Abstract
The primate visual system contains two major streams of visual information processing. The ventral stream is directed into the inferior temporal cortex and is concerned with visual object cognition, whereas the dorsal stream is directed into the posterior parietal cortex and is concerned with visuospatial cognition. Both of these processing streams send projections to the basal ganglia, and the ventral stream may also receive reciprocal connections from the basal ganglia. Although a role for the basal ganglia in visual object and visuospatial cognition has been suggested, little work has been carried out in this area in humans. The primary site of neuropathology in Huntington's disease is the basal ganglia, and hence Huntington's disease provides an important model for the role of the human basal ganglia in visual object and visuospatial cognition, and its breakdown in disease. We examined performance on a wide battery of tests of both visual object and visuospatial recognition memory, working memory, attention, associative learning and perception, enabling us to specify more fully the role of the basal ganglia in visual object and visuospatial cognition, and the disruption of these processes in Huntington's disease. Huntington's disease patients exhibited deficits on tests of pattern and spatial recognition memory; showed impaired simultaneous matching and delay-independent delayed matching-to-sample deficits; showed spared accuracy but impaired reaction times in visual search; were impaired in spatial but not visual object working memory; and showed impaired pattern–location associative learning. The results of our investigations suggest a particular role for the striatum in context-dependent action selection, in line with current computational theories of basal ganglia function.
Introduction
A key feature of the primate visual system is the separation of visual areas into two major corticocortical processing pathways: the so-called dorsal and ventral processing streams (Ungerleider and Mishkin, 1982). The dorsal pathway is directed into the posterior parietal cortex and is important for spatial perception/localization (Ungerleider and Mishkin, 1982; but see Goodale and Milner, 1992). The ventral pathway is directed into the inferior temporal cortex and is important for visual object recognition (Ungerleider and Mishkin, 1982).
The basal ganglia have been suggested to form part of both dorsal (spatial) and ventral (object) visual processing streams (Baizer et al., 1993; Yeterian and Pandya, 1995), and in one view information from these sources may come together to bear upon motor processes, for the first time, in the striatum (Johnstone and Rolls, 1990). Indeed, there is some evidence from primate studies to suggest that the basal ganglia play a role in both complex visual object cognition and visuospatial cognition (for review, see Yeterian and Pandya, 1993, 1995). For example, neurons in the ventrocaudal striatum show firing patterns similar to those of neurons in the ventral stream from which they receive input (Caan et al., 1984; Brown et al., 1995) and lesions to the tail of the caudate and putamen in monkeys result in impaired performance of difficult visual discriminations (Divac et al., 1967; Buerger et al., 1974). In contrast, lesions to the anterodorsal portion of the striatum, which receives input from the dorsal stream, lead to deficits in spatial delayed alternation and spatial delayed response (Divac et al., 1967; Woodburne, 1971).
The primary site of neuropathology in Huntington's disease is the basal ganglia, and hence Huntington's disease provides an important model for the role of the human basal ganglia in visual object and visuospatial cognition, and its breakdown in disease. The aim of the present study was to examine the performance of clinically symptomatic Huntington's disease patients, early in the course of their disease, on a wide-ranging battery of tests of visual object and visuospatial processing, in order to help clarify the role of the basal ganglia in such cognitive processes. We examined performance on tests of visual object and visuospatial perception, recognition memory, working memory, attention and associative learning.
To assess visual object and visuospatial perceptual processes, we used the Visual Object and Space Perception Battery (VOSP) (Warrington and James, 1991), which includes a number of tasks designed to assess specific, dissociable aspects of object and space perception, relatively independently of other cognitive and motor processes.
To examine recognition memory, we employed a test of pattern recognition memory based on the serial recognition tasks used to assess recognition memory in monkeys (Gaffan, 1974) and an analogous spatial recognition memory task, which in functional imaging experiments (Owen et al., 1996b) have been shown to activate the inferior temporal cortex and posterior parietal cortex, respectively.
In addition, we employed a more sophisticated test of recognition memory. The classical test of stimulus recognition memory is delayed matching-to-sample (DMTS) or its monkey analogue, delayed non-matching-to-sample. This task has been much used to study the functions of the medial temporal lobe memory system in monkeys (Murray, 1996). The present study employed a trial-unique DMTS task with varying delays. Such a task is optimally designed to differentiate true visual mnemonic function from other functions (Horel, 1994). A true deficit in visual memory would result in Huntington's disease patients forgetting more than controls, i.e. the decline in memory as a function of delay would not be parallel in patients and controls. Deficits in DMTS could potentially arise from problems other than memory, for example in associative memory (i.e. learning the matching rule) or in visual attention. To examine such functions, a control task of matching-to-sample/visual search was included, using stimuli similar to those used in the DMTS task.
Recent experiments in primates suggest that the basal ganglia may also form part of the circuitry for visual object and visuospatial working memory (Goldman-Rakic, 1995; Levy et al., 1997). We have shown previously that there are impairments in spatial working memory in Huntington's disease (Lawrence et al., 1996) on a self-ordered task which in functional imaging experiments has been shown to activate regions of the dorsolateral prefrontal cortex and posterior parietal cortex (Owen et al., 1996a), which send projections to the dorsal striatum (Selemon and Goldman-Rakic, 1988), suggesting a role for the striatum in spatial working memory in humans. To extend this work, in the present study we employed both the spatial working memory task used by Lawrence and colleagues (Lawrence et al., 1996) and an analogous visual object working memory task, which is sensitive to damage to the temporal lobe, but not the frontal lobe, in humans (Owen et al., 1996c), to see if the deficit in spatial working memory in Huntington's disease was also present in the visual object domain.
In addition to assessing visual object and visuospatial functions separately in Huntington's disease, we also examined performance on an associative learning task that involved learning to associate visual objects with spatial locations. Given that the basal ganglia receive input from both the dorsal and the ventral streams, such associative learning could, in part, be mediated via the basal ganglia (Parker and Gaffan, 1998).
For many of the tasks selected, neuroimaging studies have determined the neural networks involved in their performance (Owen et al., 1996a, b; Elliott and Dolan, 1999). Therefore, the use of such a comprehensive, well-validated battery of tests, allowing the assessment of both visual object cognition and visuospatial cognition, both independently and cooperatively, helps to specify more fully the role of the basal ganglia in visual object and visuospatial cognition and the disruption of these processes in Huntington's disease. The results of our investigations suggest a particular role for the striatum in the mechanisms of context-dependent action selection, in line with current computational theories of striatal function (Lawrence et al., 1998).
Methods
Subjects
Patients
Three groups of Huntington's disease patients participated in this experiment. One group consisted of 19 patients with genetically confirmed, clinically symptomatic Huntington's disease. These patients were tested on DMTS, visual search/matching-to-sample (VSMTS), visual working memory and the VOSP Battery (Warrington and James, 1991). Their mean age was 47.5 years (SD 9.8 years). Patients were recruited from the Cambridge Huntington's Disease Clinic at Addenbrooke's Hospital. The mean age at onset was 42.5 years (SD 9.2 years) and the mean disease duration was 5 years (SD 2.7 years). Five patients were maintained on anti-dopaminergic medication, two on benzodiazepines and two on antidepressant medication. Patients were administered the functional assessment scale of the Unified Huntington's Disease Rating Scale (UHDRS) (Huntington Study Group, 1996). The patients' mean score was 21.5 (SD 3.4), indicating mild disability. Their mean score in the Mini-Mental State Examination (MMSE) (Folstein et al., 1975) was 28.3 (SD 2.2) out of a possible maximum of 30. None of the patients scored below the cut-off score for dementia of 24. CT scans were not available for any subject.
A second group of Huntington's disease patients was tested on the paired-associates learning task (PAL). This group consisted of 19 subjects, 11 of whom were also in the above group of genetically confirmed cases of Huntington's disease. A further eight subjects were recruited from the Department of Clinical Neurology at the National Hospital, London, and were diagnosed on the basis of clinical signs and a positive family history (Folstein et al., 1986). The mean age of this group was 42.7 years (SD 11.3 years) and the mean age at onset was 38.6 years (10.4 years). The 11 Cambridge subjects were administered the UHDRS activities of daily living questionnaire and the MMSE test as above, scoring 21.2 (SD 3.5) and 28.1 (SD 2.7), respectively. The eight London subjects were administered the chorea scale of the Quantified Neurological Examination (David et al., 1987). This scale ranges from 0 to 25 (high scores denoting severe impairment) and the patients had a mean score of 6.1 (SD 1.4). In addition, the London group were administered the Kendrick Object Learning Test, a dementia screening test taken from Kendrick's cognitive tests for the elderly (Kendrick, 1985); their mean score was 35.9 (SD 2.5), which is well above the cut-off score for dementia of 22.
A third group of Huntington's disease patients was tested on the pattern and recognition memory tasks, and also the spatial working memory task. This group consisted of 21 subjects, the majority of whom were also in the above two groups. The mean age of this group was 41.8 years (SD 9.2 years). Patients were recruited from the Cambridge Huntington's Disease Clinic, Addenbrooke's Hospital. The mean age at onset was 39.5 years (SD 8.3 years) and the mean disease duration was 5 years (SD 1.9 years). Patients were administered the UHDRS and MMSE as above, with mean scores of 21.58 (SD 3.14) and 28.33 (SD 2.53), respectively.
Controls
Three groups of control subjects were chosen to match the patient groups as closely as possible with respect to both age and premorbid verbal IQ, as measured by the National Adult Reading Test (NART) (Nelson, 1982). One group was tested on DMTS, paired-associates learning and visual working memory; the second group was tested on just the visual search task. A third group, several of whom were also included in group 1, was tested on the pattern and spatial recognition memory and spatial working memory tasks. A number of subjects were included in more than one group. These subjects were drawn from a pool of control volunteers recruited through advertisements placed in the Cambridge employment centre or from the North East Age Research Panel in Newcastle-upon-Tyne. Unpaired t-tests revealed that the groups did not differ in terms of age or estimated premorbid verbal IQ.
(i) For those Huntington's disease patients tested on all tasks except the PAL versus controls: age, t(37) = 0.95, P > 0.3; NART, t(37) = 1.26, P = 0.2.
(ii) For those Huntington's disease subjects tested only on PAL versus controls: age, t(37) = 0.99, P > 0.3; NART, t(37) = 1.0, P > 0.3.
(iii) For those Huntington's disease subjects tested on pattern and spatial recognition memory and spatial working memory: age, t(36) = 1.03, P = 0.31; NART, t(36) = 1.34, P = 0.19. Full details of patient and control subject characteristics are presented in Table 1.
This study was approved by the Cambridge University Research Ethics Committee and all subjects gave written, informed consent.
Procedure
All computerized tests were carried out with a portable CarryI 486 microcomputer (CarryI, Taiwan) fitted with a Datalux© touch-sensitive screen. Six of the tests, using pattern and spatial recognition memory, spatial working memory, visuospatial paired-associates learning (PAL), delayed matching-to-sample (DMTS) and visual search/matching-to-sample (VSMTS) were taken from the CANTAB© battery. The other computerized task, visual working memory, was designed by A. M. Owen and T. W. Robbins. Finally, the Visual Object and Space Perception Battery (VOSP) (Warrington and James, 1991) was administered in booklet form. Screen shots from the computerized tests are shown in Fig. 1.
Pattern and spatial recognition (Sahakian et al., 1988)
Two tasks designed to assess recognition memory for both patterns and spatial locations were administered. In the pattern recognition task, subjects were presented with a series of 12 abstract patterns, and their task was to remember them. Following a delay of 5 s, the 12 patterns were re-presented in reverse order, paired with a novel pattern, and subjects were asked to touch the pattern they had seen previously. This procedure was then repeated with a further 12 patterns.
In the spatial recognition task, five squares were presented sequentially in different locations around the screen. In the recognition phase, each location was re-presented, paired with a novel location, and subjects were asked to touch the location at which they had seen a square appear. This procedure was repeated a further three times.
Delayed matching-to-sample (Owen et al., 1995)
At the outset of each trial, a sample stimulus—a complex, abstract visual stimulus consisting of four quadrants, each differing in colour and form—appeared in the centre of the screen for 4.5 s. Subjects were told to study the pattern carefully as they would be required to identify it later from amongst four patterns (including three distractor patterns).
In the simultaneous matching condition (SMTS), four choice patterns appeared beneath the sample, one in each of four white boxes. Subjects had to touch the one choice stimulus that matched the sample exactly. One of the other choices was a novel distractor differing in both form and colour from the sample. The remaining two choices were `partial distractors', in that one shared the same colours as the sample but not its form, and vice versa. Additionally, each of the four choices had one (random) quadrant in common to discourage the use of mnemonic strategies based on remembering the colour and form of a single quadrant of the stimulus. Feedback was provided for each response in the form of a green cross (correct) or red tick (wrong) accompanied by a high- and a low-pitched tone, respectively. After an incorrect touch, each subject had to continue until the matching choice stimulus was touched.
The delayed matching condition (DMTS) was identical to the SMTS condition in every way with the exception that, after initial presentation, the sample stimulus disappeared from the screen and there was then a delay of 0, 4, or 12 s before presentation of the choice stimuli. Thus, the subjects had to remember the sample pattern in order to pick the matching choice stimulus. The test began with three practice trials (SMTS, 0 s DMTS, 12 s DMTS), which were followed by 40 test trials consisting of 10 trials of each condition (SMTS; 0, 4, 12 s, DMTS) scheduled in fixed, pseudorandom order.
Subjects were scored according to the proportion of trials correct on the first choice, latencies to respond and type of error made (Ef, choosing form distractor; Ec, choosing colour distractor; En, choosing novel distractor).
Visual search/matching-to-sample (VSMTS) (Downes et al., 1989)
In addition to the use of the touch-sensitive screen, this task required subjects to use a non-latching switchpad fitted with a lighter spring for comfortable hand operation, which was attached to the microcomputer via an interface lead to the user port.
The stimuli used were similar to those used in the DMTS task, with the exception that one quadrant was not common to all. Depression of the hand-held switchpad opened a central red box to reveal the sample stimulus, and 2 s later the choice stimuli (1, 2, 4 or 8) appeared in an array of eight surrounding white boxes. This procedure allowed the subjects to set their own pace throughout the task. As with DMTS, one stimulus was identical to the sample. Initially, the subjects were shown the array of filled boxes and were instructed to find the choice stimulus matching the sample as quickly as possible. The subjects were required to hold down the switchpad and, when they were ready to make a response, to release the pad and touch the matching choice stimulus. This procedure allowed the calculation of both choice reaction time (latency to remove the hand from the press-pad) and movement time (latency to move from pad to screen). Auditory and visual feedback were presented as in the DMTS task, and practice trials involving four examples in order of increasing set size (1, 2, 4, 8) were given before 48 test trials (12 of each set size, presented in fixed, pseudorandom order).
The proportion of correct choices, as a function of set size, response and movement latency, was scored for each subject.
Pattern–location paired-associates learning (Sahakian et al., 1988)
Subjects were required to remember up to eight pattern–location associations. Initially, six white boxes appeared around the screen and the subject was told that each of them would open in turn, revealing what was inside. The task was to look for coloured abstract patterns in the boxes and to remember which pattern appeared in which box. Each of the boxes opened for 3 s and then closed, in a randomized sequence. On the initial trial, only one of the boxes contained a pattern. Immediately after the last box had opened, this pattern was presented in the centre of the screen and the subject was required to respond by touching the box in which this pattern appeared. If the choice was incorrect, the boxes were successively reopened for 2 s each and the subject was allowed a second attempt to touch the correct box. On each trial, the subject was allowed a maximum of 10 attempts before the test was terminated.
After the initial trial with 1 pattern, there was another 1-pattern trial, then two 2-pattern trials, two 3-pattern trials, and then one 6-pattern trial. Finally, two additional boxes were added to the array on the screen and the subject was required to locate 8 patterns correctly.
Performance was assessed according to three measures. (i) `Trials' represents the total number of presentations required to locate all the patterns correctly, summed over the entire task. (ii) `Errors' represents the total number of errors (incorrect placements) summed across the whole task. (iii) `Memory score' represents the total number of patterns located after the first presentation, summed over the entire 8 trials of the task (range 0–26).
Visual object working memory (Owen et al., 1996c)
The subject was required to search through an array of difficult-to-verbalize abstract coloured shapes presented on the touch-sensitive screen (by touching each shape in turn) in order to locate a blue token hidden behind one of the shapes, and once found, to place it in a store on the right-hand side of the screen. Subjects were instructed that only one token was hidden at any particular time, and once they had found the token and placed it in the store, another token would be hidden. It was made explicit that, once a token had been found behind a particular shape, that shape would not be used again to hide a token, and they should avoid returning there. If the subject made consecutive responses to the same shape within a search sequence, a bleep sounded.
Every shape was used once, so that on every trial the number of tokens to be found corresponded to the number of shapes on the screen. The shapes to be searched altered their spatial locations after each response, ensuring that the test could not be solved on the basis of spatial cues. The number of incorrect shapes selected (excluding errors) before a token was found was determined by the computer.
Each shape was trial-unique, and within each trial all shapes were the same colour. After successful completion of each trial, this colour changed. There were three levels of difficulty. The subjects completed two practice 3-shape trials initially, then four 4-shape trials and finally four 6-shape trials. Performance was assessed in terms of between- and within-search errors. Between-search errors represent a return to a shape behind which a token has been found on a previous search in the same trial. Within-search errors represent a return to a shape which has already been found not to hide a token within the same search sequence.
Spatial working memory (Owen et al., 1996c)
This is a test of spatial working memory for humans, which is formally analogous to the Olton radial arm maze (Olton et al., 1979), an optimal foraging-for-reward task. In this task, subjects were required to search through a number of coloured boxes presented on the monitor screen (by touching each one) in order to find blue tokens (reinforcers) which were hidden inside. On any one trial, only a single token was hidden in one of the boxes. Once found, the next token was hidden. The key instruction was that, once a token had been found within a particular box, that box would not be used again to hide a token. Two types of error were possible. First, a subject may have returned to open a box in which a token had already been found (between-search error). Secondly, a subject may have returned to a box already opened and shown to be empty earlier in the same trial (within-search error). There were four trials with each of 4, 6 and 8 boxes. The task was scored according to the number of between- and within-search errors at each level of difficulty and also for the use of an efficient search strategy (Owen et al., 1996c). A particularly efficient strategy for completing this task is to follow a predetermined search sequence, starting with a particular box and then returning to start each new sequence with that same box as soon as a token is found (editing the sequence when a token is found in that box). The extent to which such a strategy was used was estimated from the number of search sequences starting with a novel box for just the more difficult 6- and 8-box problems. The total of these scores provided a measure of strategy for each subject, a high score (many sequences starting with a novel box) representing poor use of a strategy and vice versa.
Visual Object and Space Perception Battery (Warrington and James, 1991)
This battery consisted of four visual-object and four space perception tests, preceded by a basic visual screening task. All stimuli are presented black-on-white in booklet form. Each task is devised to focus on one particular component of visual perception, whist minimizing the involvement of other cognitive skills. Full details of all tests can be found in the manual provided with the test (Warrington and James, 1991). All subjects received the tests in the order described in the VOSP instruction manual, visual object perception being tested before space perception.
Results
Pattern and spatial recognition memory
For data on recognition memory accuracy, analysis of variance (ANOVA) with test (pattern and spatial) as the within-subject factor and group as the between-subject factor revealed a statistically significant effect of test [F(1,36) = 98.07, P < 0.001] and a significant effect of group [F(1,36) = 14.72, P < 0.001], but no group × test interaction [F(1,36) = 0.36, P = 0.55]. Thus, patients were impaired relative to controls, but not differentially with respect to pattern/spatial recognition. The data are represented graphically in Fig. 2. Of note is that pattern recognition and spatial recognition were themselves correlated (r = 0.62, P < 0.01) in Huntington's disease patients but not in controls (r = 0.05, n.s.).
In terms of response latency, there was a statistically significant effect of group [F(1,36) = 8.48, P = 0.006] and a marginally significant effect of test [F(1,36) = 3.88, P = 0.056], but no interaction [F(1,36) = 0.31, P = 0.58]. Mean latencies for pattern recognition were 3276.0 ms (SEM 242.5) for Huntington's disease subjects and 2088.5 ms (SEM 106.9) for control subjects. Mean latencies for spatial recognition were 3491.2 ms (SEM 401.4) for Huntington's disease patients and 2474.2 ms (SEM 251.4) for control subjects.
Delayed matching-to-sample
Data were analysed separately for the simultaneous and delayed matching conditions. Data were analysed in terms of the proportion correct scores, and were then arcsin-transformed [y = 2arcsin(p1/2)] to meet the assumptions of ANOVA.
In simultaneous matching, Huntington's disease patients performed significantly less well than controls [t(35) = 2.57, P = 0.015]. They were also slower to respond than controls [t(35) = 4.17, P < 0.001].
In delayed matching, repeated measures ANOVA with group and delay as factors revealed a statistically significant main effect of group [F(1,35) = 10.97, P = 0.002], a significant effect of delay [F(2,70) = 9.51, P < 0.001], but no interaction between group and delay [F(2,70) = 1.11, P = 0.33]. Post hoc t-tests revealed that Huntington's disease patients were impaired at both the 4-s delay [t(35) = 3.05, P = 0.004] and the 12-s delay [t(35) = 2.58, P = 0.014], but the 0-s delay failed to reach statistical significance [t(35) = 1.68, P = 0.1]. Of note, performances in SMTS and DMTS were not significantly correlated (r = 0.3).
In terms of response latency, again there was a statistically significant effect of group [F(1,35) = 21.23, P < 0.001] and a significant effect of delay [F(2,70) = 18.45, P < 0.001], but no interaction between the two factors [F(2,70) = 1.42, P = 0.25].
Data for both the simultaneous and the delayed matching conditions are presented graphically in Fig. 3A and B.
An additional analysis of error types was undertaken to see if any error type was particularly associated with poor performance. In general, the subjects made more form errors (Huntington's disease patients, 67%; control subjects, 67%) than either colour (Huntington's disease patients, 23.1%; control subjects, 28.8%) or distractor errors (Huntington's disease patients, 9.8%; control subjects, 4.2%]. The percentages of errors of different types did not differ between patients and controls, although there was a tendency for Huntington's disease patients to make more distractor errors [P = 0.07], i.e. responses to stimuli of incorrect colour and form.
Visual search/matching-to-sample
The proportions correct at set sizes 4 and 8 were included in a repeated measures ANOVA with group as the between-subject factor. Data were arcsin-transformed to meet the assumptions of ANOVA. Performance decreased significantly with increasing set size [F(1,36) = 4.22, P = 0.047], but there was no effect of group [F(1,36) = 1.06, P = 0.31] and no interaction between the two factors [F(1,36) < 0.001, P = 0.98]. Data are presented graphically in Fig. 4A.
Reaction times were also submitted to repeated measures ANOVA. There was a significant effect of group [F(1,36) = 31.52, P < 0.001], a significant effect of set size [F(1,36) = 114.01, P < 0.001] and a significant interaction between the two factors [F(1,36) = 6.12, P = 0.018]. Analysis of simple effects revealed that Huntington's disease patients were slower than controls at set sizes of both 4 [t(36) = 5.96, P < 0.001] and 8 [t(36) = 5.18, P < 0.001], the Huntington's disease patients showing evidence of disproportionate slowing as the number of choices increased (Fig. 4B).
Paired-associates learning
Huntington's disease patients performed significantly less well than control subjects on all three measures of task performance: total trials, t(35) = 2.78, P = 0.009; total errors, t(35) = 2.61, P = 0.01; memory score, t(35) = 2.34, P = 0.025. Data are presented graphically in Fig. 5.
Visual object working memory
Between- and within-search errors were subjected to repeated measures ANOVA after square-root [y = (x + 0.5)1/2] transformation, with group and level (4 or 6 shapes) as factors. For between-search errors, there was a statistically significant increase in errors as a function of level [F(1,30) = 57.24, P < 0.001], but no effect of group [F(1,30) = 1.56, P = 0.22] and no interaction between the two factors [F(1,30) = 0.39, P = 0.54]. Similarly, for within-search errors there was an effect of level [F(1,30) = 22.36, P < 0.001] but no effect of group [F(1,30) = 1.66, P = 0.21] and no interaction [F(1,30) = 0.24, P = 0.63]. Between- and within-search errors are presented graphically in Fig. 6.
Spatial working memory
Between-search errors were subjected to repeated measures ANOVA after square-root [y = (x + 0.5)1/2] transformation, with group and level (4, 6 or 8 boxes) as factors. This revealed a statistically significant effect of group [F(1,36) = 6.24, P = 0.017], a significant effect of level [F(1,36) = 132.72, P < 0.001] and a statistically significant interaction term [F(1,36) = 3.42, P = 0.038]. Analysis of simple effects revealed that Huntington's disease patients made significantly more errors than controls only at the 8-box level [t(36) = 3.03, P = 0.005]. Data are presented in Fig. 7.
In terms of strategy score, Huntington's disease patients made slightly less use of the most efficient strategy than did controls [t(36) = 1.85, P = 0.07]. Strategy scores correlated negatively with between-search errors in both groups (Huntington's disease patients, r = 0.7, P < 0.01; control subjects, r = 0.5, P < 0.05). The mean strategy score for Huntington's disease patients was 35.5 (SEM 1.2), whereas for controls it was 32.2 (SEM 1.2).
Within-search errors were distributed non-normally, and so they were subjected to non-parametric analysis (Mann–Whitney U-test). Total within-search errors did not differ significantly between the groups [U = 151, P = 0.41]. The mean number of within-search errors for Huntington's disease patients was 2.76 (SEM 0.54), whereas for controls it was 2.58 (SEM 0.89).
Visual Object And Space Perception Battery
The number of subjects performing below the fifth percentile of control norms in the Huntington's disease group was examined. Only performance on the object decision was impaired, seven of 18 Huntington's disease patients performing below the fifth percentile on the object decision task.
Correlation analysis
To examine the relationship between clinical measures and cognitive performance, second-order partial correlations (partialling for age and NART IQ) were computed between clinical and cognitive variables. Activities of daily living scores correlated reliably with DMTS accuracy, collapsed across delays (r = 0.72, P = 0.001). Disease duration correlated with response latencies on the DMTS task, collapsed across delays (r = 0.66, P = < 0.01).
Discussion
This study represents the first detailed analysis of visual object and visuospatial processing in early-stage Huntington's disease, and has resulted in a number of novel findings with important implications for information processing within anatomically defined corticostriatal circuits.
The nature of impaired performance in Huntington's disease
Perception
The poor performance exhibited in the SMTS condition of the DMTS task by Huntington's disease patients raises the issue of possible perceptual disturbances. Perceptual deficits are possibly responsible for the SMTS deficit but, importantly, SMTS and DMTS performance were not reliably correlated, suggesting that a perceptual deficit is not responsible for impaired DMTS performance. In addition, choice accuracy was unimpaired on the VSMTS task, which places a heavy load on perceptual processing. At least some of the differences in simultaneous matching performance compared with the MTS visual search task are probably due to the use of stimuli in the former which included stimuli with common elements, thus making the choice alternatives less discriminable.
Perceptual deficits have been reported previously in Huntington's disease (Biber et al., 1981; Hodges et al., 1991; Büttner et al., 1994; Jacobs et al., 1995; Sprengelmeyer et al., 1996). Impaired perceptual function in Huntington's disease patients might also be inferred from poor performance on the VOSP Battery. As performance was within normal limits on the majority of subtests from the VOSP, a generalized perceptual dysfunction does not occur in Huntington's disease. Rather, Huntington's disease patients were impaired only on the object decision task, which requires the recognition of visually degraded objects presented from an unusual view, and is the most demanding discrimination problem in the VOSP Battery, placing demands on attentional and decisional mechanisms in addition to perceptual processes. Thus, the deficit in Huntington's disease does not appear to be one of perception per se, but rather in attentional and/or decisional processes.
Recognition memory and attention
Consistent with our earlier study (Lawrence et al., 1996), patients with Huntington's disease were impaired on both pattern and spatial recognition memory. This might suggest that both visual and spatial short-term memory processes are impaired in Huntington's disease. Indeed, a recent metabolic imaging study in behaving primates, using tasks similar to the pattern and spatial recognition memory tasks used here, showed increased metabolic activity in the dorsal striatum during performance of a spatial delayed response task and in the ventrocaudal striatum during performance of a delayed object alternation task (Levy et al., 1997), suggesting a role for the striatum in short-term memory processes in both the spatial and object domains (but see Postle and D'Esposito, 1999).
Performance on the two tasks was correlated reliably in patients but not in control subjects, suggesting a processing deficit common to the two tasks in Huntington's disease. In addition to domain-specific memorial processes, the tasks require a series of binary decisions to be made as to whether a visual object or spatial location has been seen previously, thus placing demands on attentional and/or selection processes. Thus, the deficits in Huntington's disease in these tasks might be at the level of attention/selection rather than memory per se. This interpretation is supported to some extent by the findings from the DMTS task. The version of the DMTS task used here systematically varies the load on memory by imposing different delays in a counterbalanced series of trials, enabling forgetting curves (performance versus delay) to be calculated and compared between groups. There was no group × delay interaction, suggesting parallel rates of forgetting in patients and controls. However, the patients' performance was worse overall than that of controls, and they were also impaired on simultaneous matching when there was no delay between the presentation of the stimulus and choice array. Thus, some factor other than poor memory must be the cause of the impaired performance of the Huntington's disease patients.
The performance of patients with Huntington's disease thus contrasts with that of temporal lobe patients on the same task. Temporal lobe patients show normal simultaneous matching but are impaired on delayed matching. Frontal lobe patients are unimpaired on this task, although there is some evidence of an impairment at zero delay in patients with bilateral frontal lobe lesions (Owen et al., 1995). The performance of Huntington's disease patients most resembles that of Parkinson's disease patients, who also show delay-independent deficits on this task (Sahakian et al., 1988), suggesting that the deficit on matching-to-sample seen in basal ganglia patients is not mnemonic in nature. Of note, a recent functional MRI study has shown an increased BOLD (blood oxygenation level-dependent) signal in the caudate nucleus related to delayed matching performance, independent of delay (Elliott and Dolan, 1999).
In order to perform a matching-to-sample task, individuals must perform a number of presumably dissociable computations (Desimone, 1996; Meunier et al., 1997; Moody et al., 1998). First, they must attend to and discriminate among the different stimuli. Secondly, they must retain the memory of the sample for the length of the trial. Thirdly, they must make a comparison between and reach a decision about whether the current test stimuli match the sample held in memory, and finally they must make the appropriate response selection. In addition, over a large number of trials, the subject must acquire the DMTS principle that familiar stimuli are associated with reinforcement. The finding that Huntington's disease patients are impaired even at simultaneous matching and do not show delay-dependent deficits suggests that the problem that they have relates not to memory but to one or more of these other components of DMTS.
The design of the DMTS stimuli enables detailed error analysis, which helps to specify in more precise terms the nature of the deficit. This analysis indicated that, when errors were made, subjects remembered the correct colours (but not shapes) about two-thirds of the time, and correct shapes (but not colours) a little under one-third of the time. However, Huntington's disease patients made considerably more errors to the novel distractor than did healthy volunteers.
Important insights into the current deficit can be gleaned from an electrophysiological recording study of monkey striatal neurons during performance of DMTS (Johnstone and Rolls, 1990). The authors found that neurons in the ventrocaudal striatum did not exhibit firing related either to the encoding of the sample stimulus to be remembered or to the delay-period. Rather, striatal neurons fired only in relation to test stimuli. The firing of the striatal neurons was related to `response decision' functions, such as initiating a response to the test stimulus when a match had been detected or withholding a response when a non-match was detected. Thus, the role of the striatum in DMTS may include processes of response selection. Such a deficit in Huntington's disease patients would lead to delay-independent deficits, as reported here.
Further insights into the nature of the DMTS deficit in Huntington's disease patients comes from neural network simulation studies. Levine and colleagues (Levine and Prueitt, 1991Levine and Prueitt, 1992; Levine and Kant, 1998) have described a network that can perform the DMTS task, which consists of a reinforcement (reward) mechanism combined with an attentional priming mechanism that influences the response selection process. This helps the network to selectively pay attention to those features (e.g. shape and colour) that are strongly correlated with reward. They have shown that a simulated lesion weakens the influence of reward or punishment on behaviour and leads to behaviour that is no longer guided by those features relevant to successful performance but rather by hard-wired tendencies, such as a strong attraction to novelty. We have applied this framework previously to our findings of perseveration in visual discrimination learning tasks in Huntington's disease, in which a weakening of the links between reinforcement, attentional and response selection mechanisms leads to the enhanced persistence of acquired habits (Lawrence et al., 1999).
An alternative explanation for the tendency of Huntington's disease patients to pick the distractor stimulus is that they are unable to learn the matching principle itself. However, this seems unlikely as Huntington's disease patients were unimpaired in terms of choice accuracy on the VSMTS task, which involves learning and maintaining the same match principle. Huntington's disease patients did, however, evidence some problems in this task, as their reaction times as a function of set size showed a greater slope compared with control subjects. We interpret this as a problem in attentional and/or selection mechanisms, for as the number of potential targets in the array increases, greater demands are placed on mechanisms of selective attention and response selection.
Working memory
In line with our earlier study (Lawrence et al., 1996), Huntington's disease patients were impaired on the self-ordered spatial working memory task. Huntington's disease patients showed an increase in the number of between-search errors, but were only marginally impaired in their use of an effective search strategy and did not show an increase in within-search errors. Thus, the deficit in Huntington's disease patients seems to share features of both the deficit seen in patients with damage to the temporal lobe (possibly including the hippocampus) and that seen in patients with frontal lobe lesions (Owen et al., 1996c).
By contrast, Huntington's disease patients were unimpaired on the visual object working memory task, which is sensitive to temporal lobe lesions but not frontal lobe lesions (Owen, et al., 1996c). This is in contrast to the findings of Rich and colleagues, who found impaired performance in Huntington's disease patients on the visual variant of the Self-Ordered Pointing Task (Rich et al., 1996). However, the visual stimuli used by Rich and colleagues were more complex visually and more numerous than those used in the present study, thus placing greater demands on perceptual and mnemonic processes, respectively.
The present result suggests that Huntington's disease patients do not have a general deficit in visual memory per se. One likely reason for the lack of deficit in the object working memory task is that, unlike the self-ordered spatial working memory task, the implementation and sequencing of a fixed search strategy is not required in the visual object working memory task, which instead relies more heavily on purely mnemonic functions. Although their performance was not reduced to the same extent as that of patients with frontal lobe damage, Huntington's disease patients were less efficient in their use of a repetitive search strategy on the spatial working memory task than controls. Such complex sequencing behaviour has been considered a major function of the basal ganglia (Dominey et al., 1995; Beiser and Houk, 1998; Berns and Sejnowski, 1998).
A related explanation for the lack of impairment in visual object memory but not of spatial working memory in Huntington's disease relates to the notion of `set' (Gibson, 1941). This concept has for a long time been associated with striatal function (Buchwald et al., 1975; Kimura et al., 1984; Robbins and Brown, 1990; Miller and Wickens, 1991). Sets are dispositions or biases and their effect on cognition is one of facilitation or guidance (Gibson, 1941). Buchwald and colleagues define a cognitive set as `the ability to discriminate a situational context and make an appropriate response to a given signal' (Buchwald et al., 1975). The structure of the spatial working memory task, with a fixed spatial array representing a persistent behavioural context, may enable subjects to use a set to help organize their responses. The breakdown of a cognitive set, or failure in the recognition of a persistent behavioural context, in Huntington's disease patients would thus lead to impaired spatial working memory performance. However, in the visual object working memory task, the shapes to be searched through alter their spatial locations after each response. Hence, the build-up of a cognitive set may be less strong in the object working memory test, and so performance in Huntington's disease is spared.
Pattern–location associative learning
It has been argued that a significant reduction in computational complexity is achieved if the recognition of an object can be separated from its localization (Ruekl et al., 1989; Ballard et al., 1997). However, tight integration of `what' and `where' pathways is essential: at any time, it is necessary that the information in one pathway can be correlated unambiguously with that in the other for the purposes of action selection (Goodale and Arbib, 1998; Niebur and Koch, 1998). The pattern–location PAL task used in the present study allowed us to examine the integration of visual and object memory processing. Huntington's disease patients were significantly impaired on the present pattern–location PAL task, in agreement with previous studies of impaired associative learning in Huntington's disease patients (Butters et al., 1983; Sprengelmeyer et al., 1995; Tucker et al., 1996). Impairments on this task have been reported after damage to both the frontal lobes and the temporal lobes in humans (Owen et al., 1995). In addition, a recent functional imaging study (Büchel et al., 1999) has shown that, during the learning of object–location associations, the `effective connectivity' between the dorsal and ventral streams is increased. This linking process, which requires the selection of objects based upon a spatial context, could be mediated at the level of the basal ganglia (Graybiel and Kimura, 1995).
Neuroanatomical considerations
Anatomical data (Cavada and Goldman-Rakic, 1991; Baizer et al., 1993; Yeterian and Pandya, 1995; Gerfen and Wilson, 1996) suggest that input from the dorsal and ventral visual streams is directed both to the prefrontal cortex and the basal ganglia. In addition, the basal ganglia have outputs directed not only to areas of the prefrontal cortex (Alexander et al., 1986; Yeterian and Pandya, 1991; Middleton and Strick, 1994), but also to certain posterior cortical visual areas (Middleton and Strick, 1996).
With respect to corticostriatal connectivity, the medial and dorsolateral extrastriate areas, as well as posterior parietal regions, the medial and dorsolateral prefrontal regions, are related to the dorsal portions of the head and body of the caudate nucleus and to the dorsal putamen (Yeterian and Pandya, 1991, 1995). These pathways have been suggested to play a role in visuospatial cognition. The corticostriatal connections of the ventral extrastriate and inferotemporal regions, as well as those of the ventral prefrontal region, are directed to the ventral portions of the head of the caudate and putamen. This circuitry has been considered to have a role in visual object cognition.
The degree of segregation/overlap of the inputs from the two visual streams at the level of the basal ganglia is uncertain (Gerfen and Wilson, 1996). Anatomical and functional studies suggest a considerable degree of segregation, at least at the level of the input nuclei of the basal ganglia, the striatum (the so-called segregated loops or channels hypothesis) (Alexander et al., 1986; Baizer et al., 1993; Yeterian and Pandya, 1995; Strick and Middleton, 1999).
At this moment, it is unclear whether integration of corticostriatal information processing streams takes place at the level of the striatum, via thalamic relays, or by way of corticocortical interactions (Graybiel and Kimura, 1995; Alexander, 1997; Giménez-Amaya and Scarnati, 1999), although the deficit in pattern–location associative learning in Huntington's disease in the present study is suggestive of some integration of processing at the level of the basal ganglia. Although parietal and temporal projections to the striatum appear to be largely segregated (Baizer et al., 1993), there is a degree of overlap of temporal and parietal lobe inputs to the basal ganglia, which would allow at least a limited degree of convergence (Baizer et al., 1993; Yeterian and Pandya, 1995).
The present results suggest that both visuospatial and visual object processing are impaired in Huntington's disease. Neuropathological studies suggest a dorsal-to-ventral, anterior-to-posterior and medial-to-lateral progression of neuronal death in Huntington's disease (Vonsattel and DiFiglia, 1998), and so one might expect visuospatial function to be impaired earlier in the course of disease progression than visual object processing, given the above anatomical findings from primates. Indeed, previous studies from our laboratory have suggested that functions associated with the dorsal prefrontal cortex are impaired before those associated with the ventral prefrontal cortex (Lawrence et al., 1996, 1999; Watkins et al., 2000). The finding that spatial, as opposed to visual object, working memory is impaired in Huntington's disease might also be taken to support such a position. However, both pattern recognition memory and DMTS appear to be impaired even in early Huntington's disease, and these tests clearly depend on the intact functioning of the ventral processing stream (Owen et al., 1995). Additional studies must address the detailed time course of cognitive dysfunction in Huntington's disease in relation to the proposed progression of striatal pathology in the disorder.
Implications for theories of corticostriatal information processing
The present results have important implications for the nature of corticostriatal information processing.
Current conceptual and computational models of the basal ganglia suggest a prominent role for these structures in context-dependent selection or `set' processes (Robbins and Brown, 1990; Hikosaka et al., 1993; Rolls, 1994; Dominey et al., 1995; Houk and Wise, 1995; Wickens and Kötter, 1995; Mink, 1996; Wise et al., 1996; Boussaoud and Kermadi, 1997; Lawrence et al., 1998; Redgrave et al., 1999). It has been argued (Mitchell et al., 1991; Houk and Wise, 1995; Houk, 1997) that the functions of the corticobasal ganglia loops stem from the capacity of striatal spiny neurons for pattern recognition. There is convergence of input from several cortical columns onto a cluster of striatal spiny neurons. Via dopaminergic reinforcement signals (Schultz, 1997), this neuronal architecture learns to recognize and register complex contextual patterns that are relevant to behaviour (Wickens and Kötter, 1995). Furthermore, since the temporal cortex, motor cortex and prefrontal cortex all receive input from basal ganglia output structures, the basal ganglia are well positioned to regulate the dynamic activity in these cortical areas (Pribram, 1977; Braitenberg, 1978; Wickens and Kötter, 1995). Context-relevant information processing allows the striatum to `instruct' cortical areas as to which sensory inputs or patterns of motor output are behaviourally significant (Beiser et al., 1997) in a given (task) context and should therefore be attended to. Hence, the basal ganglia may be performing a form of pattern classification computation, modifying coarsely coded cortical representations of memories, sensory features or motor intentions into representations which are context-appropriate (Beiser et al., 1997; Lawrence et al., 1998; Amit, 1999).
The current deficits in Huntington's disease, especially on the DMTS task, pattern–location PAL and spatial working memory tasks, which require subjects to inhibit or enable actions based on a recognized environmental context, are compatible with such context registration hypotheses, as is the sparing of performance on the visual object working memory task, in which there is a lack of persistent behavioural context to guide the selection of the appropriate response strategy during the task. Thus, the presence or absence of task deficits in Huntington's disease relates to a deficit in context recognition/registration, rather than a deficit in visuospatial/visual object processing per se.
The cognitive deficit in Huntington's disease is thus consistent with earlier `set' models and related computational models of basal ganglion function, which suggest that corticostriatal circuits compute the patterns of sensory input and response output which are of behavioural significance within a particular environmental context. Future studies in patients at different stages of the disease, especially in a functional neuroimaging context, are now required in order to refine current theories of basal ganglia function.
Group characteristics
| Group | n | Age | NART | Onset | Durn | ADL | MMSE | QNE | KOLT |
|---|---|---|---|---|---|---|---|---|---|
| Data are mean (standard deviation). HDa = Huntington's disease patients tested on DMTS, visual search MTS, visual working memory and VOSP; HDb = separate group of Huntington's disease patients who were tested on PAL; HDc = Huntington's disease patients tested on pattern and spatial recognition memory and spatial working memory; CS1 = control group for HDa; CS2 = separate group of controls for HDb; CS3 = controls for HDc; NART = National Adult Reading Test estimate of premorbid IQ (Nelson, 1982); Onset = age at onset of Huntington's disease; Durn = disease duration in years; ADL = UHDRS activities of daily living score; MMSE = Mini-Mental State Examination (Folstein et al., 1975); QNE = score on chorea scale of Quantified Neurological Examination (David et al., 1987); KOLT = score on Kendrick Object Learning Test (Kendrick, 1985). | |||||||||
| HDa | 19 | 47.5 (9.8) | 110.6 (8.2) | 42.5 (9.2) | 5.0 (2.7) | 21.5 (3.4) | 28.3 (2.2) | n/a | n/a |
| HDb | 19 | 42.7 (11.3) | 111.2 (8.6) | 38.6 (10.4) | 4.1 (1.9) | 21.2 (3.5) | 28.1 (2.7) | 6.1 (1.4) | 35.9 (2.5) |
| HDc | 21 | 41.8 (9.2) | 111.8 (7.8) | 39.5 (8.3) | 5.1 (1.9) | 21.6 (3.1) | 28.3 (2.5) | n/a | n/a |
| CS1 | 20 | 43.4 (16.7) | 107.0 (8.3) | n/a | n/a | n/a | n/a | n/a | n/a |
| CS2 | 20 | 43.4 (15.1) | 112.6 (6.6) | n/a | n/a | n/a | n/a | n/a | n/a |
| CS2 | 17 | 37.7 (15.3) | 108.5 (7.8) | n/a | n/a | n/a | n/a | n/a | n/a |
| Group | n | Age | NART | Onset | Durn | ADL | MMSE | QNE | KOLT |
|---|---|---|---|---|---|---|---|---|---|
| Data are mean (standard deviation). HDa = Huntington's disease patients tested on DMTS, visual search MTS, visual working memory and VOSP; HDb = separate group of Huntington's disease patients who were tested on PAL; HDc = Huntington's disease patients tested on pattern and spatial recognition memory and spatial working memory; CS1 = control group for HDa; CS2 = separate group of controls for HDb; CS3 = controls for HDc; NART = National Adult Reading Test estimate of premorbid IQ (Nelson, 1982); Onset = age at onset of Huntington's disease; Durn = disease duration in years; ADL = UHDRS activities of daily living score; MMSE = Mini-Mental State Examination (Folstein et al., 1975); QNE = score on chorea scale of Quantified Neurological Examination (David et al., 1987); KOLT = score on Kendrick Object Learning Test (Kendrick, 1985). | |||||||||
| HDa | 19 | 47.5 (9.8) | 110.6 (8.2) | 42.5 (9.2) | 5.0 (2.7) | 21.5 (3.4) | 28.3 (2.2) | n/a | n/a |
| HDb | 19 | 42.7 (11.3) | 111.2 (8.6) | 38.6 (10.4) | 4.1 (1.9) | 21.2 (3.5) | 28.1 (2.7) | 6.1 (1.4) | 35.9 (2.5) |
| HDc | 21 | 41.8 (9.2) | 111.8 (7.8) | 39.5 (8.3) | 5.1 (1.9) | 21.6 (3.1) | 28.3 (2.5) | n/a | n/a |
| CS1 | 20 | 43.4 (16.7) | 107.0 (8.3) | n/a | n/a | n/a | n/a | n/a | n/a |
| CS2 | 20 | 43.4 (15.1) | 112.6 (6.6) | n/a | n/a | n/a | n/a | n/a | n/a |
| CS2 | 17 | 37.7 (15.3) | 108.5 (7.8) | n/a | n/a | n/a | n/a | n/a | n/a |
Screen shots from the computerized cognitive tasks. (Top row) Left, pattern recognition memory; right, delayed matching-to-sample. (Middle row) Left, visual search; right, paired-associates learning. (Bottom row) Left, spatial working memory; right, visual object working memory.
Screen shots from the computerized cognitive tasks. (Top row) Left, pattern recognition memory; right, delayed matching-to-sample. (Middle row) Left, visual search; right, paired-associates learning. (Bottom row) Left, spatial working memory; right, visual object working memory.
Percentage correct performance on pattern and spatial recognition memory tasks. Error bars represent the standard error of the mean. Filled columns = patients with Huntington's disease; open columns = control subjects.
Percentage correct performance on pattern and spatial recognition memory tasks. Error bars represent the standard error of the mean. Filled columns = patients with Huntington's disease; open columns = control subjects.
(A) Proportion correct on simultaneous and delayed matching-to-sample. Error bars represent the standard error of the mean. (B) Response latencies on simultaneous and delayed matching-to-sample. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
(A) Proportion correct on simultaneous and delayed matching-to-sample. Error bars represent the standard error of the mean. (B) Response latencies on simultaneous and delayed matching-to-sample. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
(A) Proportion correct as a function of set size on the matching-to-sample/visual search task. Error bars represent the standard error of the mean. (B) Reaction times as a function of set size on the matching-to-sample/visual search task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
(A) Proportion correct as a function of set size on the matching-to-sample/visual search task. Error bars represent the standard error of the mean. (B) Reaction times as a function of set size on the matching-to-sample/visual search task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
Paired-associates learning task. Error bars represent the standard error of the mean. Filled columns = patients with Huntington's disease; open columns = control subjects.
Paired-associates learning task. Error bars represent the standard error of the mean. Filled columns = patients with Huntington's disease; open columns = control subjects.
(A) Number of between-search errors as a function of number of shapes on the visual object working memory task. Error bars represent the standard error of the mean. (B) Number of within-search errors as a function of number of shapes on the visual working memory task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
(A) Number of between-search errors as a function of number of shapes on the visual object working memory task. Error bars represent the standard error of the mean. (B) Number of within-search errors as a function of number of shapes on the visual working memory task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
Number of between-search errors as a function of number of boxes on the spatial working memory task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
Number of between-search errors as a function of number of boxes on the spatial working memory task. Error bars represent the standard error of the mean. Filled circles = patients with Huntington's disease; open circles = control subjects.
We wish to thank all the subjects who took part in this research. We would also like to thank Professor E. H. Yeterian for helpful discussions on corticostriatal anatomy, the two anonymous referees for their comments on the manuscript, and Drs J. Harrison and A. M. Owen for providing task images. This work was supported by a Wellcome Trust Programme Grant to T.W.R., B.J.S., Professor B. J. Everitt and Dr A. C. Roberts. A.D.L. was supported by the MRC and L.H.A.W. by The Huntington's Disease Association and the British Brain and Spine Foundation.
