggested greater impairments following left frontal damage compared with right frontal damage (Milner, 1964; Drewe, 1974), others have demonstrated either the opposite pattern of results (e.g. Robinson et al., 1980), or no significant difference (Anderson et al., 1991). Moreover, neither early nor more recent brain imaging experiments with the WCST provide data which can resolve this issue (Weinberger et al., 1986; Berman et al., 1986, 1995). However, in this context, it is relevant that one recent PET experiment with normal, non-brain damaged subjects found that acquisition of visual discriminations involving shifts of attention towards previously irrelevant attributes (so-called ‘extra-dimensional shifts’; Owen et al., 1991, 1993; Dias et al., 1996) produced larger regional CBF activations in left polar frontal cortex and right superior frontal cortex than acquisition of discriminations not involving such shifts (so-called ‘intra-

836

R. D. Rogers et al.

dimensional shifts’). These data suggest that the process of over-riding an acquired attentional bias or ‘set’ towards reinforced attributes of stimuli, an important component of successful WCST performance, may involve distinct bilateral frontal cortical fields (R. D. Rogers, T. C. Andrews, P. M. Grasby, D. Brooks and T. W. Robbins, unpublished observations). Finally, complementary evidence, that switches between tasks require the integrity of left frontal cortex, has been obtained recently (J. Rubinstein, J. E. Evans and D. E. Meyer, unpublished observations). These researchers used the original switching procedure of Jersild (1927) to compare RTs when alternating between sorting WCST-like stimuli by shape, size or number with RTs when sorting by just one of these rules by itself. Significantly, patients with lesions of the left, but not the right, frontal cortex exhibited larger switch costs compared with control subjects, and especially so in the presence of interference between the two sorting rules. In the main, those results are congruent with the results presented here. However, as noted earlier, the ‘mixed’ block versus ‘pure’ procedure (Jersild, 1927; Spector and Biederman, 1976) provides a measure of ‘switch cost’ that may be contaminated by factors independent of the particular control processes required to shift between tasks. Consequently, the deficits demonstrated in neurological patients with this procedure cannot be attributed solely to the failure of control processes responsible for reconfiguring between task-sets but may, in addition, reflect the failure of distinct, but probably frontally mediated, executive processes required to keep two task-sets active simultaneously (see Corbetta et al., 1991; d’Esposito et al., 1995) and/or betweenblock differences in subject arousal or effort. Moreover the present results have isolated the specific deficits of LF patients in reconfiguring between well-established task-sets from the more general organizational deficits seen in both LF and RF patients when first introduced to task conditions that require the co-ordinated and sequential switching between two simple RT tasks.

Effects of mild Parkinson’s disease on task-set control In the main, there was little evidence that switches between simple letter- and digit-naming tasks were significantly harder for mild, medicated, early stage Parkinson’s disease patients compared with age- and IQ-matched control subjects. However, one intriguing result of this experiment is that the same Parkinson’s disease patients showed progressive increases in the error costs associated with switching between these tasks as a function of the number of trials performed in each block, while age- and IQ-matched control subjects showed only progressive reductions in the error cost. Both overall RT, and the time cost of task switches, also tended to show greater increases across trials compared with the control subjects. Like the LF patients, the Parkinson’s disease

patients were not generally impaired in their performance of the task-switching procedure; they were only impaired under certain experimental conditions. Specifically, the increases in the error costs of a task switch were absent in those conditions in which there was no interference or crosstalk between the tasks, but present in those conditions in which there was crosstalk, and they were not reduced in those conditions in which the task cues were words as opposed to the colour of the stimulus window. Thus, in contrast to several previous results involving Parkinson’s disease patients (e.g. Brown and Marsden, 1988b), there was no indication that this particular change in task performance could be reversed with stronger task cues that somehow ameliorated a deficit in the internal control of attention or action. The finding that, relative to control performance, the error cost of the patients differentially increased over successive trials is reminiscent of early clinical reports of parkinsonian fatigue in motor responding, as well as more recent experiments demonstrating disrupted force–time curves in the force characteristics of parkinsonian patients (e.g. Stelmach et al., 1989). For example, Schwab et al. (1959) and Schwab (1960) reported that the isotonic pressure exerted by repeated grasps of a ‘bulb’ ergograph reduced faster in Parkinson’s disease patients than in control subjects, suggesting that the persistence of a motor set might be subject to fatiguing processes in Parkinson’s disease. Convergent evidence for the effects of ‘time on task’ in the specific context of the control of cognitive processing is also available from studies requiring Parkinson’s disease patients to shift between entire blocks of trials, applying the Odd-Man-Out rule first to one stimulus set and then to another (Flowers and Robertson, 1985; Richards et al., 1993). These studies have shown that the error rates of Parkinson’s disease patients tend to increase over successive between-block shifts, together with the strength of an association between those error rates and the presence and severity of motor symptoms (e.g. extra-pyramidal signs). The patients who took part in those studies had, on the whole, higher Hoehn and Yahr ratings than the patients in the present experiment. Therefore, further experimentation is needed to explore the precise role of dopaminergic function in these trial-by-trial effects in Parkinson’s disease patients later in the course of the disease, and in studies involving the acute withdrawal of L-dopa medication. Finally, it is possible that task-switching experiments with more severely affected Parkinson’s disease patients will reveal more substantial increases in the switch costs associated with repeated shifts between tasks. However, in the context of the limited impairments seen in the present study, it may be important to note that the chosen tasks involved only vocal responses. These responses were chosen specifically to avoid the grosser disruptions to motor behaviour that might have affected, for example, manual response repertoires such as finger presses on a computer keyboard. One possibility is that parkinsonian deficits in task switching are exacerbated to the extent that executive control processes have to remap

Frontostriatal mechanisms of task control cognitive processes (e.g. response selections) with overtly disrupted forms of motor output (cf. Robertson and Flowers, 1990). Thus, future research needs to examine the relationship between the efficiency of task-switching performance in Parkinson’s disease patients and the mode of response.

Neural mechanisms of task-set control These results have several important implications for our understanding of the neural bases of task-set control, and in particular, the executive processes mediating dynamic shifts between cognitive tasks. In the first instance, some aspects of the data support the idea that the human frontal cortex makes its greatest impact on cognition early in practice (Knight, 1984; Raichle et al., 1994 Berns et al., 1997), when the novelty of situations requires the deployment of ruleinduction and problem-solving processes (Shallice and Burgess, 1996). Both the LF and RF patients showed slow, disorganized performance over the initial practice blocks of the colour-cue and crosstalk condition that then stabilized over the following four blocks of the test condition (see Fig. 6B and C). Thus, it appears that both the left and right frontal cortex appear to participate in the global organization of behaviour that is required when subjects are introduced to new, untried combinations of task conditions. In this context, it is interesting to note that acquisition of conditional rules to action (e.g. of the form ‘if condition X do action Y’) has been shown to be impaired in patients with frontal lesions (Petrides, 1985), and to activate frontal cortex (with the use of PET technology; Petrides et al., 1993). Similarly, in order to perform the colour-cue and crosstalk condition of this experiment, subjects had to establish essentially arbitrary connections between two discriminative cues (colours ‘green’ and ‘red’) and two response sets (‘letter-naming’ and ‘digitnaming’), and thereby learn to respond conditionally on the basis of the cue currently present in the display. These data suggest that damage to left and right frontal cortex disrupts this acquisition process. However, once learning has proceeded past some point, it is also clear that the time costs of the RF patients were not significantly different from those of the control subjects (see Fig. 6C) while those of the LF patients were increased, indicating a selective deficit in task control that resists at least limited amounts of practice. What kind of control mechanism does the left frontal cortex support? Recently, Rogers and Monsell (1995) proposed that reconfiguring from one task-set to another depends upon two distinct types of control operation. ‘Endogenous’ control operations can be initiated by the subjects in advance of an expected and imminent stimulus in order to ‘set’ themselves so that when the stimulus actually arrives, they have accomplished at least some of the reorganization of cognitive and motor resources needed to perform the new task. However, these endogeneous operations are capable of completing only part of the necessary reorganization. Additional ‘exogenous’ control operations are needed to complete the reconfiguring process, and these can only begin

837

with the presentation of the task stimulus associated with the new task (e.g. letters with letter-naming, Stroop stimuli with word naming). Once these exogenous control operations have been completed, typically, on the very first trial, the new task is fully operational. Thus, reconfiguring task-set is achieved by a mixture of executive mechanisms, some of which are dependent on task-relevant stimulus input, some which are not. [For a full discussion of the theory’s supporting data, see Rogers and Monsell (1995).] In this way, the common observation that stimuli activate a tendency to perform associated but currently inappropriate tasks, as seen in the mundane action slips or capture errors of everyday life (Reason, 1979, 1990) and in the occasional pathological ‘utilization behaviour’ of patients with frontal lesions (Lhermitte, 1983; Shallice et al., 1989) is conceptualized as the result of the unmodulated operation of exogenous taskset activity. We propose that the deficits seen here in LF patients involve the failure of an executive system responsible for endogenously controlling task-sets that are exogenously activated by task-relevant stimuli. Thus, when the display contains attributes that activate both the currently appropriate and inappropriate task-sets (as in the colour and crosstalk condition), some further control mechanism is needed to intervene, to modulate the resulting conflict and ensure that the correct task-set is selected. Where that intervening mechanism is damaged, it takes longer for this competition between the task-sets to be resolved appropriately, and so the time cost associated with reconfiguring to this task-set is increased. Notice that this deficit disappears under conditions in which the display contains attributes that activate only the currently appropriate task (as in the colour-cue and nocrosstalk condition), and is reduced in the presence of stronger, more direct, task cues (as in the word-cue and crosstalk condition). Therefore, we conclude that left frontal cortex, in tandem with other brain structures, plays an important role in the control of competing task-set activity in the context of reconfiguring task-set. However, other aspects of these data are informative about the further properties of this system and the way it is disrupted following damage to frontal cortex. First, task-relevant interference in the colour-cue and crosstalk condition did not significantly increase mean nonswitch RTs or error rates for the patients with left frontal cortical damage compared with either patients with right frontal cortical damage or control subjects. Therefore, it does not appear to be the case that the left frontal cortex is involved in the modulation of crosstalk between tasks per se but only when such crosstalk interferes with processes involved in reconfiguring between task-sets. Secondly, it is significant that, in the colour-cue and crosstalk condition, the patients with left frontal damage showed severely increased time costs of a task switch not only on trials on which there actually was interference between tasks but also on trials on which crosstalk was expected but, in fact, absent. This suggests that the fronto-cortical contribution to task-set

838

R. D. Rogers et al.

control is activated not only when there is a requirement to inhibit interference from particular responses that are currently activated by available stimuli, but rather by an incidence of interference over an extended period, during which interference may occur only inconsistently. Thus, it seems that left frontal control mechanisms, once activated by task-relevant interference, make a relatively tonic, and apparently rather inflexible, contribution to the executive control processes involved in the reconfiguring between different task-sets. Further clues about the nature of the executive control mechanism that is damaged by left frontal cortical lesions are available from the finding that, on switch trials, the LF patients were more sensitive than the control subjects to the facilitatory and inhibitory effects of having ignored the nowrelevant stimulus category on the preceding non-switch trial. Specifically, when there was interference on the current switch trial from a task-associated irrelevant character, the LF patients were slow to name a member of the just-ignored category. However, when there was no interference on the current switch trial, they were faster to name a member of the just-ignored category (see Fig. 4B). Both the inhibitory and facilitatory effects are consistent with the results of experimental research into the phenomenon of negative priming, i.e. the finding that responses to a stimulus can be slowed if the same or a related stimulus has recently been ignored (Allport et al., 1985). The finding that the slowing effect can be reversed in the absence of current interference has been demonstarted by several studies (e.g. Lowe, 1979; Allport et al., 1985) suggesting that the detection of negative priming depends on the subject being in some sort of ‘selection state’ (Tipper and Cranston, 1985). On this view, selective attention to one stimulus at the expense of another leaves the representations of both highly activated. Negative priming on subsequent trials is then apparent when the subject’s selection state temporarily prevents the representation of the just-ignored stimulus from making direct contact with associated response mechanisms (leading to slower responses). However, if the selection state is dropped, the activated representations of the ignored stimulus are allowed easy access to response mechanisms (leading to faster responses/positive priming). The exaggeration of these effects in the LF patients suggests that the mechanisms underlying the modulation of such facilitatory and inhibitory effects is disrupted in these patients. One possibility is that the behavioural demands of the colour-cue and crosstalk condition (rapid configuring between task-sets on every other trial in the presence of interference from task-associated distractors) require the coordinated enabling and disabling of the linkages between the stimulus representations of each task and their associated responses (i.e. stimulus–response linkages). Thus, shifts between task-sets might be achieved by excitation of the now-appropriate stimulus–response linkages, accompanied by inhibition of the now-inappropriate stimulus–response linkages. Effective performance then depends on the balance

of this excitation and inhibition being appropriate. Too much inhibition will slow subsequent shifts of task-set (as seen in the performance of the LF patients); too little will lead to instability between two task-sets. Clearly further research, using specifically tailored experimental procedures, is required in order to discover whether or not the left frontal cortex is indeed the site of such a modulatory mechanism. Although there was little sign of substantial deficits in the task switching of medicated, early stage Parkinson’s disease patients, careful analysis of trial-by-trial performance indicated significantly increasing error costs with increasing time on the task in the presence of crosstalk between the tasks. This effect raises a number of possibilities that require further investigation. The first is that dopaminergic transmission along the nigrostriatal pathway has an important modulating effect on those executive processes that sustain and, when necessary, reorganize currently activated motor and cognitive processes, especially in the face of task interference. In this context, it is particularly striking that the most evident change in the task-switching performance of the Parkinson’s disease patients was expressed in terms of the error cost as opposed to the time cost (cf. the LF patients). An increase in errors on switch trials relative to non-switch trials can be interpreted either as the tendency to respond on switch trials without reconfiguring to the new task-set at all, or to respond before task-set reconfiguration has been completed. Thus, these data may suggest that repeated task performance in Parkinson’s disease patients produces a progressive behavioural inflexibility, as if the effect of depleted dopamine in the striatum is to ‘lock up’ cortically initiated control operations that normally allow newlyreconfigured cognitive (e.g. attentional or decisional) processes to make contact with output mechanisms controlled in other portions of the corticostriatal circuitry. An alternative interpretation of the increase in error cost shown by the Parkinson’s disease patients relates to the necessary involvement of crosstalk between tasks. Perhaps, proactive interference from previous task-relevant stimuli accumulates and is able to disrupt switches from one task to another (see Allport et al., 1994), while depleted dopamine transmission into the striatum produces a failure to control this interference adequately on trials requiring a task shift. A final possibility is that practice through a sequence of trials might, once the task-sets have been effectively established, lead to reduced demands on attentional processes, which are then subject to increased disruption by distracting information (see Lavie, 1995). In summary, further research addressing the task-switching performance of Parkinson’s disease patients over series of trials may reveal important information about the neural bases, and functional characteristics, of the executive control processes that provide continuing control of task performance. Finally, the precise nature of the frontostriatal interactions in dynamic task control will also require further detailed study of patients with well-specified striatal damage to allow a full analysis of the contribution of the striatum to task-set

Frontostriatal mechanisms of task control control, as distinct from its dopamine innervation. In this context, it is interesting to note, given the dependence of increased time costs on the laterality of frontal damage, that seven of the Parkinson’s disease patients in the study presented with right unilateral onset, two with left unilateral onset and three with bilateral onset. Thus, the preponderance of right and bilateral onsets (i.e. greater left striatal dysfunction) suggests that dopaminergic depletion in the left striatum early in the course Parkinson’s disease does not simply reproduce the major deficit seen following damage to left frontal cortex. On the other hand, indications of marked task-switching deficits following cellular damage in Huntington’s disease is supplied in a study by Sprengelmeyer et al. (1995; their experiment 2) requiring left–right discriminations over letters and digits presented in a way similar to the stimuli of the present study. Huntington’s disease patients showed significantly increased RTs and error rates in mixed blocks of trials requiring alternation of tasks compared with pure blocks requiring only one task by itself. In the wider context, it may also be of practical value to examine the functions of cerebellar–cortical loops in taskswitching procedures, given the vital role of the cerebellum in motor programming and timing (Leiner et al., 1991). In any case, the use of this task-set reconfiguring paradigm has allowed the isolation of distinct executive mechanisms for the moment-by-moment control of task performance, some of which have been shown to depend on the left frontal cortex and to be modulated by dopamine dependent mechanisms of the striatum.

839

Sorting Test performance as a measure of frontal lobe damage. J Clin Exp Neuropsychol 1991; 13: 909–22. Baddeley AD. Working memory. Oxford: Clarendon Press; 1986. Benecke R, Rothwell JC, Dick JP, Day BL, Marsden CD. Performance of simultaneous movements in patients with Parkinson’s disease. Brain 1986; 109: 739–57. Benton AL. Differential behavioral effects in frontal lobe disease. Neuropsychologia 1968; 6: 53–60. Berman KF, Zee RF, Weinberger DR. Physiologic dysfunction of dorsolateral prefrontal cortex in schizophrenia. II. Role of neuroleptic treatment, attentio and mental effort [see comments]. Arch Gen Psychiatry 1986; 43: 126–35. Comment in: Arch Gen Psychiatry 1991; 48: 282–3. Berman KF, Ostrem JL, Randolph C, Gold J, Goldberg TE, Coppola R, et al. Physiological activation of a cortical network during performance of the Wisconsin Card Sorting Test: a positron emission tomography study. Neuropsychologia 1995; 33: 1027–46. Berns GS, Cohen JD, Mintun MA. Brain regions responsive to novelty in the absence of awareness. Science 1997; 276: 1272–75. Bianchi L. The mechanism of the brain, and the function of the frontal lobes. Edinburgh: Livingstone; 1922. Bowen FP, Kamienny RS, Burns MM, Yahr M. Parkinsonism: effects of levodopa treatment on concept formation. Neurology 1975; 25: 701–4. Brown RG, Marsden CD. An investigation into the phenomenon of ‘set’ in Parkinson’s disease [see comments]. Mov Disord 1988a; 3: 152–61. Comment in: Mov Disord 1990; 5: 178–9. Brown RG, Marsden CD. Internal versus external cues and the control of attention in Parkinson’s disease. Brain 1988b; 111: 323–45.

Acknowledgements The authors would like to thank Dr Stephen Monsell, Dr John Duncan, Dr Ros McCarthy and two anonymous reviewers for helpful comments on an earlier draft of this manuscript. This research was supported by a Programme Grant from the Wellcome Trust to Drs T. W. Robbins, B. J. Everitt, A. C. Roberts and B. J. Sahakian, and a separate Programme Grant to Professor C. Kennard.

Brown RG, Marsden CD. Dual task performance and processing resources in normal subjects and patients with Parkinson’s disease. Brain 1991; 114: 215–31. Brown VJ, Schwarz U, Bowman EM, Fuhr P, Robinson DL, Hallett M. Dopamine dependent reaction time deficits in patients with Parkinson’s disease are task specific. Neuropsychologia 1993; 31: 459–69. Burgess PW, Shallice T. Response suppression, initiation and strategy use following frontal lobe lesions. Neuropsychologia 1996; 34: 263–72.

References Alexander GE, DeLong MR, Strick PL. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. [Review]. Annu Rev Neurosci 1986; 9: 357–81. Allport DA, Tipper SP, Chmiel NRJ. Perceptual integration and postcategorical filtering. In: Posner MI, Marin OSM, editors. Attention and Performance XI. Hillsdale (NJ): Lawrence Erlbaum; 1985: 107–32. Allport A, Styles EA, Hsieh S. Shifting intentional set: exploring the dynamic control of tasks. In: Umilta C, Moscovitch M, editors, Attention and Performance XV. Cambridge (MA): MIT Press; 1994. p. 421–52. Anderson SW, Damasio H, Jones RD, Tranel D. Wisconsin Card

Channon S, Jones MC, Stephenson S. Cognitive strategies and hypothesis testing during discrimination learning in Parkinson’s disease. Neuropsychologia 1993; 31: 75–82. Cicerone KD, Lazar RM, Shapiro WR. Effects of frontal lobe lesions on hypothesis sampling during concept formation. Neuropsychologia 1983; 21: 513–24. Cochran WG, Cox GM. Experimental designs. 2nd ed. New York: John Wiley; 1957. Cockburn J. Task interruption in prospective memory: a frontal lobe function? Cortex 1995; 31: 87–97. Cohen JD, Braver TS, O’Reilly RC. A computational approach to prefrontal cortex, cognitive control and schizophrenia: recent

840

R. D. Rogers et al.

developments and current challenges. [Review]. Philos Trans R Soc Lond B Biol Sci 1996; 351: 1515–27. Cools AR, van den Bercken JH, Horstink MW, van Spaendonck KP, Berger HJ. Cognitive and motor shifting aptitude disorder in Parkinson’s disease. J Neurol Neurosurg Psychiatry 1984; 47: 443–53. Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE. Selective and divided attention during visual discriminations of shape, color, and speed: functional anatomy by positron emission tomography. J Neurosci 1991; 11: 2383–402. d’Esposito M, Detre JA, Alsop DC, Shin RK, Atlas S, Grossman M. The neural basis of the central executive system of working memory. Nature 1995; 378: 279–81. Delis DC, Squire LR, Bihrle A, Massman P. Componential analysis of problem-solving ability: performance of patients with frontal lobe damage and amnesic patients on a new sorting test. Neuropsychologia 1992; 30: 683–97. Dias R, Robbins TW, Roberts AC. Dissociation in prefrontal cortex of affective and attentional shifts. Nature 1996; 380: 69–72. Downes JJ, Roberts AC, Sahakian BJ, Evenden JL, Morris RG, Robbins TW. Impaired extra-dimensional shift performance in medicated and unmedicated Parkinson’s disease: evidence for a specific attentional dysfunction. Neuropsychologia 1989; 27: 1329–43. Downes JJ, Sharp HM, Costall BM, Sagar HJ, Howe J. Alternating fluency in Parkinson’s disease: an evaluation of the attentional control theory of cognitive impairment. Brain 1993; 116: 887–902. Drewe EA. The effect of type and area of brain lesion on Wisconsin Card Sorting Test performance. Cortex 1974; 10: 159–70. Drewe EA. Go-no go learning after frontal lobe lesions in humans. Cortex 1975; 11: 8–16. Duncan J. Disorganisation of behaviour after frontal lobe damage. Cogn Neuropsychol 1986; 3: 271–90. Eslinger PJ, Damasio AR. Severe disturbance of higher cognition after bilateral frontal lobe ablation: patient EVR. Neurology 1985; 35: 1731–41. Filoteo JV, Delis DC, Demadura TL, Salmon DP, Roman MJ, Shults CW. Abnormally rapid disengagement of covert attention to global and local stimulus levels may underlie the visuoperceptual impairment in Parkinson’s patients. Neuropsychology 1994; 8: 210–7. Flowers KA, Robertson C. The effect of Parkinson’s disease on the ability to maintain a mental set. J Neurol Neurosurg Psychiatry 1985; 48: 517–29. Frith CD, Friston KJ, Liddle PF, Frackowiak RS. A PET study of word finding. Neuropsychologia 1991; 29: 1137–48. Fuster JM. The prefrontal cortex. 2nd ed. New York: Raven Press; 1989. Goldman-Rakic PS. Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In: Mountcastle VB, Plum F, editors. Handbook of physiology, Sect. 1. Vol. 5, Pt. 1. Bethesda (MD): American Physiological Society; 1987. p. 373–417.

Goldman-Rakic PS. Prefrontal cortical dysfunction in schizophrenia: the relevance of working memory. In: Carroll BJ, Barret JE, editors. Psychopathology and the brain. New York: Raven Press; 1991. p. 1–23. Goldman-Rakic PS. The issue of memory in the study of prefrontal function. In: Thierry AM, Glowinski J, Goldman-Rakic PS, Christen Y, editors. Motor and cognitive functions of the prefrontal cortex. Berlin: Springer-Verlag; 1994. p. 112–21. Goldman-Rakic PS. The prefrontal landscape: implications of functional architecture for understanding human mentation and the central executive. [Review]. Philos Trans R Soc Lond B Biol Sci 1996; 351: 1445–53. Grafman J, Jonas B, Salazar A. Wisconsin Card Sorting Test performance based on location and size of neuroanatomical lesion in Vietnam with penetrating head injury. Percept Mot Skills 1990; 71: 1120–2. Grant DA, Berg EA. A behavioral analysis of degree of reinforcement and ease of shifting to new responses in a Weigltype card-sorting problem. J Exp Psychol 1948; 38: 404–11. Heck ET, Bryer JB. Superior sorting and categorizing ability in a case of bilateral frontal atrophy: an exception to the rule. J Clin Exp Neuropsychol 1986; 8: 313–6. Hermann BP, Wyler AR, Richey ET. Wisconsin Card Sorting Test performance in patients with complex partial seizures of temporal lobe origin. J Clin Exp Neuropsychol 1988; 10: 467–76. Hitch GJ. The role of short-term working memory in mental arithmetic. Cogn Psychol 1978; 10: 302–23. Hoehn MM, Yahr MD. Parkinsonism: onset, progression and mortality. Neurology 1967; 17: 427–42. Howell DC. Statistical methods for psychology. 2nd ed. Boston: Duxbury Press; 1987. Jenkins IH, Brooks DJ, Nixon PD, Frackowiak RS, Passingham RE. Motor sequence learning: a study with positron emission tomography. J Neurosci 1994; 14: 3775–90. Jersild AT. Mental set and shift. Arch Psychol 1927; No. 89. Joel D, Weiner I. The organization of the basal gangliathalamocortical circuits: open interconnected rather than closed segregated. [Review]. Neuroscience 1994; 63: 363–79. Josiassen RC, Curry LM, Mancall EL. Development of neuropsychological deficits in Huntington’s disease. Arch Neurol 1983; 40: 791–6. Kanehara. Ishihara’s design charts for colour blindness. Tokyo: Kanehara; 1992. Kapur S, Craik FI, Tulving E, Wilson AA, Houle S, Brown GM. Neuroanatomical correlates of encoding in episodic memory: levels of processing effect [see comments]. Proc Natl Acad Sci USA 1994; 91: 2008–2011. Comment in: Proc Natl Acad Sci USA 1994; 91: 1989–91. Kimberg DY, Farah MJ. A unified account of cognitive impairments following frontal lobe damage: the role of working memory in complex, organized behaviour. [Review]. J Exp Psychol: Gen 1993; 122: 411–28.

Frontostriatal mechanisms of task control Kimura D. Left-hemisphere control of oral and brachial movements and their relation to communication. Philos Trans R Soc Lond B Biol Sci 1982; 298: 135–49. Kimura D, Archibald Y. Motor functions of the left hemisphere. Brain 1974; 97: 337–50. Knight RT. Decreased response to novel stimuli after prefrontal lesions in man. Electroencephalogr Clin Neurophysiol 1984; 59: 9–20. Kolb B, Milner B. Performance of complex arm and facial movements after focal brain lesions. Neuropsychologia 1981; 19: 491–503. Lavie N. Perceptual load as a necessary condition for selective attention. J Exp Psychol: Hum Percept Perform 1995; 21: 451–68. Lawrence AD, Sahakian BJ, Hodges JR, Rosser AE, Lange KW, Robbins TW. Executive and mnemonic functions in early Huntington’s disease. Brain 1996; 119: 1633–45. Lees AJ, Smith E. Cognitive deficits in the early stages of Parkinson’s disease. Brain 1983; 106: 257–70. Lee CH, Brown R. Use of advance information in Parkinson’s disease. In: Vallar G, Cappa SF, Wallesch CW, editors. Neuropsychological disorders associated with subcortical lesions. Oxford: Oxford University Press; 1992. p. 190–203. Leiner HC, Leiner AL, Dow RS. The human cerebro-cerebellar system: its computing, cognitive and language skills. [Review]. Beh Brain Res 1991: 44: 113–28. Lhermitte F. ‘Utilization behaviour’ and its relation to lesions of the frontal lobes. Brain 1983; 106: 237–55. Liepmann H, Maas O. Fall von linksseitiger Agraphie und Apraxie bei Rechtsseitiger Lahmung. J Psychol Neurol 1907; 10: 214–27. Lowe DG. Stategies, context, and the mechanism of response inhibition. Memory Cogn 1979; 7: 382–27. Luria AR, Pribram, KH, Homskaya ED. An experimental analysis of the behavioral disturbance produced by a left frontal arachnoidal endothelioma (meningioma). Neuropsychologia 1964; 2: 257–80. Luria AR. Higher cortical functions in man. London: Tavistock; 1966. Malapani C, Pillon B, Dubois B, Agid Y. Impaired simultaneous cognitive task performance in Parkinson’s disease: a dopamine related dysfunction. Neurology 1994; 44: 319–326. Milner B. Effects of different brain lesions on card sorting. Arch Neurol 1963; 9: 100–10. Milner B. Some effects of frontal lobectomy in man. In: Warren JM, Akert K, editors. The frontal granular cortex and behaviour. New York: McGraw-Hill; 1964. p. 410–44. Miotto EC, Bullock P, Polkey CE, Morris RG. Spatial working memory and strategy formation in patients with frontal lobe excisions. Cortex 1996; 32: 613–30. Monsell S. Control of mental processes. In: Bruce V, editor, Mysteries of the mind: tutorial essays on cognition. Hove (UK): Lawrence Erlbaum; 1996. p. 93–148.

841

Moscovitch N, Winocur G. Frontal lobes, memory, and aging. [Review]. Ann NY Acad Sci 1995; 769: 119–50. Nelson HE. A modified card sorting test sensitive to frontal lobe defects. Cortex 1976; 12: 313–24. Nelson HE. National Adult Reading Test (NART) Test Manual. Windsor (UK): NFER-Nelson; 1982. Owen AM, Downes JJ, Sahakian BJ, Polkey CE, Robbins TW. Planning and spatial working memory following frontal lobe lesions in man. Neuropsychologia 1990; 28: 1021–34. Owen AM, Roberts AC, Polkey CE, Sahakian BJ, Robbins TW. Extra-dimensional versus intra-dimensional set shifting performance following frontal lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy in man. Neuropsychologia 1991; 29: 993–1006. Owen AM, Roberts AC, Hodges JR, Summers BA, Polkey CE, Robbins TW. Contrasting mechanisms of impaired attentional setshifting in patients with frontal lobe damage or Parkinson’s disease. Brain 1993; 116: 1159–75. Owen AM, Sahakian BJ, Semple J, Polkey CT, Robbins TW. Visuospatial short-term recognition memory and learning after temporal lobe excisions, frontal lobe excisions or amygdalo-hippocampectomy in man. Neuropsychologia 1995; 33: 1–24. Owen AM, Morris RG, Sahakian BJ, Polkey CE, Robbins TW. Double dissociations of memory and executive functions in working memory tasks following frontal lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy in man. Brain 1996a; 119: 1597–615. Owen AM, Doyon J, Petrides M, Evans AC. Planning and spatial working memory: a positron emission tomography study in humans. Eur J Neurosci 1996b; 8: 353–64. Owen AM, Evans AC, Petrides M. Evidence for a two-stage model of spatial working memory processing within the lateral frontal cortex: a positron emission tomography study. Cereb Cortex 1996c; 6: 31–8. Parent A, Hazrati L-N. Functional anatomy of the basal ganglia. I. The cortico-basal ganglia-thalamo-cortical loop. [Review]. Brain Res Rev Brain Res 1995; 20: 91–127. Paulesu E, Frith, CD, Frackowiak, RS. The neural correlates of the verbal component of working memory. Nature 1993; 362: 342–44. Perret E. The left frontal lobe of man and the suppression of habitual responses in verbal categorical behaviour. Neuropsychologia 1974; 12: 323–30. Petrides M. Deficits on conditional associative-learning after frontaland temporal-lobe lesions in man. Neuropsychologia 1985; 23: 601– 6-14. Petrides M, Alivisatos B, Evans AC, Meyer E. Dissociation of human mid-dorsolateral from posterior dorsolateral frontal cortex in memory processing. Proc Natl Acad Sci USA 1993; 90: 873–7. Petrides M, Alivisatos B, Evans AC. Functional activation of the human ventrolateral frontal cortex during mnemonic retrieval of verbal information. Proc Natl Acad Sci USA 1995; 92: 5803–7. Raichle ME, Fiez JA, Videen TO, MacLeod AM, Pardo JV, Fox

842

R. D. Rogers et al.

PT, et al. Practice-related changes in human brain functional anatomy during nonmotor learning. Cereb Cortex 1994; 4: 8–26.

Shallice T. From neuropsychology to mental structure. Cambridge: Cambridge University Press; 1988.

Reason J. Actions not as planned: the price of automatization. In: Underwood G, Stevens R, editors. Aspects of consciousness, Vol. 1. London: Academic Press; 1979. p. 67–89.

Shallice T. Multiple levels of control processes. In: Umilta C, Moscovitch M, editors. Attention and performance XV. Cambridge (MA): MIT Press; 1994. p. 395–420.

Reason J. Human error. Cambridge: Cambridge University Press; 1990.

Shallice T, Burgess P. The domain of supervisory processes and temporal organization of behaviour. [Review]. Philos Trans R Soc Lond B Biol Sci 1996; 351: 1405–12.

Richards M, Cote LJ, Stern Y. Executive function in Parkinson’s disease: set-shifting or set-maintenance? J Clin Exp Neuropsychol 1993; 15: 266–79.

Shallice T, Burgess PW, Schon F, Baxter DM. The origins of utilization behaviour. Brain 1989; 112: 1587–98.

Robertson C, Flowers KA. Motor set in Parkinson’s disease. J Neurol Neurosurg Psychiatry 1990; 53: 583–92.

Shallice T, Fletcher P, Frith CD, Grasby P, Frackowiak RS, Dolan RJ. Brain regions associated with acquisition and retrieval of verbal episodic memory. Nature 1994; 368: 633–5.

Robbins TW, Brown VJ. The role of the striatum in the mental chronometry of action: a theoretical review. Rev Neurosci 1990; 2: 181–213. Robinson AL, Heaton RK,Lehman RA, Stilson DW. The utility of the Wisconsin Card Sorting Test in detecting and localizing frontal lobe lesions. J Consult Clin Psychol 1980; 48: 605–14. Rogers RD. The costs of switching between cognitive tasks: a performance analysis [PhD thesis]. Cambridge: University of Cambridge; 1993.

Spector A, Biederman I. Mental set and mental shift revisited. Am J Psychol 1976; 89: 669–79. Sprengelmeyer R, Lange H, Homberg V. The pattern of attentional deficits in Huntington’s disease. Brain 1995; 118: 145–52. Stelmach GE, Teasdale N, Phillips J, Worringham CJ. Force production characteristics in Parkinson’s disease. Exp Brain Res 1989; 76: 165–72. Stuss DT, Benson DF. The frontal lobes. New York: Raven Press; 1986.

Rogers RD, Monsell S. Costs of a predictable switch between simple cognitive tasks. J Exp Psych: Gen 1995; 124: 207–31.

Talland GA, Schwab RS. Performance with multiple sets in Parkinson’s disease. Neuropsychologia 1964; 2: 45–53.

Rubinstein J, Meyer DE, Evans JE. Executive control of cognitive processes in task switching. J Exp Psychol Gen 1998. In press.

Taylor AE, Saint-Cyr JA, Lang AE. Frontal lobe dysfunction in Parkinson’s disease: the cortical focus of neostriatal outflow. Brain 1986; 109: 845–83.

Sahakian BJ, Morris RG, Evenden JL, Heald A, Levy R, Philpot M, et al. A comparative study of visuospatial memory and learning in Alzheimer-type dementia and Parkinson’s disease. Brain 1988; 111: 695–718.

Teuber HL, Battersby WS, Bender MB. Performance of complex visual tasks after cerebral lesions. J Nerv Ment Dis 1951; 114: 413–29.

Sandson J, Albert ML. Perseveration in behavioral neurology. Neurology 1987; 37: 1736–41. Schwab RS. Progression and prognosis in Parkinson’s disease. J Nerv Ment Dis 1960; 130: 556–66. Schwab RS, England AC, Peterson E. Akinesia in Parkinson’s disease. Neurology 1959; 9: 65–72. Schwartz MF, Reed ES, Montgomery M, Palmer C, Mayer NH. The quantitative description of action disorganisation after brain damage: a case study. Cogn Neuropsychol 1991; 8: 381–414. Shallice T. Specific impairments of planning. Philos Trans R Soc Lond B Biol Sci 1982; 298: 199–209.

Tipper SP, Cranston M. Selective attention and priming: inhibitory and facilitatory effects of ignored primes. Q J Exp Psychol 1985; 37A: 591–611. Tulving E, Kapur S, Craik FI, Moskovitch M, Houle S. Hemispheric encoding/retrieval asymmetry in episodic memory: positron emission tomography findings. Proc Natl Acad 1994; 91: 2016–2120. Weinberger DR, Berman KF, Zee RF. Physiologic dysfunction of dorsolateral prefrontal cortex in schizophrenia. I. Regional cerebral blood flow evidence [see comments]. Arch Gen Psychiatry 1986; 43: 114–24. Comment in: Arch Gen Psychiatry 1991; 48: 182–3. Received February 13, 1997. Revised November 10, 1997. Accepted December 4, 1997

Dissociating executive mechanisms of task control ... - Semantic Scholar

Dissociating executive mechanisms of task control ... - Semantic Scholar

Suggest Documents

Supramodal executive control of attention - Semantic Scholar

Mechanisms of control of Mycobacterium ... - Semantic Scholar

Neurocognitive mechanisms of cognitive control - Semantic Scholar

Dissociating the neural mechanisms of pain ... - EUROHEADPAIN

Dissociating semantic and phonological ... - Semantic Scholar

Dissociating Markers of Senescence and ... - Semantic Scholar

Dissociating stimulusdriven semantic and ... - Semantic Scholar

Dissociating stimulus-driven semantic and ... - Semantic Scholar

Task Scheduling for Control Oriented ... - Semantic Scholar

Dissociating person-specific from general ... - Semantic Scholar

Dissociating Barrel Development and Lesion ... - Semantic Scholar

Control mechanisms in mitochondrial oxidative ... - Semantic Scholar

Molecular Mechanisms to Control Post ... - Semantic Scholar

TCP Symbiosis: Congestion Control Mechanisms ... - Semantic Scholar

Executive Dysfunction Assessed with a Task ... - Semantic Scholar

Dissociating Verbal and Nonverbal Conceptual ... - Semantic Scholar

Dissociating Hippocampal versus Basal Ganglia ... - Semantic Scholar

Photorefractoriness in Mammals: Dissociating a ... - Semantic Scholar

Dissociating Memory Retrieval Processes Using ... - Semantic Scholar

Dissociating compatibility effects and distractor ... - Semantic Scholar

PTSD Modifies Performance on a Task of Affective Executive Control ...

Plasticity of executive control through task switching ... - ScienceOpen

Two Mechanisms for Task Switching in the ... - Semantic Scholar

Rule-Guided Executive Control of Response ... - Semantic Scholar