OUP user menu

Subthalamic nucleus stimulation affects striato‐anterior cingulate cortex circuit in a response conflict task: a PET study

U. Schroeder, A. Kuehler, B. Haslinger, P. Erhard, W. Fogel, V. M. Tronnier, K. W. Lange, H. Boecker, A. O. Ceballos‐Baumann
DOI: http://dx.doi.org/10.1093/brain/awf199 1995-2004 First published online: 1 September 2002


The subthalamic nucleus (STN) has generally been considered as a relay station within frontal‐subcortical motor control circuitry. Little is known about the influence of the STN on cognitive networks. Clinical observations and studies in animals suggest that the STN participates in non‐motor functions which can now be probed in Parkinson’s disease patients with deep brain stimulation of the STN, allowing selective and reversible modulation of this nucleus. Using PET, we studied changes in regional cerebral blood flow (rCBF) associated with a response conflict task (Stroop task) in Parkinson’s disease patients ON and OFF bilateral STN stimulation. The Stroop task requires subjects to name the font colour of colour words (e.g. ‘blue’) printed in an incongruent colour ink (e.g. yellow). During STN stimulation, impaired task performance (prolonged reaction times) was associated with decreased activation in both right anterior cingulate cortex (ACC) and right ventral striatum. Concomitant increased activation in left angular gyrus indicative of ongoing word processing during stimulation is consistent with an impairment to inhibit habitual responses. ACC and ventral striatum are part of the ACC circuit associated with response conflict tasks. The decreased activation during STN stimulation in the ACC circuit, while response conflict processing worsened, provides direct evidence of STN modulating non‐motor basal ganglia‐thalamocortical circuitry. Impairment in ACC circuit function could account for the subtle negative effects on cognition induced by STN stimulation.

  • Keywords: anterior cingulate cortex; deep brain stimulation; PET; Parkinson’s disease; subthalamic nucleus
  • Abbreviations: ACC = anterior cingulate cortex; DBS = deep brain stimulation; NT = neutral control task; rCBF = regional cerebral blood flow; SMA = supplementary motor area; SPM = statistical parametric mapping; ST = stroop task; STN = subthalamic nucleus

Received November 2001. Revised March 11, 2002. Accepted March 22, 2002


The subthalamic nucleus (STN) has been generally associated with motor function. Its role in cognition has not yet been investigated in detail. Lesions at the level of the STN lead to hemiballism and involuntary movements of the contralateral extremities (Martin, 1927; Guridi and Obeso, 2001). The function of the STN in the current model of basal ganglia organization was developed from animal models and neurodegenerative diseases with hypo‐ and hyperkinetic movement disorders (Penney and Young, 1983; Alexander et al., 1986). The model came to be regarded as one of the principal networks of the brain, which can explain many aspects of motor activity, eye movements and behaviour (Alexander et al., 1986; Mega and Cummings, 1994). In this model, the STN represents a relay in the so‐called ‘indirect’ pathway of the parallel basal ganglia‐thalamocortical circuits. The ‘indirect’ pathway links the striatum to the basal ganglia output through the external globus pallidus (GPe) and the STN. STN excitatory efferents reinforce the inhibition of basal ganglia output nuclei, the internal globus pallidus (GPi) and the reticular substantia nigra (SNr) on thalamo‐frontal neurones. A ‘direct’ striato‐pallidal pathway opposes the STN by normally inhibiting basal ganglia outflow, facilitating thalamo‐frontal drive.

In animal models of Parkinson’s disease it has been shown that excitatory neurones from STN to the GPi are overactive, and that lesioning the STN results in a marked improvement in motor function (Bergman et al., 1990; Wichmann et al., 1994). In rats with dopaminergic lesions, the positive effect of STN lesions on motor impairments was accompanied by non‐motor deficits like premature responding in a simple reaction time task (Baunez et al., 1995).

High frequency deep brain stimulation (DBS) of the STN improves motor function in Parkinson’s disease patients and has therefore become a strategy to treat medically intractable Parkinson’s disease patients (Bergman et al., 1990; Limousin et al., 1995; Deep‐Brain Stimulation for Parkinson’s Disease Study Group, 2001). Using functional imaging, it has previously been shown that STN stimulation enhances movement‐associated activation of supplementary motor area (SMA) (Limousin et al., 1997; Ceballos‐Baumann et al., 1999). Pallidotomy and STN stimulation have been shown to enhance movement‐associated activation in premotor cortex in two studies (Grafton et al., 1995; Ceballos‐Baumann et al., 1999), whereas this was not the case in two other studies (Limousin et al., 1997; Samuel et al., 1997b). Activation studies of Parkinson’s disease have found underactivation of the SMA during joystick movements (Playford et al., 1992; Haslinger et al., 2001), self‐initiated simple finger movements (Jahanshahi et al., 1995) and sequential finger movements (Rascol et al., 1992; Samuel et al., 1997a). Overactivation of the lateral premotor cortex has been shown in an event‐related fMRI study investigating joystick movements (Haslinger et al., 2001), as well as in block design fMRI (Sabatini et al., 2000) and PET (Samuel et al., 1997a; Catalan et al., 1999) studies on complex finger movements.

However, facilitation of movement‐related frontal activity is not necessarily coupled with improvement in frontal executive function as suggested by neuropsychological studies in Parkinson’s disease with DBS. Bilateral STN‐stimulation for Parkinson’s disease can cause subtle impairments in various aspects of frontal executive functioning, particularly in older patients (Saint‐Cyr et al., 2000). Two studies examined the impact of DBS on cognitive functions compared performances with either ON‐ versus OFF‐stimulation (Jahanshahi et al., 2000) or pre‐ versus postoperatively (Dujardin et al., 2001), and both reported impaired performance in the Stroop task (ST). This task requires subjects to name the font colour of colour words (e.g. ‘blue’) that are printed in an incongruent colour ink (e.g. red). The ST induces a conflict between the tendency to read the colour word and the actual task of naming the colour of the font, which is reflected in prolonged reaction times compared with a control task, a phenomenon well known as the Stroop effect (MacLeod, 1991). This task probes non‐automated and flexible behaviour, which was predicted to be impaired after stereotactic surgery for Parkinson’s disease (Marsden and Obeso, 1994). In the study by Jahanshahi and colleagues (Jahanshahi et al., 2000), patients made a greater number of self‐corrected errors on the interference condition of the Stroop task during stimulation ON compared with OFF. Dujardin and colleagues (Dujardin et al., 2001) showed that reactions times worsened significantly in the interference condition of the Stroop task when tested 1 year after implantation of the electrodes in the STN compared with the preoperative status.

The aim of the present study was to provide direct evidence of STN participation in a cognitive network. The effects of STN stimulation on the anterior cingulate cortex (ACC) circuit were studied using PET measurements of regional cerebral blood flow (rCBF), as the Stroop task has been studied extensively with functional imaging and has consistently been shown to be associated with activation of the ACC (Pardo et al., 1990; Carter et al., 1995; Bush et al., 1998). We hypothesized that STN stimulation would impair the performance in the Stroop task and reduce task‐specific ACC activation.



The study included 10 patients (five women and five men) with idiopathic Parkinson’s disease (UK Parkinson’s Disease Society Brain Bank clinical diagnostic criteria) (Hughes et al., 1992) and bilateral DBS of the STN. Seven patients [mean age ± standard deviation (SD) = 59.14 ± 12.98, range = 39–78 years; average age of onset = 34.86 ± 11.55 years, range = 27–58 years] could be fully analysed. The data of three patients were excluded due to excessive movement (two patients) or inability to perform the task during scanning (one patient) (for individual scores see Table 1).

View this table:
Table 1

Patient characteristics

Patient codeSexAge (years)Age of onset (years)Medication (mg/day)Handedness*Depression
01Male5939350 l‐dopa; 3 cabergoline Right4
02Female6453900 l‐dopa; 7.5 pergolide; 1000 entacapone;     200 amantadineRight7
03Female5739450 l‐dopa; 6 ropiniroleRight7
04Male3927650 l‐dopa; 4 cabergolineRight4
05Male6955475 l‐dopa; 4 pergolideRight4
06Female7858800 l‐dopaLeft8

*Handedness was measured with the Edinburgh Handedness Inventory. Depression was measured with the Beck Depression Inventory (BDI).

Six patients were taking levodopa, the mean dosage being 518 mg/day (range = 0–800 mg/day); entacapone was used as supplement in one patient. In addition, five out of these six patients were treated with dopamine agonists (pergolide, cabergoline and ropinirole) and one patient with amantadine. One patient did not take any medication (for individual medication see Table 1).

All but one subject were right‐handed following the criteria of the Edinburgh Handedness Inventory (Oldfield, 1971). According to self reports on the Beck Depression Inventory (Beck et al., 1961) none of the patients was depressed (mean = 5.57 ± 1.72) (for individual data see Table 1).

Informed written consent according to the Declaration of Helsinki was obtained from each subject. The local Ethics Committee of the Technische Universität München gave approval for this study. Permission to administer radioactive substances was obtained from the radiation protection authorities (Bundesamt für Strahlenschutz and the Bayerisches Amt für Umweltschutz).


Electrode implantation was guided stereotactically in accordance with preoperative MRI scans and intraoperative stimulation. All patients received quadripolar electrodes (3387; Medtronic, Inc., Minneapolis, USA) with four platinum–iridium cylindrical surfaces (1.27 mm diameter, 1.52 mm length, 3‐mm centre‐to‐centre separation). Contact 0 was the deepest and most distal contact. A brain MRI performed a few days after the procedure verified the positions of the active contact in the STN (Krause et al., 2001).

The stimulator type was Kinetra® in three patients, Itrel®‐II in two patients and Itrel®‐III in the remaining two patients (Medtronic, Inc.). In all patients stimulation was monopolar, using one contact of the quadripolar electrode. The stimulation characteristics were as follows: mean pulse width 60 µs for all patients for the right side and 64.29 µs for the left side (SD = 11.34), mean frequency 152.86 Hz (SD = 28.56) for the right as well as the left side, and mean voltage 2.51 V (SD = 0.80) for the right side and 2.90 V (SD = 0.83) for the left side (for individual stimulator parameters see Table 2).

View this table:
Table 2

Stimulator parameters

NumberStimulatorElectrodesFrequency (Hz)Amplitude (V)Impulse duration (µs)

Electrodes = stimulated (–) and non‐stimulated (0) contacts of the quadripolar electrode from distal (right) to ventral (left).

Neuropsychological and motor assessment

Neuropsychological and motor assessments were performed at the PET centre, independently from the neurosurgical team, at least 5 months after implantation of the electrodes (mean ± SD = 1.07 ± 0.35 years, range = 0.5–1.5 years) and the day before PET scanning with medication and stimulator ON. Global intellectual function was measured with a German short version of the Wechsler adult intelligence scale (Dahl, 1986), consisting of the subtests ‘Information’, ‘Similarities’, ‘Picture Completion’ and ‘Block Design’. The participants were shown to have an average IQ of 109.43 (SD = 16.53; individual scores on neuropsychological tests are shown in Table 3). One patient scored below published norms, however this patient understood the instructions of the Stroop task. Inspection of individual errors (see Table 4 below) confirmed that the patient could perform the task.

View this table:
Table 3

Neuropsychological performance

Patient codeWIPVCSTVOSPColour
ScreeningIncomplete lettersPos. DisDot counting Number location Cube analysis

aImpaired performances according to published norms. 0 = no impairment; WIP = short version of the Wechsler intelligence scale; VCST = visual contrast test system; VOSP = Visual Object and Space Perception; Colour = colour perception measured with Ishihara’s templates; Pos. Dis = position discriminator.

View this table:
Table 4

Averaged and rounded individual errors for false responses, self‐corrected errors and omissions for the Stroop test

Patient numberFalse responsesSelf‐corrected errorsOmissions

ON = activated STN stimulation; OFF = inactivated STN stimulation.

Visual contrast sensitivity and vision acuity were measured with the Visual Contrast Test System (VCTS 6000; Vistech Consultants Incorporation, Dayton, OH, USA) and were normal in all subjects.

Higher visual functions were measured with the Visual Object and Space Perception (VOSP) test, including the subtests ‘Screening’, ‘Incomplete Letters’, ‘Dot Counting’, ‘Position Discrimination’, ‘Number Location’ and ‘Cube Analysis’ (Warrington and James, 1991). For these subtests mean performance fell within the published norms.

Colour perception was assessed with the Ishihara’s tests for colour blindness (Ishihara, 1983), with no patient showing any deficit.

In summary, no decline in contrast sensitivity, visual acuity, perception of objects and space or colour perception was observed in the group.

Motor assessment and PET scanning were performed after overnight withdrawal of medication for 12–13 h. Motor symptoms were assessed with STN stimulation ON or OFF using the motor score of the Unified Parkinson’s Disease Rating Scale (UPDRS) (Fahn et al., 1987) and the Hoehn & Yahr rating (Hoehn and Yahr, 1967). Motor symptoms for ON STN‐stimulation were compared with the OFF‐state using the non‐parametric Wilcoxon tests, with one‐tailed = 0.05.

PET activation paradigm

We used two tasks as activating conditions: a single‐trial Stroop task, consisting of colour words [blau (blue), grün (green), rot (red), gelb (yellow)] and a single‐trial neutral control task (NT) with German animal names (hund, maus, bär, gans), controlling for letter and word processing. Animal words and colour words were matched for the number of letters and syllables. The differences in word frequency were acceptable using standard German Corpora (Mannheimer Corpora 1+2; http://corpora.ids‐mannheim.de/∼cosmas/) to identify word frequencies, including 2.54 million word forms. We determined frequencies for German colour words (grün: 68 = 0.0027%; blau: 66 = 0.0026%; rot: 142 = 0.0056%; and gelb: 50 = 0.0020%) and animal words (Bär: 10 = 0.0004%; Maus: 21 = 0.0008%; Hund: 119 = 0.0047%; and Gans: 9 = 0.0004%). In the ST, the colour of the font never matched the colour name. No font colour and no colour name were ever repeated consecutively. Each word was printed in blue, green, red and yellow capital letters. The words were presented individually against a black background on an LCD monitor with a distance of 40 cm to the patient. Each stimulus was presented with the Experimental Run Time System (ERTS; BeriSoft Cooperation, Frankfurt, Germany) for 1600 ms, followed by a black screen of 500 ms. Reaction times were registered by a voice key and were based on correct as well as on false responses. They were included in the analysis if they occurred between 100 and 2100 ms after word presentation. Additionally, different types of errors were monitored: (i) false responses, if patients read the colour word instead of naming the colour of the font; and (ii) self‐corrected errors, if they corrected false responses and omissions. To compare the effect of stimulation on reaction times and error rates, a non‐parametric Wilcoxon test with one‐tailed = 0.05 was applied for statistical analysis.

Before data acquisition, patients performed a practice block consisting of 10 trials inside the scanner. Participants were instructed to name the font colour of the words presented as quickly and as accurately as possible. Afterwards patients underwent up to 12 consecutive three‐dimensional (3D) H215O PET measurements, with the stimulator either ON effective stimulation or OFF stimulation. The stimulator was turned OFF shortly before scanning and was turned ON again immediately after scanning in order to keep the unpleasant OFF‐state as short as possible.

A 2 × 2 factorial within subject design was used. The factors were ‘task’ (ST versus NT) and ‘stimulation’ (ON versus OFF), resulting in four ‘conditions’ (ST OFF, ST ON, NT OFF, NT ON) that were repeated three times each. The ‘tasks’ were pseudorandomly assigned to the 12 scans, without repetition of one task more than three times. The initial ‘stimulation’ setting was counterbalanced across subjects, and afterwards ‘stimulation’ settings were alternated between scans.

PET data acquisition

PET measurements were performed using a Siemens ECAT HR+ Scanner (Siemens CTI, Knoxville, TN, USA) in 3D mode with a total axial field of view of 15.5 cm without interplane septa. A 10‐min transmission scan using rotating rods of 68Ge/68Ga was acquired for attenuation correction with the septa extended. Two‐dimensional blank and transmission scans were used to reconstruct a 3D attenuation map. Oblique lines of coincidence for which the attenuation correction factor had not been measured were obtained by forward projection through the 3D map (Townsend et al., 1991). To measure rCBF, 10 mCi H215O (5 mSi) were administered intravenously over 30 s with a semibolus injection using an infusion pump. Single frames were acquired for 60 s, starting with the appearance of the tracer in the brain. The behavioural task was started parallel to the injection pump. The interval between successive H215O administrations was 10 min. Following corrections for randoms, dead time and scatter, images were reconstructed by filtered back projection with a Hanning filter (cut‐off frequency 0.4 cycles per projection element), resulting in 63 slices with a 128 × 128 pixel matrix (pixel size 2.0 mm) and interplane separation of 2.5 mm.

PET image transformation and statistical analysis

Statistical analysis was performed using statistical parametric mapping (SPM) software (SPM99; Wellcome Department of Cognitive Neurology, London, UK) based on the general linear model (Friston, 1997). First, scans of each subject were aligned to the first scan of the session to account for head motion in time. This generated an aligned set of images and a mean image (each of 31 planes) for each patient. In a second step, to allow for group analysis, the realigned images were stereotactically normalized into a standard space approximating that of Talairach and Tournoux (Talairach and Tournoux, 1988). Finally, images were smoothed with an isotropic Gaussian kernel of 14 mm to increase the signal to noise ratio and to compensate for interindividual anatomical variability. To remove the effect of variations in global flow across subjects and scans, mean rCBF was proportionally scaled to an arbitrary level of 50 ml/100 ml/min.

For analysing the effect of stimulation on interference processing, a ‘conditions × subject interaction × covariates’ PET model was used. For each subject the design matrix contains four columns, representing the scans for each condition, resulting in a total of 28 columns for all seven subjects. All comparisons were specified by appropriately weighted categorical contrasts [(ST/OFF–NT/OFF)–(ST/ON–NT/ON)] and a SPM{T} map was computed for each contrast on a voxel‐by‐voxel basis. Based on the results of prior PET studies and our a priori hypothesis, significance was accepted for voxels surviving an uncorrected P‐value of <0.001 in both analyses.

All coordinates reported are based on the Talairach atlas and were transformed applying procedures developed by Matthew Brett (http://www.mrc‐cbu.cam.ac.uk/Imaging).


Significant improvement in motor function associated with ON‐ compared with OFF‐stimulation was reflected in lower motor scores of the UPDRS (ON: mean ± SD = 26.57 ± 11.49; OFF: mean = 45.86 ± 16.95; P < 0.01) and the Hoehn & Yahr rating (ON: mean ± SD = 2.79 ± 0.64; OFF: mean = 3.21 ± 0.81; P < 0.05).

The Stroop effect was effectively induced with and without STN stimulation. For STN‐stimulation ON, the mean difference between reaction times for ST (mean ± SD = 1113.12 ± 132.44 ms) and NT (mean ± SD = 885.39 ± 144.95 ms) was 227.73 ms (SD = 73.05; P < 0.05). For STN‐stimulation OFF, the average difference between reaction times for ST (mean ± SD = 1056.47 ± 119.63 ms) and NT (mean ± SD = 913.16 ± 116.19 ms) was 143.31 ms (SD = 55.84; P < 0.05). There was a significant difference between the Stroop effect during ON‐ and OFF‐stimulation (227.73 ms and 143.31 ms, respectively; P < 0.05) (see Fig. 1; for individual data see Table 5).

Fig. 1 Reaction times (ms) during the Stroop and the neutral task, with the stimulator setting ON or OFF.

View this table:
Table 5

Averaged reaction times during Stroop and control task (Patient 1 not recorded)

Patient numberStroop taskNeutral task

ON = activated STN stimulation; OFF = inactivated STN stimulation.

Error data did not differ between ON and OFF stimulation, nor for false responses (ON: mean ± SD = 2.4 ± 3.4; OFF: mean ± SD = 1.9 ± 2.47), self‐corrected errors (ON: mean ± SD = 0.9 ± 1.1; OFF: mean ± = 1.1 ± 1.3) or omissions (ON: mean ± SD = 0.8 ± 1.18; OFF: mean ± SD = 1.0 ± 1.2). Individual errors are shown in Table 4.

When OFF‐stimulation was compared with the ON condition, significant increases in activation (P < 0.001, uncorrected) associated with the Stroop effect (ST–NT) were observed in right ACC (Fig. 2) [Brodmann area (BA) 24/BA 32; x = –10, y = 28, z = 15; Z‐value = 3.70; 12 voxels] and right ventral putamen (x = –20, y = 9; z = –11; Z‐value = 3.51; 80 voxels; see Table 6).

Fig. 2 Increased activation during the stimulator OFF compared with the stimulator ON associated with the Stroop effect (ST–NT; P < 0.001, uncorrected). x, y, z express the position of the voxel in mm relative to the anterior comissure (AC). x = lateral distance from the midline (– right, + left); y = anteroposterior distance from the AC (+ anterior, – posterior); z = height relative to the AC line (+ above, – below). Activation differences were localized in the left anterior cingulate cortex (BA 24/32; x = –10, y = 28, z = 15) and the right ventral striatum (x = –20, y = 9, z = –11).

Fig. 3 Increased activation during the stimulator ON and stimulator OFF associated with the Stroop effect (ST–NT; P < 0.001, uncorrected). x, y, z express the position of the voxel in mm relative to the anterior commissure (AC). x = lateral distance from the midline (– right, + left); y = anteroposterior distance from the AC (+ anterior, – posterior); z = height relative to the AC line (+ above, – below). Activation differences were localized in the left angular gyrus (BA 39, x = 48, y = –70, z = 35).

View this table:
Table 6

Areas with stroop‐associated activation (ST–NT) for OFF>ON and ON>OFF

AreasTalairach coordinates (x, y, z)Z‐scoreVoxel per clusterP (uncorrected)
OFF>ONRight ACC BA 24/ 32–10, 28, 153.7012<0.001
Right ventral striatum–20, 9, –113.5180<0.001
ON>OFFLeft angular gyrus BA 3948, –70, 353.6425<0.001

x, y, z express the position of the voxel in millimetres relative to the anterior commissure (AC). x = lateral distance from the midline (– right, + left); y = anteroposterior distance from the AC (+ anterior, – posterior); z = height relative to the AC line (+ above, – below).

When the ON condition was compared with the OFF state, significant increased ST–NT associated with the Stroop effect (P < 0.001, uncorrected) was observed within the left angular gyrus (BA 39; see Table 4).


As predicted, DBS in the STN decreased rCBF responses in the ACC during the Stroop task, while impairing task performance at the same time. The peak of the differences in activation fell well within the part of the ACC associated with cognitive function (Bush et al., 2000). Decreased activation was also located in the ventral striatum, the structure receiving input from the ACC in the current model of basal ganglia organization. The angular gyrus showed higher activation during ON stimulation compared with the OFF state. The angular gyrus is associated with encoding of words (Menard et al., 1996; Peterson et al., 1999). The increased activation during stimulation therefore most likely corresponds to difficulties in suppressing the habitual response, i.e. word processing, when patients had actually to name the font colour of the colour word. Alternatively, during OFF stimulation, ACC might effectively modulate language processing in the angular gyrus, resulting in decreased activation compared with the ON state. This is in good correspondence to a factorial analysis of cognitive processes in the Stroop task, revealing suppression of language processing to facilitate colour naming (Peterson et al., 1999).

Five separate parallel basal ganglia‐thalamocortical circuits have been proposed (Alexander et al., 1986) to form one of the principal networks of the brain and to mediate motor activity, eye movements and behaviour (Alexander et al., 1986; Mega and Cummings, 1994). These circuits link the basal ganglia via the ventral and dorsomedial thalamus to premotor areas (‘motor circuit’), frontal area 8 (‘oculomotor circuit’), to the dorsolateral prefrontal cortex (‘DLPFC circuit’), the orbitofrontal cortex (‘OFC circuit’) and the anterior cingulate cortex (‘ACC circuit’). The ACC (BA 24) projects to the ventral striatum and forms, together with the ventral pallidum and the dorsomedial thalamus, the ACC circuit (Alexander et al., 1990). Clinical studies suggest that the ACC circuit is associated with response inhibition (Mega and Cummings, 1994), and functional imaging studies with Stroop and similar response conflict tasks show that the ACC is also involved in conflict monitoring (Carter et al., 1998; Botvinick et al., 1999) and ‘attention to action’ (Posner and Dehaene, 1994). Thus, impaired performance in the ST reported here is most likely linked to disturbances within the ACC circuit.

Stimulation‐induced disruption of cortico‐STN afferents from the ACC could well account for the impaired performance and reduced activation seen in this study. Work in STN‐lesioned rats suggests that the cortico‐STN‐output nuclei pathway helps to erase the effect of the previous responses, thus facilitating performance of the currently selected action (Baunez et al., 2001). Cortical efferents to the STN are fast conducting excitatory pathways that can bypass cortico‐striato‐pallidal‐STN projections (Maurice et al., 1998). Accordingly, STN stimulation may disrupt these critical cortico‐STN pathways, which are irrelevant for simple motor function.

Our results are consistent with the emerging identification of the role of the STN as an important basal ganglia input and output structure not only in the motor, but also in the cognitive/limbic domain. Lateral parts of the STN receive somatotopically organized projections from primary motor cortex (Monakow et al., 1978), whereas the remaining nucleus is occupied by projections from premotor and prefrontal areas (Canteras et al., 1990; Feger et al., 1994). The medial part of the STN is intimately and partly reciprocally connected with limbic parts of the basal ganglia (Groenewegen and Berendse, 1990; Maurice et al., 1998). Direct cortico‐STN projections and indirect connections via the nucleus accumbens and ventral pallidum from prelimbic and medial orbital cortex have an excitatory influence on the STN (Maurice et al., 1998).

According to the classical basal ganglia model, STN stimulation should result in facilitation of projections to cortical areas in the five circuits. Previous activation studies fitted well with this prediction, reporting increased movement‐associated activation of the frontal motor association areas (SMA, lateral premotor cortex and ACC) coupled to clinical improvement in motor function during STN stimulation (Limousin et al., 1997; Ceballos‐Baumann et al., 1999). Here, we find the reverse effect, namely impaired cognitive performance coupled to reduced activation in the ACC circuit. In contrast to previous PET studies of motor activation with joystick movements and sequential finger movements, the Stroop task requires a high level of conflict monitoring and response inhibition. According to these differences in task demands, different parts of the ACC are activated. The increased activation associated with joystick movements during STN stimulation fell within the ‘caudal’ ACC and extended into the SMA (Limousin et al., 1997). This part of the ACC has previously been found to be underactive in Parkinson’s disease during finger movement (Playford et al., 1992; Jahanshahi et al., 1995). In contrast, the Stroop task has been unanimously associated with the rostral or cognitive part of the ACC (Pardo et al., 1990; Carter et al., 1995; Taylor et al., 1997), which was found to be underactive during STN stimulation in our study.

The differential impact of STN stimulation on cognitive and motor circuitry supports the two opposite predictions of the so‐called paradox of stereotactic surgery in Parkinson’s disease (Marsden and Obeso, 1994). The motor system devoid of the important basal ganglia input structure STN could continue to function in routine, predictable and automatic movement. ‘Releasing the brake’ on frontal function with STN stimulation improves aspects of motor function (Aziz et al., 1991; Limousin et al., 1995). However, in new and unexpected situations requiring non‐automated behaviour, disruption of basal ganglia input at the level of the STN may lead to inflexibility of mental and motor responses, as shown in the Stroop task.

In summary, here we have provided direct evidence that STN participates in a cognitive circuitry involving ACC and ventral striatum. We have shown that STN stimulation has a disruptive effect on this cognitive circuit while motor function was generally improved. However, the functional impairment in this cognitive network was clinically not apparent and was noted during response conflict processing, which is only one facet of cognitive control. Other executive functions and their underlying neuronal networks may not be influenced by STN stimulation. Future studies should probe neural circuits of other executive functions and also examine their clinical significance.


This study was supported by the Deutsche Parkinson Vereinigung–Bundesverband and the Deutsche Forschungsgemeinschaft (SFB 462, TP C3). We wish to thank our radiochemistry group and cyclotron staff for their reliable supply of H215O, and B. Dzewas, H. Fernolendt and C. Kruschke for their technical assistance.


View Abstract