Embed Size (px)
Citation preview
Exp Brain Res (2007) 179:457–474
DOI 10.1007/s00221-006-0802-2RESEARCH ARTICLE
Visuomotor learning in immersive 3D virtual reality in Parkinson’s disease and in aging
Julie Messier · Sergei Adamovich · David Jack · Wayne Hening · Jacob Sage · Howard Poizner
Received: 11 October 2005 / Accepted: 13 November 2006 / Published online: 5 December 2006© Springer-Verlag 2006
Abstract Successful adaptation to novel sensorimotorcontexts critically depends on eYcient sensory process-ing and integration mechanisms, particularly thoserequired to combine visual and proprioceptive inputs.If the basal ganglia are a critical part of specialized cir-cuits that adapt motor behavior to new sensorimotorcontexts, then patients who are suVering from basalganglia dysfunction, as in Parkinson’s disease shouldshow sensorimotor learning impairments. However,this issue has been under-explored. We tested the abil-ity of 8 patients with Parkinson’s disease (PD), oVmedication, ten healthy elderly subjects and tenhealthy young adults to reach to a remembered 3Dlocation presented in an immersive virtual environ-ment. A multi-phase learning paradigm was used hav-ing four conditions: baseline, initial learning, reversal
learning and aftereVect. In initial learning, the com-puter altered the position of a simulated arm endpointused for movement feedback by shifting its apparentlocation diagonally, requiring thereby both horizontaland vertical compensations. This visual distortionforced subjects to learn new coordinations betweenwhat they saw in the virtual environment and theactual position of their limbs, which they had to derivefrom proprioceptive information (or eVerence copy).In reversal learning, the sign of the distortion wasreversed. Both elderly subjects and PD patientsshowed learning phase-dependent diYculties. First,elderly controls were slower than young subjects whenlearning both dimensions of the initial biaxial discor-dance. However, their performance improved duringreversal learning and as a result elderly and young con-trols showed similar adaptation rates during reversallearning. Second, in striking contrast to healthy elderlysubjects, PD patients were more profoundly impairedduring the reversal phase of learning. PD patients wereable to learn the initial biaxial discordance but were onaverage slower than age-matched controls in adaptingto the horizontal component of the biaxial discordance.More importantly, when the biaxial discordance wasreversed, PD patients were unable to make appropri-ate movement corrections. Therefore, they showedsigniWcantly degraded learning indices relative to age-matched controls for both dimensions of the biaxialdiscordance. Together, these results suggest that theability to adapt to a sudden biaxial visuomotor discor-dance applied in three-dimensional space declines innormal aging and Parkinson disease. Furthermore, thepresence of learning rate diVerences in the PD patientsrelative to age-matched controls supports an impor-tant contribution of basal ganglia-related circuits in
J. Messier (&)Département de kinésiolgie, Université de Montréal, 2100, boul. Édouard-Montpetit, bureau 8225, H3T 1J4 Montreal, QC, Canadae-mail: [email protected]
D. JackCenter for Molecular and Behavioral Neuroscience, Rutgers University, Newark, NJ, USA
H. PoiznerInstitute for Neural Computation, University of California, San Diego, CA, USA
S. AdamovichDepartment of Biomedical Engineering, New Jersey Institute of Technology, Newark, NJ, USA
W. Hening · J. SageDepartment of Neurology, UMDNJ/Robert Wood Johnson Medical School, Piscataway, NJ, USA
123
458 Exp Brain Res (2007) 179:457–474
learning novel visuomotor coordinations, particularlythose in which subjects must learn to adapt to sensori-motor contingencies that were reversed from those justlearned.
Keywords Visuomotor learning · 3D reaching movements · Virtual reality · Parkinson’s Disease · Normal aging
Introduction
Our capacity to learn novel arbitrary associationsbetween vision and action allows for eYcient interac-tions with the world. For instance, the precision withwhich we manipulate a novel tool or drive a new cardepends greatly on our ability to learn and adjustmovements to altered external conditions. A commonapproach to studying how novel sensorimotor associa-tions are formed involves perturbing the relationshipbetween the movement and its sensory consequences,either by altering the relation between the actual andvisually presented hand positions (visual distortions),or by changing the relationship between applied forcesand resulting hand displacements (mechanical distor-tions).
In recent years, such adaptation learning has beenintensively studied using anatomical, imaging and psy-chophysical approaches (Imamizu et al. 2000; Ghilardiet al. 2000; Krebs et al. 1998, 2001; Houk and Wise1995; Grafton et al., 1995; Alexander et al. 1994; Hooverand Strick 1993; Doyon et al. 2003). An increasingbody of evidence suggests a substantial contribution ofthe basal ganglia in sensorimotor learning (Rauch et al.1997; Krebs et al. 1998; Shadmehr and Holcomb 1999;Graybiel 2004). Of particular interest, a number ofrecent studies have provided insights about the speciWcsensorimotor contexts requiring the greatest participa-tion of the basal ganglia (Shadmehr and Holcomb1999; Krebs et al. 1998, 2001; Contreras-Vidal andBuch 2003; Krakauer et al. 2004). For example, Krebset al. (1998) examined how patterns of regional cere-bral blood Xow (rCBF) changed as a function of learn-ing phase when subjects reached to visual targets in aforce Weld. They found a signiWcant contribution of thecorticostriatal loop during early implicit motor learn-ing, whereas late implicit motor learning was associ-ated with a large increase in the corticocerebellarloops. However, when the force Weld was reversed,both the corticostriatal and the corticocerebellar loopsshowed a signiWcant change in rCBF. These observa-tions provide evidence that basal ganglia are moreinvolved at the beginning of the acquisition process
(early exposure) or when a novel situation is imposed(reversal learning). These conclusions were corrobo-rated by Shadmehr and Holcomb (1999) who used asimilar force-Weld task with PET. Likewise, usingfMRI, Seidler et al. (2006) found initial basal gangliaengagement when subjects adapted to visuomotorrotations. However, other studies have failed to showsuch initial basal ganglia activation (Inoue et al. 2000),have found no initial but only late basal ganglia activa-tion (Graydon et al. 2005), or have found initial basalganglia engagement in gain adaptation only (Krakaueret al. 2004).
If the basal ganglia are a critical part of specializedcircuits that adapt motor behavior to new sensorimotorcontexts, then patients who are suVering from basalganglia dysfunction, as in Parkinson’s disease shouldshow sensorimotor learning impairments. Successfuladaptation to novel sensorimotor contexts criticallydepends on eYcient sensory processing and integrationmechanisms, particularly those required to combinevisual and proprioceptive inputs. Studies of adaptationlearning in PD patients have also led to sometimes con-Xicting results, with some studies showing mild or nosensorimotor learning impairments, whereas othersreported marked learning deWcits (Stern et al. 1988;Fucetola and Smith 1997; Contreras-Vidal et al. 2002;Teulings et al. 2002; Contreras-Vidal and Buch 2003;Fernandez-Ruiz et al. 2003; Krebs et al. 2001). More-over, very few studies have examined the speciWc sen-sorimotor learning phases within which PD patientsshow deWcits. To further investigate the speciWc eVectsof Parkinson disease and normal aging on visuomotoradaptation-learning abilities, we compared the point-ing accuracy of young controls, elderly controls and PDpatients in a novel visuomotor multistage learning task.
Finally, almost all previous studies have analyzedmovements restricted to a single plane or have pre-sented visual feedback in a diVerent plane from that inwhich the movements were performed. Such require-ments can alter movement accuracy compared to con-ditions of unconstrained movements with feedbackpresented in the same space as the movements (Mess-ier and Kalaska 1997; Desmurget et al. 1997). In thepresent experiment, we used a three-dimensional vir-tual immersive environment to dissociate visual fromproprioceptive information. The position of a simu-lated arm endpoint used for movement feedback wasaltered by shifting its apparent location diagonally,requiring thereby both horizontal and vertical compen-sations. No study has yet examined the ability of PDpatients to compensate for a visual distortion duringnatural unconstrained movements within three-dimen-sional space. This becomes particularly important since
123
Exp Brain Res (2007) 179:457–474 459
reaching in the vertical dimension requires compensa-tion for gravity, a dimension that requires the mostdependence on processing proprioception. Since thissense is impaired in PD patients (e.g., Maschke et al.2003), it is predicted that PD patients should show diY-culties in a sensorimotor learning task in which novelvisuomotor relationships are imposed. To further testthe nature of adaptation learning in PD patients and toexamine whether it is globally impaired or selectivelydeWcient, we used a four phase learning paradigm. Thisparadigm comprised baseline performance, initiallearning, reversal learning (in which the sign of the dis-tortion in initial learning was reversed), and aftereVect(baseline performance again). The comparison ofpointing errors made during the baseline before expo-sure to the visual distortion to those made in initiallearning will allow the dissociation of any diYculty PDpatients have in utilizing and integrating propriocep-tive information from deWcits in learning the novelvisuomotor relationships. Importantly, the reversalphase of learning required subjects to switch from onenewly learned motor pattern to a diVerent motor pat-tern, since subjects were exposed to opposite biaxialdiscordances in quick succession. If the basal gangliaplay a critical role in facilitating learning of novel sen-sorimotor contexts, then Parkinsonian subjects shouldshow impairments in initial and reversal learning, per-haps with the greatest impairments during the reversallearning phase.
Methods
Subjects
Ten neurologically normal young adults (meanage = 27), eight PD subjects (mean age = 71; range 61–79 years) and ten age-comparable healthy controls
(mean age = 68.5; range 63–78 years) participated inthis study.1 There was no signiWcant diVerence in agebetween elderly healthy subjects and PD patients (t =–1.056, P > 0.05). All subjects were right-handed indi-viduals. The PD subjects were evaluated by a move-ment disorders specialist at the time of testing andwere found to have mild to moderate Parkinson’s dis-ease (Hoehn and Yahr Stages II and III; Hoehn andYahr 1967), and showed motor scores ranging from25.5 to 48 on the United Parkinson’s Disease RatingScale (UPDRS, Fahn and Elton 1987) (see Table 1).Parkinsonian subjects were also evaluated with neuro-psychological tests. PD subjects showing signs ofdementia or depression as revealed by a battery of teststhat included the Mini-Mental Test (cut-oV score <25/30) and Beck Depression Inventory (cut-oV score >10)were excluded from the experiments. PD subjects werestudied the morning before taking their daily anti-par-kinsonian medication, so that they were at least 12 h oVmedication. Subjects were informed about the generalnature of the study and signed an institutionallyapproved consent form. However, no information wasprovided about the speciWc nature of the visual distor-tion applied in the VR world.
Virtual reality environment
Software routines were developed to present subjectswith a VR visual world, to graphically simulate themotions of their arms, and to calibrate the VR worldusing movements sensing devices and an SGI Octane/sse workstation. To view the VR world, the subjectwore a head-mounted display unit (Virtual Technolo-
1 An eleventh elderly subject was run, but his data were excludedsince analysis revealed that his spatial errors fell well outsidethose of all other elderly control subjects; they were greater thantwo standard deviations from the mean
Table 1 Clinical features of patients with Parkinson’s disease
a Motor section. Higher scores indicate greater impairmentsb Duration is years since Wrst remembered parkinsonian symptomc Medication codes: LevR Carbidopa/levodopa sustained release, Lev Carbidopa/levodopa (regular formulation), Pr Pramipexole, SelSelegiline, Ent Entacapone, Br Bromocriptine, Rop Ropinirole
PD patient Sex Most aVected side Stage Age UPDRSa Duration (years)b Medicationsc
SA F Right 2.5 67 25.5 12 Lev, Pr, EntRB M Right 2.5 61 29 14 Lev, LevR, PrSC M Left 2.5 79 36 8 Lev, Pr, SelRF M EqualAV 3 73 30.5 7 Lev, Sel, BrHK M Right 2.5 74 40 14 Lev, Pr, SelBP F Left 2.5 71 30 11 Lev, PrGS M Right 2.5 69 25.5 6 LevAS M Left 3 75 48 10 LevR, Ent, Rop
123
460 Exp Brain Res (2007) 179:457–474
gies, Inc. V–8 headset) that provided immersive stereovisual input. The arm’s position and orientation wasmonitored by signals from “Flock of Birds” electro-magnetic sensors (Ascension Technology, Inc.) posi-tioned on the right shoulder, wrist, and index Wngertip,and on the head-mounted display.
General experimental procedure
Subjects were seated with their backs resting on theback of a straight-back chair. The VR world wasadapted to each participant by a calibration procedure.First, the position of the left eye and the inter-pupillarydistance were computed by appropriate placement ofFlock of Birds markers and a software routine. Thisallowed computation of the viewpoint, which was deW-ned to be the point directly between the two eyes, andallowed creation of a coherent immersive stereo visualinput. Second, the recorded position of the markerplaced on the headset was translated to the position ofthe viewpoint, so that rotations of the head producedrotations of the virtual world centered about the view-point. Finally, the length of the subject’s arm fromshoulder to wrist was measured to normalize acrosssubjects the placement of the target (a green sphere) inthe virtual world at a distance reachable without fullarm extension and close to subject’s midline at eyelevel.
The experiment consisted of a series of conditions.The Wrst was a “familiarization” condition in which par-ticipants executed ten movements in the VR world inthe absence of any perturbations and with continuousvisual feedback in the form of a stick Wnger represent-ing the subject’s index displayed in VR. The start posi-tion of each movement was indicated with a Velcromarker (7 mm £ 7 mm) placed on the right thigh ofparticipant 10 cm from the knee. For a starting trial,subjects were asked to point in front of them at eyelevel in the VR environment. No target was displayed inthe VR world for this Wrst trial. After each movement, a3D reconstruction of the subject’s hand trajectory wasshown simultaneously with the target location (Fig. 1a).Also, the subject’s Wnger movement was replayed. Thatis, subjects could see the displacement of the stick Wngermoving along the trajectory path, thus providingdynamic visual information about the previous trial.
Subjects were then required to point to the spatialposition that would minimize the distance between thetarget and their trajectory endpoint, and to bring backtheir arm to the initial position. Subjects were encour-aged to make straight and uncorrected movements atnatural speed. The “go” signal for the subject to initiatemovements was given verbally by the experimenter to
insure that subjects were ready to execute their move-ments. There was no instruction to initiate responsesrapidly, which might have inXuenced reaction time.
The second condition consisted of “baseline” trialsduring which subjects performed twelve movements inabsence of any perturbations. In contrast to the famil-iarization block, visual feedback was removed duringthe movements, thereby forcing participants to relyonly on the visual feedback (trajectory and target)shown after each trial to maintain and/or improve theiraccuracy level. The third condition consists of the “ini-tial learning”. Subjects performed twenty trials whileexposed to a biaxial visuomotor discordance. The samevisual feedback mode (trajectory and target) was usedas in the baseline condition; however this perturbationshifted the trajectory used for movement feedbackdiagonally, 10 cm to the right and 10 cm higher thanthe veridical feedback provided in the Baseline condi-tion, requiring thereby both horizontal and verticalcompensations. Thus, this visual distortion forced sub-jects to learn new coordinations between what they seein VR and the actual position of their arm, which theymust derive from proprioceptive information since novision of the arm was given during the movement.
The fourth condition was “reversal learning”. Thiscondition required the learning of the reverse biaxialdiscordance; the trajectory endpoint was now shifted10 cm to the left and 10 cm lower than the veridical feed-back provided in the baseline condition. Fifteen trialswere performed during the reversal learning condition.
In the Wfth condition or “aftereVect”, participantsperformed ten movements in the undistorted VRworld as in baseline.
Fig. 1 3D reconstruction of the hand trajectory shown simulta-neously with the target location for feedback after each trial.Subject sees display in 3D and sees dynamic replay of the Wngermotion along the path
123
Exp Brain Res (2007) 179:457–474 461
Kinematic recordings and data analysis
As mentioned previously, the arm’s position and orien-tation was monitored by signals from “Flock of Birds”electromagnetic sensors (Ascension Technology, Inc.)positioned on the right shoulder, elbow, wrist, andWngertip, and on the head-mounted display. The posi-tion of each marker was sampled at 100 Hz. The posi-tion series were then digitally low-pass Wltered using aButterworth Wlter with a cutoV frequency of 8 Hz.Movement onset was deWned as 5% of peak velocity ofthe index Wngertip. After the hand accelerated fromrest, and continued until the velocity returned to nearzero at the end of the outward movement, where thepath reversed direction and began returning towardthe subject’s body. The point of path reversal wasdetermined by a minimum in tangential velocity and/orby visually determining the spatial reversal of the tra-jectory, i.e., the Wrst time the handpath changed direc-tion and returned toward the subject’s body. Thisreversal point was deWned to be movement oVset. Themovement paths that corresponded to the outwardmovements toward the target were selected for furtheranalysis. All trajectories were visualized in 3D andcould be rotated, translated, scaled, and viewportmapped in real-time for interactive analysis (Poizneret al. 1998).
Performance indices and statistics
Since the distortion of the VR environment wasrestricted to the frontal plane, constant and absolutehorizontal and vertical pointing errors were analyzed.Constant horizontal and vertical errors were calculatedas the deviation between the coordinates of the targetand those of index Wnger endpoint in the horizontal(lateral direction ‘x’) and vertical (vertical direction ‘z’)dimensions, respectively. Absolute horizontal and ver-tical errors were absolute values of the constant hori-zontal and vertical errors respectively.
Constant errors (signed errors) can cancel eachother out when averaged across trials and subjects,should some errors be positive and others negative,thereby reducing the magnitude of the average errors.Therefore, absolute errors were used to indicate thesize of errors, whereas the constant horizontal and ver-tical errors indicated the spatial location of the handrelative to the target in all task phases.
The standard deviations of constant horizontal andvertical errors obtained for each subject for the base-line condition and the last Wve trials of initial learning,reversal learning and aftereVect were used as variabil-ity indexes to test for group diVerences in trial-to-trial
variability. This allowed us to test how subject groupsdiVered in their ability to attain a ‘stable’ level of accu-racy by the end of all task phases.2
Baseline
The control subjects and PD patients could easily“touch” the targets with close to zero horizontal andvertical constant error in VR after 5–10 trials of train-ing in the familiarization condition. However, in thebaseline condition, after the removal of visual feed-back of hand displacement during the movement, accu-racy and inter-trial variability increased for the Wrstbaseline trials before the performance stabilized. Forthis reason, only the eight last trials of baseline condi-tion were analyzed in this study.
To test whether control subjects and PD patientsshowed similar levels of accuracy during the baselinecondition, separate two factor analyses of variance(group £ trial) were performed on the constant horizon-tal and vertical errors of each subject over the last eighttrials. Further, to test for group diVerence in trial-to-trialvariability, single factor analyses of variance were per-formed on the standard deviation of horizontal and ver-tical errors made by each subject during baseline trials.For the two analyses mentioned above as well as for allvariance analyses used in this study a signiWcance levelof 0.05 was used. Post-hoc pairwise comparisons acrossgroups were then performed with a Newman–Keuls test.For these post-hoc analyses, we reduced the probabilityof type I error by dividing the original alpha level (0.05)by the number of planned comparisons. Thus, given thatwe compared young controls to elderly controls andolder controls to PD patients, an alpha level of 0.025 wasused for these analyses.
2 The number of trials tested is diVerent for the initial learning (20trials) and the reversal learning phases (15 trials). To comparemore directly performance indices between the two learningphases, we also performed statistical analyses on the learningphases using only the Wrst 15 trials. For instance, we evaluated thevariability indexes and the adaptation magnitude scores on trials10–15 of initial learning as was done for reversal learning. Similarresults were obtained as when using all 20 trials of initial learning.The adaptation score for the horizontal component of the initialbiaxial discordance was similar in all three subject groups(F(2,25) = 0.285; P > 0.05) as was the level of trial-to-trial variabil-ity (F(2,25) = 0.012; P > 0.05). Also, there was a signiWcant betweengroup diVerence in the adaptation magnitude score for the verti-cal component of the initial discordance (F(2,25) = 8.49; P < 0.05).Young and elderly controls showed a similar level of adaptation(P > 0.05), while PD patients showed signiWcantly smaller adapta-tion magnitude score (P < 0.025). In a similar manner, we per-formed the ANOVAs on constant errors for trials 1–15 of initiallearning. Again, these analyses conWrmed results found whentrials 1–20 were analyzed
123
462 Exp Brain Res (2007) 179:457–474
Learning conditions
In order to test if all three subject groups made a simi-lar constant horizontal and vertical error when Wrstexposed to both initial and reversal visuomotor biaxialdiscordances a single factor ANOVA was performedon the Wrst trial of exposure to each visual distortion.
To test for diVerence in learning rates exhibited byeach subject group, separate two factor analyses of var-iance (group £ trial) were performed on the constantand absolute horizontal and vertical errors obtainedduring both initial and reversal learning for each sub-ject of each group.
To further determine how subject groups diVered intheir ability to learn the biaxial visuomotor discor-dances, we computed an adaptation magnitude score.This was calculated as the diVerence between the con-stant horizontal and vertical errors obtained when Wrstexposed to each visual distortion (trial 1) and the aver-age value of the last Wve trials of exposure to eachvisual distortion.2
AftereVect
To further evaluate adaptation to the reversed visuo-motor biaxial discordance, an aftereVect test was per-formed separately on horizontal and vertical constanterrors. Comparisons between the average baselineconstant errors and the constant error made when thediscordance was Wrst removed was tested using t-testsfor independent samples for both the horizontal andvertical constant errors for each subject group. Also, toassess any fatigue eVect, i.e., whether subjects wereable to return to their baseline level of accuracy by theend of aftereVect, we compared the average baselineconstant errors (last 5 trials) with the average after-eVect constant errors (last 5 trials) using t tests for inde-pendent samples, for both the horizontal and verticalconstant errors for each subject group.
Results
PD patients showed learning deWcits along both the vertical and the horizontal dimensions of the biaxial discordance
To investigate how exposure to the sequential biaxialdiscordances inXuenced the accuracy and learning rateof controls and PD patients, separate ANOVAs wereperformed on the horizontal and vertical dimensions ofthe endpoint positions of subjects for all task phases.This allowed the evaluation of the possible diVerence
in the time course and magnitude of adaptation to thevertical and horizontal dimensions of the biaxial dis-cordances.
Accuracy along the horizontal dimension
Figure 2a and b shows constant and absolute reachingerrors made along the horizontal dimension during the12 trials of the baseline (pre-exposure phase), duringinitial learning, learning of the reversed discordanceand aftereVect conditions for the three subject groups.In the baseline condition, PD patients as well as bothgroups of control subjects made very small horizontalconstant errors that were of similar magnitude, withmean constant errors ranging from ¡1.08 cm to1.21 cm. Moreover, all groups showed a stable level ofhorizontal accuracy across baseline trials. There was nosigniWcant diVerence in the magnitude of horizontalconstant errors both among groups (F(2,25) = 0.50;P > 0.05) and between trials (F(7,175) = 0.54; P > 0.05).However, the magnitude of trial-to-trial variability sig-niWcantly diVered between groups (F(2,25) = 7.97;P < 0.05). Post-hoc tests indicated that young and eld-erly controls displayed similar levels of intertrial vari-ability along the horizontal dimension (P > 0.05),whereas intertrial variability was signiWcantly higher inthe PD subjects compared to elderly control subjects(P < 0.025).
When the initial biaxial discordance was suddenlyand unexpectedly introduced, control subjects and PDpatients made a large horizontal constant error of simi-lar size to the imposed visuomotor discordance(Fig. 2a). Although this initial error appeared, on aver-age, slightly larger for PD patients than controls, thisbetween group diVerence did not reach statistical sig-niWcance (F(2,25) = 2.54; P > 0.05).
For young controls, this initial horizontal errordeclined rapidly. During early exposure trials, youngcontrols made large compensatory movements, i.e.,movements of large amplitude that either undershot orovershot the horizontal position of the virtual target.As a result, the mean horizontal constant error acrosssubjects was close to zero after only 1 trial of interac-tion with the visuomotor discordance (Fig. 2a). In con-trast, although elderly controls and PD patients alsoimproved with training, they compensated for asmaller proportion of the distortion in each movementand thus exhibited a slower decrease in the magnitudeof constant horizontal errors (Fig. 2a). However, theaverage horizontal errors made by elderly and youngcontrols showed great overlap from intermediate tolate exposure trials, whereas those of PD patients weresystematically larger than both control groups during
123
Exp Brain Res (2007) 179:457–474 463
all trials of exposure to the initial visuomotor discor-dance. Accordingly, there was a signiWcant main eVectof both group (F(2,25) = 8.58; P < 0.05) and trials(F(19,475) = 13.87; P < 0.05), as well as a signiWcant groupby trial interaction (F(38,475) = 2.64; P < 0.05). Post-hoccomparisons across groups revealed that when con-stant horizontal errors were pooled over all exposuretrials to the initial discordance, there was no signiWcantdiVerence between young and elderly subjects(P > 0.05), but there was a signiWcant diVerencebetween elderly controls and PD patients (P < 0.025).To exclude the possibility that the observed signiWcantdiVerence between elderly controls and PD patients isdue to intersubject diVerences in the Wrst exposure trialto the initial discordance, the analysis was subse-quently performed on constant horizontal errors nor-malized for each subject to the magnitude of thehorizontal error that subject made when Wrst exposedto the initial discordance. For these normalized hori-zontal errors, the diVerence between elderly controlsand PD patients fell below the corrected signiWcancelevel (P = 0.035). However, consistent with the
observed between group diVerences in the initial rateof adaptation to the horizontal component of thevisuomotor biaxial discordance, the group by trialinteraction was still signiWcant (F(38,475) = 2.38; P < 0.05,Fig. 2).
To further evaluate these between group diVer-ences, we analyzed the horizontal constant errors madeduring the Wrst Wve exposure trials to the initial discor-dance. As expected, there was both a main eVect ofgroup (F(2,25) = 11.80; P < 0.05) and a main eVect oftrial (F(4,100) = 13.05; P < 0.05) as well as a signiWcantgroup by trial interaction (F(8,100) = 2.48; P < 0.05).During early exposure trials, young subjects madesigniWcantly smaller horizontal constant errors thanelderly controls (P < 0.025), whereas elderly controlsand PD patients made horizontal errors of similarmagnitude (P > 0.05).
The results were similar when applied on absolutehorizontal errors. There were main eVects of group(F(2,25) = 6.13; P < 0.05) and trial (F(19,475) = 21.07;P < 0.05), and the group by trial interaction was signiW-cant (F(38,475) = 2.19; P < 0.05). As was the case for the
Fig. 2 a Mean constant hori-zontal pointing errors for each of the three groups of subjects for all task phases as a func-tion of trial number. Positive horizontal constant errors correspond to movement end-points to the right of the tar-get, whereas negative horizontal constant errors correspond to movement endpoints to the left of the target. Error bars represent standard errors of the means (SE). b Mean absolute hori-zontal pointing errors for each of the three groups of subjects for all task phases as a func-tion of trial number
-30
-25
-20
-15
-10
-5
0
5
10
15
20
0 5 10 15 20 25 30 35 40 45 50 55 60H
oriz
onta
l con
stan
t err
ors
(cm
)H
oriz
onta
l abs
olut
e er
rors
(cm
)
Trial number
Baseline
Baseline
Aftereffect
Aftereffect
Initial Learning
Initial Learning
Reversal Learning
Reversal Learning
0
5
10
15
20
25
30
0 5 10 15 20 25 30 35 40 45 50 55 60
YoungOldPD
YoungOldPD
A
B
123
464 Exp Brain Res (2007) 179:457–474
normalized constant horizontal errors, the diVerencebetween elderly controls and PD patients did not reachsigniWcance. Further, the comparison of horizontalabsolute errors made during the Wrst Wve trials of expo-sure revealed main eVects of group (F(2,25) = 6.63;P < 0.05) and trial (F(4,100) = 11.42; P < 0.05) and a sig-niWcant group by trial interaction (F(8,100) = 2.19;P < 0.05). However, in contrast to horizontal constanterrors, the diVerence between the horizontal absoluteerrors made by elderly and young controls during earlyexposure to the initial discordance did not reach signiW-cance. These results indicate that while the size of hori-zontal errors (absolute) made by elderly and youngcontrols were similar during early exposure to the ini-tial discordance, young controls showed less of a spa-tial bias in their movement endpoints around thetarget, and as a result, had smaller constant horizontalerrors.
However, the adaptation magnitude score for thehorizontal component of the initial biaxial discordancewas similar in all three groups (F(2,25) = 0.02; P > 0.05).Further, elderly controls and PD patients were able tostabilize their performance by the end of exposure tothe initial discordance as revealed by their similar levelof trial-to-trial variability along the horizontal dimen-sion during late exposure trials (F(2,25) = 2.86;P > 0.05).2
These results conWrmed the patterns seen in Fig. 2aand b that elderly controls and PD patients were ableto adapt their movements to the horizontal componentof the novel visuomotor arrangement. However, theydid so less eYciently than young controls and there-fore, needed more trials to reach asymptotic perfor-mance.
Consistent with both their similar adaptation magni-tude to the horizontal component of the initial biaxialdiscordance and their comparable trial-to-trial variabil-ity level during late exposure trials to the initial discor-dance, all three subject groups showed an averageconstant horizontal error of similar magnitude whenthe visual perturbation was Wrst reversed (F(2,25) = 0.94;P > 0.05, Fig. 2a). Note that the magnitude of the hori-zontal error made for this Wrst exposure trial representsthe additive eVect of the size of the imposed visual dis-tortion and the aftereVect of initial learning.
During reversal learning, both young and elderlycontrols largely compensated for their initial horizontalerror during early exposure trials and thereafterreturned to their baseline level of horizontal accuracy(Fig. 2a, b). In contrast, PD patients reduced their hor-izontal errors more slowly and continued to producelarger horizontal errors than both control groups dur-ing intermediate and late exposure trials to the reverse
discordance. As shown in Fig. 2a and b, these largerhorizontal errors were associated with large betweensubject variability during all reversal learning exposuretrials. In accord with these observations, there was asigniWcant main eVect of both group (F(2,25) = 6.88;P < 0.05) and trials (F(14,350) = 21.38; P < 0.05) as well asa signiWcant group by trial interaction (F(28,350) = 1.90;P < 0.05) on horizontal constant errors. Post-hoc com-parisons across groups conWrmed that there was no sig-niWcant diVerence between young and elderly controls(P > 0.05) whereas the diVerence between PD patientsand elderly controls was marginally signiWcant duringreversal learning (P < 0.027).
Furthermore, examination of trial-to-trial variabilityduring reversal learning revealed a large increase inoscillatory behavior in PD patients. Indeed, PDpatients were unable to stabilize their performance bythe end of exposure to the reversed discordance suchthat the average horizontal variable errors of PDpatients (6.18 cm) was nearly three times greater thanthose of both young (1.19 cm) and elderly controls(2.21 cm) during late reversal learning trials. Accord-ingly, there was a signiWcant diVerence in horizontalvariable errors between groups for late exposure trialsto the reverse discordance (F(2,25) = 5.97; P < 0.05).Young and elderly controls displayed similar level ofintertrial variability (P > 0.05), whereas PD patientsshowed signiWcantly larger intertrial variability thanage-matched controls (P < 0.025).
The oscillatory behavior of some PD subjects was sopronounced that on some trials they made large errorsin both positive and negative directions during all trialsof exposure to the reverse discordance. As a result,horizontal constant errors canceled each other outwhen averaged across subjects and therefore largelyreduced the magnitude of horizontal constant errorsshown in Fig. 2a. Thus, for reversal learning, the signiW-cantly degraded performance of PD patients is bettercaptured by the observation of absolute horizontalerrors (Fig. 2b). Accordingly, the analysis of varianceon absolute horizontal errors revealed a main eVect ofboth group (F(2,25) = 20.60; P < 0.05) and trials(F(14,350) = 41.67; P < 0.05) and the group by trial inter-action was signiWcant (F(28,350) = 2.24; P < 0.05). Fur-ther, post hoc tests conWrmed that PD patients madesigniWcantly greater absolute horizontal errors thanelderly controls (P < 0.025) during the reversal phaseof learning.
Consistent with the results mentioned above, therewas a signiWcant main eVect of group on the adaptationmagnitude score for the reversal visuomotor discor-dance (F(2,25) = 4.26; P < 0.05). Young and elderly con-trols showed similar adaptation magnitude score
123
Exp Brain Res (2007) 179:457–474 465
(P > 0.05). However, likely due to the large amount ofintertrial and intersubject variability in the PD patientsgroup, the diVerence between elderly controls and PDpatients was only marginally signiWcant at the cor-rected alpha level (P = 0.028). This diVerence was,however, highly signiWcant when computed on abso-lute horizontal errors (P < 0.025). Further, PD patientsdisplayed post-exposure shifts (aftereVects) when com-puted in terms of signed horizontal errors. However, instriking contrast to the control subject groups whoshowed degradation in accuracy similar in magnitudeto the horizontal bias (10 cm) when the reversed dis-cordance was Wrst removed, PD patients showed anabsolute horizontal error of similar magnitude to thosemade during late exposure trials to the reversed discor-dance (Fig. 2b).
Accuracy along the vertical dimension
All three subject groups displayed a stable level of con-stant vertical errors during baseline. However, PD
patients showed a pronounced systematic downwardshift of movement endpoints (Fig. 3a). There was a sig-niWcant diVerence between groups (F(2,25) = 8.38;P < 0.05) but no overall diVerence across trials(F(7,175) = 0.48; P > 0.05) in the magnitude of the con-stant vertical errors. Young and elderly controls madeconstant vertical errors of similar magnitude(P > 0.05), whereas those of PD patients were signiW-cantly larger than those of age-matched controls(P < 0.025).
Consequently, young and elderly controls made aninitial large vertical error of similar magnitude whenthe visuomotor discordance was introduced (Fig. 3a).As for the horizontal dimension, young controlsquickly adapted their movements to the vertical com-ponent of the initial biaxial discordance and thus rap-idly reduced their initial vertical error, while elderlycontrols reduced these errors more gradually (Fig. 3a).PD patients, however, made a much smaller initial ver-tical error than control subjects and thereafter slightlydecreased the magnitude of their vertical errors
Fig. 3 a Mean constant verti-cal pointing errors for each of the three groups of subjects for all task phases as a func-tion of trial number. UnWlled triangles represent the mean constant vertical errors after subtracting the baseline downward bias of each PD subject. Positive vertical con-stant errors correspond to movement endpoints above the target, whereas negative vertical constant errors corre-spond to movement endpoints below the target. Error bars represent standard errors of the means (SE). b Mean abso-lute vertical pointing errors for each of the three groups of subjects for all task phases as a function of trial number
-30
-25
-20
-15
-10
-5
0
5
10
15
20
5 10 15 20 25 30 35 40 45 50 55 60
Ver
tical
con
stan
t err
ors
(cm
)V
ertic
al a
bsol
ute
erro
rs (
cm)
Trial number
Baseline
Baseline
Aftereffect
Aftereffect
Initial Learning
Initial Learning
Reversal Learning
Reversal Learning
YoungOldPD
YoungOldPD
0
5
10
15
20
25
30
0 5 10 15 20 25 30 35 40 45 50 55 60
A
B
123
466 Exp Brain Res (2007) 179:457–474
throughout exposure to the initial discordance. As aresult, the three subject groups displayed a similar ver-tical accuracy level during intermediate and late expo-sure trials.
The vertical component of the biaxial discordancemoved the apparent position of the simulated arm end-point used for movement feedback higher than theactual hand position, thereby forcing subjects to pointlower than the target location presented in the virtualenvironment. Since PD patients were already pointinglower during the baseline condition, there was no needfor large compensations along the vertical axis. There-fore, there was no evidence of learning along the verti-cal dimension in PD patients. Indeed, this simpliWedthe learning of the initial biaxial discordance in PDpatients as they were not required to produce largemovement adjustments along the horizontal and verti-cal axes simultaneously.
The above mentioned results were conWrmed by theanalysis of variance. For constant vertical errors in ini-tial learning, there was no main eVect of group(F(2,25) = 2.72; P > 0.05) but a signiWcant eVect of trials(F(19,475) = 8.67; P < 0.05) and a signiWcant group bytrial interaction (F(38,475) = 2.13; P < 0.05). In a similarmanner as for the horizontal dimension, we furtheranalyzed the group by trial interaction by applying theANOVA to the Wrst 5 trials following introduction ofthe visuomotor distortion. There was a main eVect ofboth group (F(2,25) = 6.23; P < 0.05) and trials(F(4,100) = 8.47; P < 0.05) and a signiWcant group by trialinteraction (F(8,100) = 3.59; P < 0.05). Post-hoc testsconWrmed that elderly controls made larger constantvertical errors than young controls during early expo-sure trials (P < 0.025). Additionally, consistent with thesmaller initial vertical errors made by PD patients,their constant vertical errors were signiWcantly smallerthan those of age-matched controls during early expo-sure trials of initial learning (P < 0.025). Similar resultswere found when ANOVA were applied on absolutevertical errors. There was no main eVect of group(F(2,25) = 3.35; P > 0.05) but a signiWcant eVect of trials(F(19,475) = 9.05; P < 0.05) as well as a signiWcant groupby trial interaction (F(38,475) = 2.29; P < 0.05).
Furthermore, in accord with these Wndings, therewas a signiWcant between group diVerence in the adap-tation magnitude score for the vertical component ofthe initial discordance (F(2,25) = 7.39; P < 0.05). Youngand elderly controls showed a similar level of adapta-tion (P > 0.05), while PD patients displayed signiW-cantly smaller adaptation magnitude (P < 0.025).
During reversal learning, both young and elderlycontrols showed an initial rapid reduction in verticalerrors, followed by a more progressive decline
throughout intermediate and late exposure trials(Fig. 3a). Therefore, although their initial vertical errorwas very large (20 cm), both control groups easilymade the required compensations bringing their Wnalvertical errors on average to 1–2.5 cm during late expo-sure trials. In striking contrast, PD patients onlyslightly moved their hand higher along the vertical axisacross learning trials, and therefore, did not completelycompensated for the reverse discordance. As a result,they showed a constant vertical error that was on aver-age nearly four times greater than that of both youngand elderly controls, even at the end of the reversalcondition (Fig. 3a). Analysis of variance on constantvertical errors revealed both a main eVect of group(F(2,25) = 56.50; P < 0.05) and trials (F(14,350) = 17.25;P < 0.05) as well as a signiWcant group by trial interac-tion (F(28,350) = 1.73; P < 0.05). Post-hoc tests showed asigniWcant diVerence between PD patients and age-matched controls (P < 0.025), but no diVerencebetween young and elderly controls (P > 0.05). TheseeVects were also found when absolute vertical errorswere analyzed. There was a main eVect of group(F(2,25) = 51.86; P < 0.05) and trials (F(14,350) = 27.24;P < 0.05) and a signiWcant interaction (F(28,350) = 2.33;P < 0.05).
Furthermore, there was a signiWcant group diVer-ence in the adaptation magnitude score (F(2,25) = 17.42;P < 0.05). Again, young and elderly controls showedsimilar adaptation levels (P > 0.05), while PD patientsdisplayed signiWcantly smaller adaptation magnitudethan age-matched controls (P < 0.025) during reversallearning. The marked impairment in PD patient’sreversal learning was also evidenced by an absence ofan aftereVect. Following removal of the reverse discor-dance, both control groups showed a large deteriora-tion in their pointing accuracy along the verticaldimension. As a result, the constant vertical errormade when the reverse discordance was Wrst removedwas signiWcantly larger than the average constant verti-cal error made during baseline for both young(t = ¡14.16; P < 0.05) and elderly (t = ¡7.80; P < 0.05)controls. By contrast, the vertical constant error madeby PD patients when the reverse discordance was Wrstremoved was similar than the average constant verticalerror made during baseline (t = ¡0.51; P > 0.05). Fur-ther, PD patients maintained this level of accuracyuntil the end of aftereVect. As a result, there was nosigniWcant diVerence between the average constantvertical error made during the last Wve trials of baselineand the average constant vertical error made duringthe last Wve trials of aftereVect (t = ¡1.04, P > 0.05).The observation that PD patients showed a similarlevel of accuracy along the vertical axis at the end of
123
Exp Brain Res (2007) 179:457–474 467
the Wrst (baseline) and the last task phase (aftereVect)suggests that fatigue was not a major contributor of theobserved degradation in performance of PD patientsduring reversal learning.
In contrast to initial learning, the reverse discor-dance shifted the apparent hand location lower thanthe actual hand position, thereby forcing subjects topoint higher than the target location presented in thevirtual environment. Therefore, instead of reducing themagnitude of the downward shift exhibited by PDpatients, as was the case during initial learning, theimplementation of the reverse discordance wouldincrease the magnitude of the downward shift.
To test whether the large baseline vertical bias ofthe PD patients is the major contributor to the muchslower adaptation rates of PD patients observed alongthe vertical dimension, the mean constant verticalerror made by each PD subject during the baseline wassubtracted from the vertical bias made by the subjectfor each trial for reversal learning. Subsequently, ananalysis of variance was performed on these ‘corrected’constant vertical errors values. This analysis revealedthat even after subtracting these baseline downwardshifts, there was still a signiWcant between group diVer-ence in the overall magnitude of vertical constanterrors made during reversal learning (F(2,25) = 15.70;P < 0.05; Fig. 3a). Importantly, constant vertical errorsmade by PD patients were still signiWcantly larger thanthose of age-matched controls (P < 0.025). Similarresults were obtained when this subtraction analysiswas performed on absolute vertical errors. There was asigniWcant main eVect of group (F(2,25) = 12.91; P < 0.05;Fig. 3b) with absolute vertical errors made by PDpatients being signiWcantly larger than those of elderlycontrols (P < 0.025). These Wnding indicates that theimpaired ability of the PD patients to learn the verticalcomponent of the reversal discordance was not entirelydue to performance factors such as those related tomoving the arm against gravitational forces actingalong the vertical axis.
Discussion
To investigate the eVect of Parkinson’s disease andnormal aging on adaptation learning, we compared thepointing accuracy and learning rates of young controls,elderly controls and PD subjects in a novel visuomotormultistage learning task. Subjects were required toadapt to an initial sudden biaxial discordance betweenthe actual hand displacement and one visually-simu-lated in a three-dimensional virtual environment. Theythen had to adapt to the reversed visuomotor discor-
dance in quick succession. Since the basal ganglia areknown to be important for sensorimotor learning(Graybiel 2005; Barnes et al. 2005), and since learningnew visuomotor mappings depends on eYcient use ofproprioceptive information which is impaired in Par-kinson Disease (Schneider et al. 1986; Jobst et al. 1997;Zia et al. 2000; Klockgether et al. 1995; Contreras-Vidal and Gold 2004; Adamovich et al. 2001), wepredicted that PD patients would show diYculty inlearning new sensorimotor mappings. In a previousstudy, a multistage force Weld learning task was used toexamine how speciWc learning phases may diVerentiallyimpact the ability of PD subjects to learn to adapt theirmovements to an imposed force Weld (Krebs et al.2001). In that study, the learning rate of PD subjectsand age-matched controls was evaluated during earlymotor learning of an initial force Weld, during latemotor learning and during the learning of a force Weldof opposite direction. The learning rate of PD patientswas less than that of control subjects in all phases.However, the greatest deterioration in the learningindices of the PD patients was found during reversallearning. Based on this prior work, we predicted thatPD patients would show more severe deWcits in thereversal phase of learning than in the initial learning ofa visuomotor biaxial discordance.
The results showed learning phase-dependent diY-culties in both aged subjects and PD patients. First, eld-erly controls were slower than young subjects whenlearning both dimensions of the initial biaxial discor-dance. However, their performance improved duringreversal learning and as a result, elderly and youngcontrols showed similar adaptation rates during rever-sal learning. Second, in striking contrast to healthy eld-erly subjects and consistent with our predictions, PDpatients were more profoundly impaired during thereversal phase of learning. PD patients were able tolearn the initial visuomotor bias but were on averageslower than age-matched controls in adapting to thehorizontal component of the biaxial discordance. ThisdiVerence, however, did not reach the corrected signiW-cance level likely due to the large amount of intersub-ject variability. More importantly, when the biaxialdiscordance was reversed, PD patients were unable tomake appropriate movement corrections. Conse-quently, they showed slower adaptation rates andsmaller adaptation scores relative to age-matched con-trols for both dimensions of the biaxial discordance.Together, these results suggest that the ability to adaptto a sudden biaxial visuomotor discordance applied inthree-dimensional space declines in normal aging andin Parkinson disease. Furthermore, the presence oflearning rate diVerences in the PD patients relative to
123
468 Exp Brain Res (2007) 179:457–474
age-matched controls supports an important contribu-tion of basal ganglia to adaptation learning, particu-larly when the task requires a rapid reconWguration ofnewly learned visuomotor coordinations.
PD subjects are impaired when required to reconWgure visuomotor mappings
Initial learning
When the initial biaxial discordance was Wrst intro-duced, young controls immediately made large move-ment compensations in the appropriate direction,whereas some elderly controls and PD patients movedin the erroneous direction which increased their spatialerrors instead of reducing it (Fig. 2a). This inappropri-ate behavior was more frequent in the PD group whichmay account, at least in part, for the observed slowerlearning rate along the horizontal dimension. Nonethe-less, PD patients displayed similar adaptation scores asage-matched controls. These observations suggest thatduring initial learning, PD patients were able toimprove their accuracy with training but were lesseYcient than age-matched controls at using knowledgeof result (KR) about their spatial errors to select andscale appropriate movement compensations for theimposed visual perturbation.
The vertical component of the initial biaxial distor-tion moved the apparent position of the simulated armendpoint used for movement feedback higher than theactual hand position, thereby forcing subjects to pointlower than the target location presented in the VRworld. Since PD patients were already pointing lowerthan the target during the baseline condition, theimposed vertical error could not be detected so thatthere was no need for PD patients to produce largemovement adaptations along the vertical axis. Thus,the accurate vertical components of PD patients’smovements during initial learning are likely due totheir baseline vertical bias. Accordingly, and in con-trast to elderly controls, no learning curve wasobserved along the vertical dimension for PD patients.
Given this downward shift in movement endpoints,the learning of the initial biaxial discordance was sim-pliWed for PD patients, as it eliminated the require-ment to simultaneously adapt their movements to biasapplied along two dimensions of the 3D space. Theperformance of PD patients usually degrades whentask complexity increases or when multiple motor actsmust be simultaneously produced (Benecke et al. 1986;Jackson et al. 2000; Poizner et al. 2000; Pessiglioneet al. 2003). Therefore, the extent to which PD patientsare impaired during initial learning of a complex novel
visuomotor coordination in absence of vision during3D movements may have been underestimated in thisstudy. Nevertheless, the observation that PD patientswere eventually able to learn the horizontal compo-nent of the biaxial discordance is in general agreementwith studies showing that the adaptation-learning abili-ties are not completely compromised in PD patients(Stern et al. 1988; Fucetola and Smith 1997; Contreras-Vidal et al. 2002; Teulings et al. 2002; Fernandez-Ruizet al. 2003; Contreras-Vidal and Buch 2003). The largerconstant horizontal errors made by PD patients duringexposure to the initial biaxial distortion reXect modestimpairments in the ability of PD patients to adapt theirmovements to a new initial mapping between visualinformation and motor responses.
Reversal learning
In striking contrast to initial learning, diVerencesbetween PD patients and age-matched controls weremarked during exposure to the reverse discordance.Indeed, PD patients were unable to eYciently use KRfrom previous trials to implement large and appropri-ate compensations to the imposed errors. As a result,both their horizontal and vertical errors were largerthan those of elderly controls during all reversal learn-ing exposure trials. Consequently, PD patients showeda smaller adaptation magnitude score compared toage-matched controls. Together, these results providestrong evidence for reduced learning in PD patientswhen confronted with the reversed visual distortion.
In contrast to initial learning, the reverse discor-dance shifted the apparent hand location lower thanthe actual hand position. This forced subjects to pointhigher than the target location presented in the virtualenvironment. Therefore, one may ask whether fatiguemay have had more impact on the reversal than the ini-tial phase of learning, as the reversal phase requiredmore force to lift the hand higher. However, a numberof observations are inconsistent with the interpretationof fatigue being a primary factor. First, as stated above,the degraded performance of PD patients during rever-sal learning was found along both the vertical dimen-sion, where subject had to lift their arm against gravityand along the horizontal dimension where there is noincrease in force requirements. This suggests that thedegraded performance of PD patients during reversallearning is, at least in part, independent from the forcerequired to reach the target. Second, consistent withthe results of the present study, two previous studieshave shown that the magnitude of the downward shiftin movement endpoints exhibited by PD patients rela-tive to elderly controls was fairly constant across target
123
Exp Brain Res (2007) 179:457–474 469
elevations (Poizner et al. 1998; Adamovich et al. 2001).That is, the vertical biases were similar whether PDpatients pointed to the central target (shoulder eleva-tion) or to the most elevated target (25 cm higher, sim-ilar in elevation to the reversal target in this study).Therefore, in contrast to what would be expected fromfatigue related to arm movements performed againstgravity, PD patient’s undershooting did not increasewith increase in the vertical distance to the target. Fur-ther, based on this previous observation, we performeda subtraction analysis to test whether the baseline ver-tical bias of the PD patients is a major contributor tothe much slower adaptation rates of PD patientsobserved along the vertical dimension. We subtractedthe mean constant vertical error made by each PD sub-ject during the baseline from the constant vertical errormade for each trial during reversal learning. This anal-ysis revealed that even after subtracting out the sys-tematic downward shift PD patients showed duringbaseline, PD patients still pointed signiWcantly lowerthan age-matched controls during reversal learning(Fig. 3a). For these reasons and because PD patientswere not presenting any joint limitations that wouldhave prevent them from pointing at the reversal targetelevation, we view the impaired ability of PD patientsto learn the vertical component of the reversal discor-dance as a learning diYculty not accounted for byfatigue due to moving against the gravitational forcesacting along the vertical axis.
Several studies have reported systematic hypometriain PD (Flowers 1976; Klockgether and Dichgans 1994).Therefore, it might be possible that the deWcitsobserved during reversal learning were due to targetundershooting. However, results of the present studymake this explanation unlikely. The vertical distancebetween the initial position of the hand (on the rightthigh) and the target location was greater during rever-sal learning than initial learning. In contrast, the hori-zontal distance between the initial position of the handand the target was greater during initial learning thanduring reversal learning. Despite these distance diVer-ences relative to start location, PD patients wereimpaired in their ability to consistently reach the targetlocation along both the vertical and the horizontaldimensions during the reversal phase of learning. Fur-ther, results from previous studies of 3D reachingmovements to the same target locations in PD patientswith similar severity levels are not consistent with thetarget undershooting explanation. In the study by Poiz-ner et al. (1998) and in the present study, movementswere initiated from a horizontal start position. There-fore, the general direction of the movements duringpointing was upward. As a result, hypometric move-
ments would predict a downward bias. In contrast, inthe study by Adamovich et al. (2001), movementstoward the same target spatial locations were initiatedfrom a vertical position such that hypometric move-ments would predict an upward bias. Instead, a system-atic depression in endpoint elevation was observed.Further and as mentioned previously, this depressionin endpoint elevation did not show any signiWcantincrease as a function of the vertical distance to the tar-get in both studies (Poizner et al. 1998; Adamovichet al. 2001). Therefore, it appears most likely that theobservation of a downward shift in three diVerent stud-ies with diVerent PD subjects, using diVerent initialarm postures, reXects inappropriate movement scalingalong the vertical dimension.
Sensory processing and integration are importantfactors that can accelerate or retard the eVective learn-ing of novel visuomotor coordinations. Therefore, onemay ask to what degree the observed learning deWcitsmay in fact be due to impaired sensory processing. Ithas been known for some time that PD patients havecertain visual-spatial deWcits, among them a lack ofprecise perception of the visual-vertical. Proctor et al.(1964), for example, showed that PD patients hadimpaired judgment of visual-vertical under conditionsof body tilt. Likewise, Azulay et al. (2002) found thatPD patients showed larger errors than controls whenadjusting a rod to visual-vertical when the rod wasinside a tilted frame. Moreover, Lee et al. (2001, 2002)report that PD patients with primarily left-sided symp-toms show both rightward and downward biases incoding left and upper visual space respectively, deWcitsakin to visual neglect. Therefore, one possibility is thatreversal learning deWcits of PD patients, i.e., greaterspatial errors when pointing at the upper right target,are due to impaired perception of visual space. How-ever, we feel that this explanation is unlikely sinceroughly half of the patients tested in the present studywere more aVected on each side (Table 1). Further,some investigators have found that PD patients shownormal perception of visual-vertical (e.g., Bronsteinet al. 1996). More telling however, is that 3D reachingstudies using similar targets to those in the presentexperiment demonstrate that visual perceptual deWcitscould not account for the pattern of reaching perfor-mance. For example, Adamovich et al. (2001) showedthat PD patients had signiWcant elevation errors whenreaching to either actual or remembered targets whenthey could not see their arms. However, when an illu-minated LED was placed on the Wngertip, so thatvision of the arm endpoint was visible throughout themovement, the PD patients’s elevation errors normal-ized despite the fact that subjects were pointing to
123
470 Exp Brain Res (2007) 179:457–474
remembered targets. Thus, the initial visual perceptionof target elevation had to be intact (at least within theaccuracy constraints of these reaching experiments)since providing vision of the Wngertip during move-ments normalized the elevation errors (see also Keij-sers et al. 2005 who found that providing PD patientswith vision of the Wngertip and an illuminated refer-ence frame normalized variable errors for reaches toremembered 3D targets). Numerous studies are nowindicating that PD patients have deWcient propriocep-tion or sensorimotor integration mechanisms (Schnei-der et al. 1986; Jobst et al. 1997; Zia et al. 2000;Klockgether et al. 1995; Adamovich et al. 2001; Mas-chke et al. 2003). Accurate positioning of the hand invertical space challenges proprioceptive processing to agreater degree than does positioning in the horizontaldimension, since vertical movements require compen-sation for gravitational forces. Thus we feel that alongthe vertical dimension, altered proprioceptive or sen-sorimotor integration processing appeared to haveinterfered with the ability of PD patients to performaccurate movements, as reXected by the large and con-stant downward bias present in all task phases.
The sudden introduction of the opposite distortionafter subjects adapted to the initial biaxial bias induceda substantial error. Indeed, there was a double errorsignal in reversal learning compared to the initial learn-ing, since once subjects had learned to point down andto the left in initial learning, they then had to learn topoint up and to the right in reversal learning. In face ofthis novel situation, subjects not only had to compen-sate for that large visuomotor error, but were requiredto reconWgure a newly learned visuomotor mapping.Simulation studies of basal ganglia function show thatunlearning a previous response set is a separate processfrom learning a new one and has a diVerential learningrate (Berns and Sejnowski 1998). All of these require-ments represent a great challenge and may explain themarked degradation of PD patient’s performancewhen the initial distortion was reversed.
In this complex reversal condition in which subjectsmust both reconWgure their visuomotor mapping andsimultaneously compensate for the horizontal and ver-tical components of the biaxial discordance, PDpatients appeared to have adopted a decompositionstrategy. Along the horizontal dimension, they imple-mented large and inappropriate movement changesfrom trial to trial (oscillations). Along the verticaldimension, however, they had both a relatively stablelevel of constant vertical errors and a similar level oftrial-to-trial variability during late exposure trials asboth control groups, despite vertical errors that werefour times greater then those of both control groups.
These results suggest that PD patients were repeti-tively pointing at similar, although inappropriate spa-tial locations. Perseveration behavior has beenpreviously reported in Parkinson’s disease (Ebersbachet al. 1994; StoVers et al. 2001). Further, a number ofstudies have suggested that PD patients are impairedwhen required to modify their behavioral response inface of changing contingencies both in the cognitiveand the motor domains. For instance, Haaland et al.(1997) studied tracking performance using three diVer-ent speeds. PD patients showed slower learning ratesduring experiments with a randomized design but nor-mal rates with a blocked design. These Wndings sug-gested that PD patients have more ‘rigid’ motorprograms and cannot rapidly switch performance ‘set’(Cools et al. 1984; Richards et al. 1993, Cronin-Golomb et al. 1994, Hayes et al. 1998, Monchi et al.2004). Likewise, Vakil and Herishanu-Naaman (1998)found that PD patients showed diYculties with learn-ing motor sequencing when required to switch fromone motor program to the next. Further, in a previousstudy, we examined in which sensorimotor learningphase PD patients’ deWcits are more apparent in amulti-stage force-Weld learning task (Krebs et al. 2001).As previously mentioned, the greatest deterioration inlearning indices of the PD patients was in reversallearning, a phase which required a rapid reconWgura-tion of sensorimotor response. These results, togetherwith those of the present study, suggest that the basalganglia might be a key center for context switching.
However, it is also possible that the impairments thatPD patients show when confronted with the reversedvisuomotor mapping are not the expression of theirgeneral inability to switch set. Rather, their markedlydeteriorated performance may reXect profound diYcul-ties in implementing appropriate compensations whenlarge visual errors are imposed. It has been recentlyproposed that adaptation to gradual and sudden newvisuomotor mappings engage two distinct and comple-mentary learning processes (Florian et al. 1997; Malfaitand Ostry 2004; Robertson and Miall 1999; Contreras-Vidal and Buch 2003). Adapting to gradual changes invisual feedback involve small movement correctionsthat require small reconWgurations of the visuomotormap. By contrast, a sudden large perturbation in visualfeedback requires a search in visuomotor space to mini-mize errors that are more likely consciously detected bysubjects. In support of distinct learning processes, theability of young controls to reconWgure visuomotormappings in gradual versus sudden experimental con-texts led to diVerent pattern of results both for learninga visual distortion and for learning a dynamic one (Flo-rian et al. 1997; Malfait and Ostry 2004). Although, no
123
Exp Brain Res (2007) 179:457–474 471
study compared adaptation to gradual and sudden dis-tortions in Parkinson Disease, Contreras-Vidal andBuch (2003) reported marked deWcits in PD patientswhen learning a sudden large visual distortion createdby a 90° cursor rotation. They concluded that theobserved deWcits reXect the critical role of the basalganglia in tasks that are initially eVortful, such as whenadapting to a large sudden distortion (Contreras-Vidaland Buch 2003). This interpretation is in agreementwith results of a recent study examining PD patient’sability to correct for movement errors induced by targetjumps of diVerent sizes (Desmurget et al. 2004). In thatstudy, PD patients were easily able to compensate forsmall subliminal target jumps but their performance sig-niWcantly deteriorated when large consciously detectedtarget jumps were imposed. Along the same lines as theadaptation studies mentioned above, Desmurget et al.(2004) suggested that corrections for these two types oferrors rely on distinct processes and that basal gangliacircuits are more critically involved in correcting largeerrors.
Given that this study and all previous studies of sen-sorimotor adaptation to sequential perturbations usedopposite distortions which double the error signal, itremains unclear whether both the marked impairmentsof PD patients during reversal learning, and theobserved greater involvement of basal ganglia duringthe reversal phase of learning, reXect mainly a majorcontribution of basal ganglia in context switching or animportant contribution of basal ganglia in correctingfor large visual errors. Future studies in which the sizeand the order of visual distortions are systematicallymanipulated are needed to isolate the participation ofbasal ganglia in these two important and complemen-tary aspects of adaptive motor control. Nevertheless,this study provides the Wrst evaluation of the eVect ofPD and normal aging on the ability to adapt naturalmovements performed in three-dimensional space. Thepresence of sensorimotor learning deWcits in PD itselfis a controversial issue. Our results that PD patientswere able to learn the initial visual distortion andshowed marked degradation in performance relative toage-matched controls during the reversal phase oflearning indicates that PD produces selective deWcits invisuomotor learning when the task is complex involv-ing either a large visual distortion or when subjectsmust learn to adapt to reversed visual distortions inquick succession. Further, this study extends previousWndings of impaired sensorimotor learning in PD andsupports imaging studies showing that the basal gangliaare involved when subjects must learn novel visuomo-tor associations sequentially presented (Shadmehr andHolcomb 1999; Krebs et al. 1998; Krakauer et al. 2004).
In the context of the present study, however, bothadapting to the initial smaller visual distortion and to thereversed larger visual distortion involved the selectionand elaboration of appropriate motor responses. Thus,the somewhat slower adaptation rate observed duringinitial learning and the markedly deteriorated perfor-mance of PD patients during reversal learning may bothinvolve, as suggested by Shadmehr and Holcomb (1999)and Krebs et al. (2001), the expression of a key role ofbasal ganglia in the selection of appropriate motor pro-grams. Such selection would involve the learning of anew motor response in a novel visuomotor context, and,in the case of reversal learning, would involve as well thesuppression of a just learned but now inappropriatemotor response (Jueptner et al. 1996; Schmidt 1998;Kropotow and Etlinger 1999; Brooks 2000).
PD patients exhibited a large downward vertical bias
Accurate reaching in the vertical dimension requirescompensation for gravity. Therefore, it is most depen-dent on proprioceptive information. Since several stud-ies have reported impaired proprioceptive processingin Parkinson disease (Schneider et al. 1986; Jobst et al.1997; Zia et al. 2000), the systematic downward shift ofmovement endpoints exhibited by PD patients in alltask phases may likely represent their inability to uti-lize proprioceptive inputs to compensate for gravita-tional force. Two previous studies of three-dimensionalreaching movements support this view (Poizner et al.1998; Adamovich et al. 2001).
Visuomotor learning in normal aging
In the present study, healthy elderly subjects showed anoverall slower adaptation rate to the initial biaxial dis-tortions compared to young controls. As mentioned inthe introduction, the eVect of normal aging on visuomo-tor learning is controversial (Fernandez-Ruiz et al.2000; Bock and Schneider 2002; Buch et al. 2003). As isthe case for PD, studies Wnd either a slower adaptationrate or a smaller aftereVect from one learning paradigmto another suggesting that the inXuence of aging onvisuomotor learning is context dependent. Recently,Buch et al. (2003) suggested that much of the variabilityin the patterns of results could be explained by thenature of the perturbation used in the diVerent studies.They tested the ability of aged subjects to reconWguretheir visuomotor mapping when exposed both to agradual and a sudden visual distortion. Elderly subjectsshowed reduced performance indices in both visuomo-tor conditions. When exposed to the gradual distortion,they produced smaller aftereVects, whereas in the
123
472 Exp Brain Res (2007) 179:457–474
sudden distortion they exhibited more slowly decreas-ing learning curves (Buch et al. 2003). These resultswere interpreted as the manifestation of two distinctlearning deWcits. The smaller aftereVect was suggestedto reXect reduced implicit learning, whereas the sloweradaptation rate was considered the expression of cogni-tive factors interfering with rapid selection of an appro-priate strategy to reduce large visual errors (Buch et al.2003). This interpretation is consistent with previousstudies using dual task paradigms suggesting that adap-tation learning requires a higher proportion of theavailable computational resources in elderly subjectsrelative to young controls (Bock and Schneider 2002).The Wnding of the present study that aged subjectsmade greater errors than young controls during earlyexposure to the initial biaxial distortion before normal-izing their performance supports these interpretations.Interestingly, however, when aged subjects were sud-denly confronted with the reverse discordance and hadto correct for large visual errors that were of doublesize, unlike PD subjects their adaptation rates were notsigniWcantly diVerent from young controls for eitherdimension of the biaxial discordance. Therefore,although aged subjects were slower learners than youngcontrols during initial learning, possibly due to reducedstrategic processes, they were able to learn generalstrategies to reconWgure their visuomotor mapping andto apply them to improve their reversal learning, anability referred as ‘learning to learn’ (Bock and Schnei-der 2001, 2002). The results of the present study areconsistent with these previous studies in the context of amultistage learning task involving two opposite visualdistortions presented in quick succession.
The capacity to Xexibly adapt movements to novelenvironmental contexts is central to normal everydayhuman behavior. By using a multistage visuomotorlearning task, this study provided the Wrst evaluation ofthe selective eVect of Parkinson Disease and normalaging on the ability to adapt natural movements per-formed in three-dimensional space. Results showedthat adaptation-learning abilities decline in both agingand Parkinson disease. However, the more stronglyimpaired performance of PD patients when confrontedwith a reversed visuomotor distortion suggests thatParkinson’s disease produces selective deWcits ratherthan a profound generalized impairment in the abilityto learn novel visuomotor coordinations. The presentWndings are consistent with the recent hypothesis thatdopamine depletion results in a loss of functional seg-regation within basal ganglia circuits which may inter-fere with a number of motor and cognitive abilitiesincluding movement selection and context switching(Pessiglione et al. 2005).
Acknowledgments This research was supported in part by anFRSQ and CIHR Postdoctoral Fellowship to JM and by NIHGrant 7 R01 NS36449 to UCSD (HP).
References
Adamovich SV, Berkinblit MB, Henning W, Sage J, Poizner H.(2001) The interaction of visual and proprioceptive inputs inpointing to actual and remembered targets in Parkinson’sdisease. Neuroscience 1027–1041
Alexander GE (1994) Basal ganglia-thalamocortical circuits:their role in control of movements. J Clin Neurophysiol11:420–431
Azulay JP, Mesure S, Amblard B, Pouget J (2002) Increased visu-al dependence in Parkinson’s disease. Percept Mot Skills95(3 Pt 2):1106–1114
Barnes TD, Kubota Y, Hu D, Jin DZ, Graybiel AM (2005) Activ-ity of striatal neurons reXects dynamic encoding and recod-ing of procedural memories. Nature 437(7062):1158–1161
Benecke R, Rothwell JC, Dick JP, Day BL, Marsden CD (1986)Performance of simultaneous movements in patients withParkinson’s disease. Brain Aug 109(Pt 4):739–57
Berns SB, Sejnowski TJ (1998) A computational model of how thebasal ganglia produce sequences. J Cogn Neurosci 10:108–121
Bock O, Schneider S (2001) Acquisition of a sensorimotor skill inyounger and older adults. Acta Physiol Pharmacol Bulg26(1–2):89–92
Bock O, Schneider S (2002) Sensorimotor adaptation in youngand elderly humans. Neurosci Biobehav Rev 26:761–767
Bronstein AM, Yardley L, Moore AP, Cleeves L (1996) Visuallyand posturally mediated tilt illusion in Parkinson's disease andin labyrinthine defective subjects. Neurology 47(3):651–656
Brooks DJ (2000) Imaging basal ganglia function. J Anat196:543–554
Buch ER, Young S, Contreras-Vidal JL (2003) Visuomotor adap-tation in normal aging. Learn Mem 10:55–63
Contreras-Vidal JL, Buch ER (2003) EVects of Parkinson’s deas-ese on visuomotor adaptation. Exp Brain Res 150:25–32
Contreras-Vidal JL, Gold DR (2004) Dynamic estimation ofhand position is abnormal in Parkinson’s disease. Parkinson-ism Relat Disord 10(8):501–506
Contreras-Vidal JL, Teulings HL, Stelmach GE, Adler CH(2002) Adaptation to changes in vertical display gain duringhandwriting in Parkinson’s disease patients, elderly andyoung controls. Parkinsonism Relat Disord 9:77–84
Cools AR, Van Den Bercken JH, Horstink MW, Van Spa-endonck KP, Berger HJ (1984) Cognitive and motor shiftingaptitude disorder in Parkinson’s disease. J Neurol NeurosurgPsychiatry 47(5):443–453
Cronin-Golomb A, Corkin S, Growdon JH (1994) Impairedproblem solving in Parkinson’s disease: impact of a set- shift-ing deWcit. Neuropsychologia 32:579–593
Desmurget M, Jordan M, Prablanc C, Jeannerod M (1997) Con-strained and unconstrained movements involve diVerentcontrol strategies. J Neurophysiol Mar 77(3):1644–1650
Desmurget M, Gaveau P, Vindras RS, Turner E, Broussolle SThobois (2004) On-line motor control in patients with Par-kinson’s disease. Brain 127(Pt 8):1755–1773 Epub Jun 23
Doyon J, Penhune V, Ungerleider LG (2003) Distinct contribu-tion of the cortico-strial and cortico cerebellar systems tomotor skill learning. Neuropsychologia 41:252–262
Ebersbach G, Hattig H, Schelosky L, Wissel J, Poewe W (1994)Perseverative motor behaviour in Parkinson’s disease.Neuropsychologia 32(7):799–804
123
Exp Brain Res (2007) 179:457–474 473
Fahn S, Elton RL (1987) UniWed Parkinson’s disease rating scale.In: Fahn S, Marsden C, Calne D Goldstein M (eds) Recentdevelopments in Parkinson’s disease. Macmillan:29304, NewYork
Fernandez-Ruiz J, Hall C, Vergara P, Diaz R (2000) Prism adap-tation in normal aging: Slower adaptation rate and larger af-ter eVect. Brain Res Cogn Brain Res 9:223–226
Fernandez-Ruiz J, Diaz R, Hall-Haro C, Vergara P, Mischner J,Nunez L, Drucker-Colin R, Ochoa A, Alonso ME (2003)Normal prism adaptation but reduced after-eVect in basalganglia disorders using a throwing task 18:689–694
Florian A, Kagerer, Contreras-Vidal JL, Stelmach GE (1997)Adaptation to gradual as compared with sudden visuo-motordistortions. Exp Brain Res Jul 115(3):557–561
Flowers K (1976) Visual “closed-loop” and “open-loop” charac-teristics of voluntary movement in patients with parkinson-ism and intention tremor. Brain 99:269–310
Fucetola R, Smith MC (1997) Distorted visual feedback eVects ondrawing in Parkinson’s disease. Acta Psychol (Amst) 95:255–266
Ghilardi MF, Alberoni M, Rossi M, Franceschi M, Mariani C,Fazio F (2000) Visual feedback has diVerential eVects onreaching movements in Parkinson’s and Alzheimer’s disease.Brain Res 876:112–113
Grafton ST, Hazeltine E, Ivry R (1995) Functional mapping of se-quence learning in normal humans. J Cogn Neurosci 7:497–510
Graybiel AM (2004) Network-level neuroplasticiy in cortico-bas-al ganglia pathways, Parkinsonism Rel Disord 10:293–296
Graybiel AM (2005) The basal ganglia: learning new tricks andloving it. Curr Opin Neurobiol 15(6):638–644
Graydon FX, Friston KJ, Thomas CG, Brooks VB, Menon RS(2005) Learning-related fMRI activation associated with arotational visuo-motor transformation. Cogn Brain Res22:373–383
Haaland KY, Harrington DL, O’Brien S, Hermanowicz N (1997)Cognitive-motor learning in Parkinson’s disease. Neuropsy-chology 11:180–186
Hayes AE, Davidson MC, Keele SW, Rafal RD (1998) Toward afunctional analysis of the basal ganglia. J Cogn Neurosci10:178–198
Hoehn MM, Yahr MD (1967) Parkinsonism: onset, progressionand mortality. Neurology 17:427–442
Hoover JE, Strick PL (1993) Multiple output channels in thebasal ganglia. Science 259:819–821
Houk JC, Wise SP (1995) Distributed modular architectures link-ing basal ganglia, cerebellum, and cerebral cortex: their rolein planning and controlling action. Cereb Cortex 5:95–110
Imamizu H, Miyauchi S, Tamada T, Sasaki Y, Takino R, Putz Bet al (2000) Human cerellar activity reXecting an acquiredinternal model of a new tool. Nature 403:192–195
Inoue K, Kawashima R, Satoh K, Kinomura S, Sugiura M, GotoR, Ito M, Fukuda H (2000) A PET study of visuomotorlearning under optical rotation. Neuroimage 11:505–516
Jackson GM, Jackson SR, Hindle JV (2000) The control ofbimanual reach-to-grasp movements in hemiparkinsonianpatients. Exp Brain Res 132:390–398
Jobst EE, Melnick ME, Byl NN, Dowling GA, AminoV MJ(1997) Sensory perception in Parkinson disease. Arch Neu-rol 54:450–454
Jueptner M, Jenkins IH, Brooks DJ, Frackowiak RS, PassinghamRE (1996) The sensory guidance of movement: a comparisonof the cerebellum and basal ganglia. Exp Brain Res 112:462–474
Keijsers NL, Admiraal MA, Cools AR, Bloem BR, Gielen CC(2005) DiVerential progression of proprioceptive and visual
information processing deWcits in Parkinson’s disease. Eur JNeurosci 21(1):239–248
Klockgether T, Dichgans J (1994) Visual control of arm move-ments in Parkinson's disease. Mov Disord 9(1):48–56
Klockgether T, Borutta M, Rapp H, Spieker S, Dichgans J (1995)A defect of kinesthesia in Parkinson’s disease. Mov Disord10:460–465
Krakauer JW, Ghilardi MD, Mentis M, Barnes A, Veytsman, Ei-delberg D, Ghez C (2004) DiVerential cortical and subcorti-cal activations in learning rotations and gains for reaching: apet study. J Neurophysiol 91:924–933
Krebs HI, Brashers-Krug T, Rauch SL, Savage CR, Hogan N,Rubin RH, Fischman AJ, Alpert NM (1998) Robot aidedfunctional imaging: application to a motor learning study.Hum Brain Mapp 6:59–72
Krebs HI, Hogan N, Hening W, Adamovich SV, Poizner H (2001)Procedural motor learning in Parkinson’s disease. Exp BrainRes 141:425–437
Kropotow JD, Etlinger SC (1999) Selection of actions in the basalganglia-thalamocortical circuits: review and model. Int J Psy-chophysiol 31:197–217
Lee AC, Harris JP, Atkinson EA, Fowler MS (2001) Evidencefrom a line bisection task for visuospatial neglect in LeftHemiparkinson’s disease. Vis Res 41:2677–2686
Lee AC, Harris JP, Atkinson EA, Nithi K, Fowler MS (2002)Dopamine and the representation of the upper visual Weld:evidence from vertical bisection errors in unilateral Parkin-son’s disease. Neuropsychologia 40:2023–2029
Malfait N, Ostry DJ (2004) Is interlimb transfer of force-Weldadaptation a cognitive response to the sudden introductionof load? J Neurosci 24(37):8084–8089
Maschke M, Gomez CM, Tuite PJ, Konczak J (2003) Dysfunctionof the basal ganglia, but not the cerebellum, impairs kinaes-thesia. Brain 126:2312–2322
Messier J, Kalaska JF (1997) DiVerential eVect of task conditionson errors of direction and extent of reaching movements.Exp Brain Res 115:469–478
Monchi O, Petrides M, Doyon J, Postuma RV, Worsley K, Da-gher A (2004) Neural bases of set-Shifting deWcits in Parkin-son’s disease. J Neurosci 24(3):702–710
Pessiglione M, Guehl D, Agid Y, Hirsch EC, Feger J, Tremblay L(2003) Impairment of context-adapted movement selectionin a primate model of presymptomatic Parkinson’s disease.Brain 126(Pt 6):1392–1408
Pessiglione M, Guehl D, Rolland AS, Francois C, Hirsch EC, FegerJ, Tremblay L (2005) Thalamic neuronal activity in dopamine-depleted primates: evidence for a loss of functional segregationwithin basal ganglia circuits. J Neurosci 25(6):1523–1531
Poizner H, Fookson O, Berkinglit MB, Hening W, Feldman G,Adamovich SV (1998) Pointing to remembered targets in 3Dspace in Parkinson’s disease. Motor Control 2:251–277
Poizner H, Feldman AG, Levin MF, Berkinblit MB, Hening WA,Patel A, Adamovich SV (2000) The timing of arm-trunkcoordination is deWcient and vision-dependent in Parkin-son’s patients during reaching movements. Exp Brain Res133(3):279–292
Proctor F, Riklan M, Cooper IS, Teuber H-L (1964) Judgmentof visual and postural vertical by Parkinsonian patients.Neurology 14:282–293
Rauch SL, Whalen PJ, Savage CR, Curran T, Kendrick A, BrownHD et al (1997) Striatal recruitment during an implicit se-quence learning task as measured by functional magneticresonance imaging. Hum Brain Mapp 5:124–132
Richards M, Cote LJ, Stern Y (1993) Executive function in Par-kinson’s Disease: set-shifting or set-maintenance? J Clin ExpNeuropsychol 15 2:266–279
123
474 Exp Brain Res (2007) 179:457–474
Robertson EM, Miall RC (1999) Visuomotor adaptation duringinactivation of the dentate nucleus; NeuroReport10(5):1029–1034
Schmidt WJ (1998) Dopamine-glutamate interactions in the basalganglia. Amino Acids 14:5–10
Schneider JS, Diamond SG, Markham CH (1986) DeWcits in oro-facial sensorimotor function in Parkinson’s disease. AnnNeurol 19:275–282
Seidler RD, Noll DC, Chintalapati P (2006) Bilateral basal gan-glia activation associated with sensorimotor adaptation. ExpBrain Res [Epub ahead of print]
Shadmehr R, Holcomb HH (1999) Inhibitory control of compet-ing motor memories. Exp Brain Res 126:235–251
Stern Y, Mayeux R, Hermann A, Rosen J (1988) Prism adapta-tion in Parkinson’s disease. J Neurol Neuroseurg Psychiatry51:1584–1587
StoVers D, Berendse HW, Deijen JB, Wolters EC (2001) Motorperseveration is an early sign of Parkinson’s disease. Neurol-ogy 57:2111–2113
Teulings HL, Contreras-Vidal JL, Stelmach GE, Adler CH(2002) Adaptation of handwriting size under distorted visu-al feedback in patients witn parkinson’s disease and elderlyand young controls. J Neurol Neurosurg Psychiatry 72:315–324
Vakil E, Herishanu-Naaman S (1998) Declarative and procedurallearning in Parkinson’s disease patients having tremor orbradykinesia as the predominant symptom. Cortex 34:611–620
Zia S, Cody F, O’Boyle D (2000) Joint position sense is impairedby Parkinson’s disease. Ann Neurol 47:218–228
123
