Impairments in the of manual skill in patients · skill and (b) achieve a degree ofautomaticity in the performance ofthat skill. Weinvestigated two kinds of learning using continuous

Journal of Neurology, Neurosurgery, and Psychiatry 1986;49:661-668

Impairments in the learning and performance of a newmanual skill in patients with Parkinson's diseaseCD FRITH, CA BLOXHAM, KN CARPENTER

From the Division ofPsychiatry, MRC Clinical Research Centre, Harrow, UK

SUMMARY Twelve patients with Parkinson's disease learned two novel skills in which they had totrack a target by moving a joystick. In task 1 they had to learn to anticipate the movements of a semipredictable target. In task 2 they had to learn a novel control system in which the movements of thejoystick were mirror reversed in relation to the computer screen.On each task they performed twosessions of three minutes continuous practice separated by a 10 minute rest. In both tasks thepatients performed much worse than the controls, but showed clear evidence of learning, particu-larly after the ten minute rest. Detailed examination of their performance suggested that the skillwas becoming automatic, releasing attention for aspects of the task that could not be learned. Themajor difference from the controls appeared during the first minute of each practice session whenthe controls showed a marked improvement in performance while the patients did not. We suggestthat this rapid but temporary improvement in performance reflects the acquisition of a motor "set"whereby existing motor programs or skills are modified to suit the task currently in hand. Weconcluded that patients with Parkinson's disease have difficulty in maintaining such sets.

A number of attempts have been made in recent yearsto characterise the impairments of motor controlassociated with Parkinson's disease in terms of mod-els of control derived from the study of normal volun-teers. Thus Flowers' has suggested that patients withParkinson's disease have a problem with "open loop"control, but not with "closed loop" control.Marsden2 has suggested that Parkinsonian patientshave difficulty in the "automatic execution of learnedmotor plans". Bloxham et al3 propose that thesepatients have difficulty in initiating a motor plan, butno difficulty in executing the plan once it has beeninitiated. Wing and Miller4 suggest that Parkinson'sdisease patients have a deficit in activating a pre-planned movement when the trigger is provided inter-nally. The implicit assumption of all these authors isthat there should be in Parkinson's disease a discreteimpairment in one of the components of motor con-trol corresponding to the relatively discrete brainlesions found in this disease. However, while there isa wealth of detailed clinical description in the litera-ture about the impairment of motor control in Par-kinson's disease there are relatively few experimentalstudies. In particular there are few studies on theAddress for reprint requests: Dr CD Frith, CRC Division ofPsychiatry, Harrow HAI 3UJ, UK.Received 5 March 1985 and in revised form 9 October 1985.Accepted 20 October 1985

ability of Parkinson's disease patients to learn a novelmotor skill. Marsden2 reports some unpublished datawhich suggests that, unlike normal controls, someParkinson's disease patients fail to learn a trackingtask when given information about target predict-ability. On the other hand, Day et al' found that Par-kinson's disease patients selected to have no cognitiveabnormalities did show an improvement in per-formance (decreased lag) when made aware that a tar-get was moving predictably. This suggests that somedegree of learning was taking place. In this paper weshall describe the learning and performance of twopursuit tracking tasks by Parkinson's disease patientsand normal controls. In one task the movement of thetarget was partially predictable while in the other thecontrol system was mirror reversed.The learning of pursuit tracking tasks has been

studied extensively since the 1920s6 and a number ofdifferent aspects of the learning and performance ofthese tasks can be distinguished. Typically, with com-puter controlled equipment, the subject controls theposition of the pointer on a screen by moving ajoystick. His task is to keep the pointer as near aspossible to a continuously moving target.The most interesting paradign for studying the

learning of such a task is one in which short periods ofwork (circa 5 minutes) are separated by periods of rest(circa 10 minutes). Using this paradigm two phenom-

661

by copyright. on June 7, 2020 by guest. P

rotectedhttp://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.49.6.661 on 1 June 1986. D

ownloaded from

http://jnnp.bmj.com/

662ena can reliably be observed.6 (1) At the beginning ofeach session of work there is a rapid improvement ofperformance completed within 30 seconds. This effectis more marked in the later sessions of work. The sizeof this upswing is also dependent on the length of thepreceding rest, being larger with longer rests. Thisthen appears to be a temporary improvement in per-formance that is acquired at the beginning of eachsession of work and dissipates during rest. To dis-tinguish it from permanent learning this phenomenonis often referred to as the acquisition of set. (2) Afterthe initial upswing, performance within sessions tendsto remain constant or even decline. However, after arest there is a marked and permanent improvement inoverall performance. This improvement from pre- topost-rest tends to be greater than the improvementduring a session of continuous work. This post-restimprovement is known as reminiscence.How these changes in performance relate to the

underlying learning processes remains controversial.Following Fitts7 it is useful to distinguish three stagesin the learning of a new skill. (1) In the cognitive stagethe subject must acquire (implicitly or explicitly) theknowledge needed to perform the skill. Thus if targetmovement is predictable the subject must discoverthis by observing that the target always moves to pos-ition X2 after it has been in position Xl and so on.Having observed this he then knows that he can antic-ipate the position of the target by moving to positionX2 when he sees the target in position Xl (thusachieving predictive performance using feedback).Similarly if his joystick control is mirror reversed hemust discover that when the target moves to positionXl he must move the joystick to position Xn- 1. Theseitems of knowledge are specific to the skill beinglearned. (2) In the strategic stage the subject mustsomehow use his knowledge to produce appropriatemovements. Since, at least in the early stages of prac-tice, the new skill is not yet available this must bedone by modifying already existing skills (or motorprograms) for this new purpose. At this stage per-formance will need much mental effort and consciousattention. (3) In the automatic stage the new skill hasfinally been learned. That is a specific motor programhas developed for this skill and the subject no longerneeds to modify programs from other skills. Atthis stage performance needs little mental effort orattention.

As a skill becomes automatic, performance of itinterferes less with other tasks that are being per-formed at the same time. Frith and Lang8 exploitedthis prediction in order to study the development ofautomaticity in a tracking task. When both the hori-zontal and the vertical movements of the target wereunpredictable then there was no improvement in per-formance. However, when one component was pre-

Frith, Bloxham, Carpenter

dictable then (a) there was an improvement in theperformance of this component during a session ofwork and (b) an improvement in the other unpredict-able component after a rest. This result indicated thatsubjects had learned to track the predictable com-ponent. In addition, after rest, the tracking of thiscomponent had become more automatic and there-fore interfered less with performance of the other,unpredictable component. The authors concludedthat the phenomenon of reminiscence could beexplained in terms of increased automaticity, whilewithin sessions improvements could be interpreted interms of strategic learning.

In the experiment reported here we have used Frithand Lang's methodology to investigate whether Par-kinson's disease patients can (a) learn a new manualskill and (b) achieve a degree of automaticity in theperformance of that skill. We investigated two kindsof learning using continuous tracking tasks. In onetask (semi-predictable) the movements of the targetwere predictable in the horizontal dimension, butunpredictable in the vertical dimensions. Thus a sub-ject could learn to predict the movements of the targetin the horizontal dimension. This learning is of partic-ular interest in relation to Parkinson's disease sinceFlowers' has proposed that Parkinson's diseasepatients have difficulty in making anticipatory move-ments and can only follow a target using visual feed-back to correct their errors. In the other task (mirrorreversed) the movements of the target were unpredict-able in both dimensions. However, the joystick con-trol was mirror reversed so that in order to stay ontarget the subject had to move the joystick to the rightwhen the target moved to the left. Thus, for the hori-zontal dimension, he had to learn to make a novelresponse to the visual feedback he received abouterrors. If Flowers is correct then Parkinson's diseasepatients should be less impaired on this mirrorreversed task than on the semi-predictable one. It ispossible to study within session (strategic)improvements in both tasks, but there is nothing inthe literature to suggest whether or not Parkinson'sdisease patients should be impaired in this aspect oflearning.

Method

Subjects had to track a target which moved in two dimen-sions on the face of a VDU with a maximum amplitude of1Ocm. The target was a 4mm square and the screen pointerwas a cross which fitted exactly inside the square. The sub-ject's task was to match the movement of the target asclosely as he could, keeping the cross, if possible, inside thesquare. The position of the cross was controlled by means ofa 17 mm long joystick which the subject held in his preferredhand. The subject's elbow rested on a foam rubber pad and



j.com/

J Neurol N


ownloaded from


Skill learning in Parkinson's diseaseTable 1 Track components (Hz)

Semi-predictable target(task 1)

Horizontal 0 30 0 30 0 30Vertical 006 0 11 023

Mirror reversed control system(task 2)

Horizontal 0-03 0 21 0-32Vertical 0 05 0 19 0-33

thus controls of the joystick required movements of the wristand forearm.Control of target movement Movement of the target in thevertical and horizontal directions was controlled indepen-dently. For each of these directions of target movement theposition of the target every 25ms was determined from acombination of sine waves.Task I The horizontal component of this task was a singlesine wave of 0 3 Hz. It was thus repetitive and predictable.The vertical component of this task was formed from a com-bination of three sine waves of unrelated frequencies (seetable 1). Thus this component was irregular and unpredict-able. This task is essentially the same as the task II used byFrith and Lang8 except that the frequencies are slower andthus the task is easier and more suitable for older subjects.Task 2 In this task both the horizontal and the verticalcomponents were composed from three unrelated sine waves(see table 1) and so all aspects of target movement wereunpredictable. In addition the relation between joystick andscreen pointer was reversed for the horizontal direction sothat the movements of the joystick to the left caused thecross on the screen to move to the right. For vertical move-ments of the joystick the conventional relationship wasmaintained. Since this "mirror reversed" task is moredifficult slower frequencies of target movement were usedthan in task 1. We aimed to choose a level of difficulty inboth tasks that enabled subjects to achieve in the region of50% time-on target, since this level of performancemaximises the discriminating power of the measure.Recording oftracking response Two voltages indicating thehorizontal and vertical position of the joystick were fed intoa PDP- 11 computer via A/D converters and these controlledthe position of the screen pointer (cross). The computer alsocontrolled the position of the target and recorded theposition of target and joystick every 25 ms.Procedure Subjects were first shown how movement of thejoystick could control the position of the cross on the screen.

663Once they had understood this control system and hadfound a comfortable position for their arm, subjects per-formed the semi-predictable task (task 1). Subjects were nottold that target movements were predictable. Each subjectperformed the task continuously for three minutes. Theythen had a ten minute rest after which they again performedthe task continuously for three minutes. After completingthis task they were asked if they noticed anything regular orpredictable in the target movement.

Subjects then performed a reaction time task9 for about15 minutes. After this they performed the mirror reversedtracking task. They were told that the effect of moving thejoystick had been altered and were shown how the new con-trol system worked. Once they had understood this they per-formed two sessions of 3 minutes practice separated by a 10minute rest.

Analysis ofperformancePerformance in the horizontal and the vertical directionswas recorded separately. In each 30 seconds the total timefor which the subject was within the target limits wasrecorded. (Obviously, since the two directions were consid-ered independently, a subject would be considered on targetvertically if he was within the limits of the vertical directioneven if he was a long way from the target in the horizontaldirection.) The two occasions of practice were divided intosix 30 second trials. A rapid increase in performance at thevery beginning of practice is supposed to reflect the acquisi-tion of set. Given the 30 second measurement periods theearliest and shortest rise in the study is that between the firstand second 30 second period. We therefore chose thedifference between the first and second trials as our measureof "set". Additional within-session learning of the task wasindicated by the increase that occurred during trials 2 to 6.Overall performance on trials 2 to 6 was used to compare thefirst and second sessions of practice to obtain an estimate ofbetween-sessions improvements. Within and between groupcomparisons were made using repeated measures Analysis ofVariance, with appropriately adjusted degrees of freedom.

SubjectsTwelve patients with Parkinson's disease were tested. Therewere five women and seven men aged between 44 and 79years. They were all outpatients with a history of the diseaselasting at least 2 years. Details of the symptoms and thetreatment regime at the time of testing are shown in table 2.None of these patients showed any evidence of cognitiveimpairment.

Table 2 Clinical status of the Parkinsonian group

Subject Sex Age (yr) Duration Tremor Rigidity Bradykinesia Social Drugs

I F 45 3-5 L severe L arm L arm ind. Sinemet2 F 66 6 L + R slight R > L mod. L + R mod. dep. slight Artane Symmetrel3 F 71 6 nil L nil ind. Sinemet4 F 72 13 L + R mod L + R L + R mild ind. Sinemet Kemadrin5 F 73 3 L+ R L+ R L+ R ind. nil6 M 44 4 nil L marked L ind. Sinemet7 M 45 3 L fine L + R mod. L mild ind. Lorazepam8 M 53 2 L mild L + R L mod. ind. Sinemet9 M 57 6 L + R mild L + R mod. L + R mod. ind. Sinemet10 M 59 10 R mild L + R L + R mod ind. Sinemet Artane11 M 61 4 L + R L + R mod. R marked ind. Sinemet Disipal12 M 79 15 L + R marked L + R marked L + R marked dep. mod. Sinemet



j.com/

J Neurol N


ownloaded from


664

All the patients had been contacted through the local Par-kinson's Disease Society and had volunteered to take part inthe experiment. Nine were right handed and three were lefthanded. They used their preferred hand to perform thetracking task. One subject found the semi-predictable track-ing task (task 1) too difficult (less than 5% time on target)and had to be excluded. For the same reason four subjectshad to be excluded from the mirror reversed tracking task(task 2).The control group consisted of 13 volunteers, seven

women and six men, from the League of Friends and from

ancilliary, scientific and technical staff at Northwick ParkHospital. They were aged between 44 and 75 years. All buttwo controls were right handed and they used their preferredhand to perform the tracking tasks. No control subject hadto be excluded for poor performance.

Informed consent was obtained from every subject.

Results

Task 1: Semi-predictable tracking testFigures 1 and 2 show the results for the semi-

55 -

50 -

o 45-

0)2 40-

0

a, 35-E_

30 j

o ControlPD

A1)

LAaL)

C3

0

I ---I

0 1 2 3 0 1 2 3Time (min)

Fig 1 Mean time on targetfor the predictable (horizontal)component of task 1 (semi-predictable target movement).(PD = Parkinson's disease patients.)

55-

50-

;- 45-

O 40-

0 35-a,

E

30-

25 -

o Controlse PD

6 1 2 3 0 1 2 3Time (min)

Fig 2 Mean time on target for the unpredictable (vertical)component of task I (semi-predictable target movement).


predictable track; fig I showing the predictable hori-zontal component and fig 2 the unpredictable verticalcomponent.The overall impression is that the control group

perform better and learn faster than the Parkinson'sdisease group, particularly during the first two trialsof each occasion. ANOVA revealed a highlysignificant improvement in performance from trial 1to 2 (F (1, 20) = 19 8, p < 0-0001). However thisincrease was significantly greater in the control group(F (1, 20) = 8 36, p < 0 01). There was no significantimprovement in the Parkinson's disease patients(F (1, 9) = 0 88).

There was a highly significant improvement duringtrials 2 to 6 (linear F (1, 20) = 14 6, p < 0 001). Therewas also a significant interaction with direction suchthat improvement was greater for the horizontal, pre-dictable component than for the vertical unpredict-able component (linear F (1, 20) = 5-49, p < 0 05).There were no significant differences between thegroups in this aspect of learning. When analysed sep-arately the Parkinson's disease patients showed asignificant improvement in performance (linear F (1,9) = 6 51, p < 0-05). There was a very highlysignificant difference between the two sessions ofpractice (F (1, 20), = 35 1, p < 00001). Thisimprovement was the same for both directions, butwas significantly greater in the control group (F (1,20) = 5-98, p < 0 05). Nevertheless the Parkinson'sdisease patients did also show an improvement acrosssessions for the unpredictable, vertical component(F (1, 9) = 5 94, p < 0 05). No one in the controlgroup and only one member of the patient groupreported noticing anything predictable or regular inthe movement of the target.

Task 2: Mirror reversedFigures 3 and 4 show the results for the mirrorreversed track; fig 3 showing the mirror reversed hor-izontal component and fig 4 the non-reversed verticalcomponent. Both components were unpredictable.Once again the controls performed better and showedmuch faster learning. The improvement shown by thecontrols during the first minute of practice was evenmore striking than that shown for the semi-predictable task. There was a highly significantimprovement in performance from trials 1 to 2 (F (1,17) = 12 5, p < 0 01). This was significantly greaterin the controls (F (1, 17) = 8 03, p < 0 02). There wasno significant improvement in the Parkinson's diseasepatients (F (1, 6) = 0 30).

There was a small but significant improvement inperformance within sessions from trials 2 to 6 (linearF (1, 17) = 4 84, p < 0 05). However there was alsoa significant interaction between groups, directionsand trials (linear F (1, 17) = 5 18, p < 0 05). This was

L

0___A.



j.com/

J Neurol N


ownloaded from


Skill learning in Parkinson's disease55 -

50 -

; 45-

240 -~4-

0 35-E30 -

25 -

o Controlslo PDl

E0

o 1 2 3 o 1 2Time (min)

Fig 3 Mean time on targetfor the reversed (horizontal)component of task 2 (mirror tracking).

55 -

50 -

;, 45 -4-

v40 -

0 35-E30 -

25 -

o Controls9 PD l

I I I I--0 1 2 3 0

Time (min)Fig 4 Mean time on targetfor the normalcomponent of task 2 (mirror tracking).

because the control group showed aifor the horizontal, reversed componethe vertical component, while the Parngroup showed an improvement in neit(F (1, 6) = 0-62). Thus in this task tdisease patients showed no significantimprovements. There was a higiimprovement from one session to the= 56 8, p < 0-0001). This improvemclarger for the control group (F (1, 170 10). However the Parkinson's diseasea marked improvement in both comp(1, 6) = 9-42, p < 0-05).

Discussion

The performance of the control groupsimilar to that observed in other contii

tasks.6 In both sessions of work there was a markedimprovement during the first 30 seconds and ratherless thereafter. This effect was more marked on thesecond session of work than the first. Performanceafter the rest, except for the very first trial, wasmarkedly better than performance before the rest.The performance of the Parkinson's disease patientswas clearly worse than that of the controls, but alsoshowed qualitative differences. Nevertheless, like thecontrols the patients showed significant between ses-sions improvements for both tasks and, in the case ofthe mirror tracking task, this improvement was onlyslightly and insignificantly less than the controls.However the patients showed almost no within ses-

3 sion improvements. This lack of improvement wasmost striking for the first two trials of each practicesession during which time the patients showed nosignificant improvement on either task. For the 3minute session as a whole the patients showed no

-° improvement on the mirror tracking task, but a slightimprovement on the semi-predictable task.

Temporary and permanent learningIn the introduction we suggested that between ses-sions improvements reflect permanent learningwhereas within session improvements reflect tempo-rary learning. The results of the present experiment

._ _---- suggest that Parkinson's disease patients are relativelyunimpaired in the permanent component of skilllearning, but are markedly impaired in the temporarycomponent. The consequences of this defect are par-ticularly striking in the mirror tracking task (figs 3

,---r---- , and 4). On the first 30 second trial after the rest the2 3 performance of the controls was much worse than on

the last 5 pre-rest trials. Presumably this is because(vertical) the temporary learning acquired during the first ses-

sion of work has been lost during the rest. In strikingcontrast the first post-rest trial of the Parkinson's dis-

n improvement ease patients was much better than any pre-test trial.nt, but not for We suggest that since the patients did not acquire akinson's disease temporary component of learning during the pre-rest'her component work they did not have any to lose during the rest.the Parkinson's Thus only the increment in performance due to per-t within session manent learning is revealed.hly significant A different pattern is found for the semi-next (F (1, 17) predictable task. While the controls do show moreent was slightly improvement than the Parkinson's disease patients at= 3-16, p < the very beginning of practice, both groups showgroup showed within session improvements thereafter. After restonents also (F there is a slight decrement in performance for both

groups also. There is a major difference between thetwo tasks in terms of the availability of the knowledgenecessary for acquiring the skill. In the mirror track-ing tasks subjects were told, and indeed it would

was essentially become obvious to them at the very beginning ofnuous tracking practice, that the joystick control had been reversed

665

0



j.com/

J Neurol N


ownloaded from


666

so that if the target moved to the left they would haveto move to the right. Thus they could attempt to usethis knowledge immediately. In the semi-predictabletask the knowledge that the target moved regularly inthe horizontal direction was not given to the subjectsand few of them became aware of it during practice.The slow within session improvements shown by bothgroups might reflect the gradual, and unconscious,acquisition of this knowledge. This difference betweenthe tasks might imply that controls can make betteruse of explicitly given knowledge about the task thancan patients with Parkinson's disease. This hypothesishas been directly investigated by Bronstein andKennard"0 who studied the tracking of a predictabletarget by eye. When subjects were unaware of the pre-dictable nature of the target movement there was littledifference between Parkinson's disease patients andcontrols and both showed a steady improvement inperformance. However, when the information aboutpredictability was given the performance of the con-trols altered far more than that of the Parkinson'sdisease patients in that they produced many moreanticipatory responses.

The automatisation of learningWe have suggested that the permanent, between ses-sions improvements in performance reflect the auto-matisation of the skill being learned, particularly if itoccurs in the unpredictable components of the task.From this point of view our Parkinson's diseasepatients were acquiring a new skill that they couldperform automatically on both tasks. Crucial to thisargument is the claim that little or no learning is pos-sible for an unpredictable component of a task, sothat any improvement must be due to reduced atten-tion to other components. This claim was justified inFrith and Lang's study8 since no improvementoccurred on an unpredictable version of the task used.However, the tasks used in the present study weredeliberately made easier (by using slower target fre-quencies) to suit less able subjects. Clearly, althoughthe movements of the target were unpredictable in thelong term, short term predictions were possible par-ticularly if the target was moving relatively slowly.For short periods the target will be moving in a con-stant direction at a constant speed and the subjectcould learn something about the average speed of thetarget. In order to check this possibility we testedadditional controls (matched for age) on unpredict-able versions of the tasks with various average targetspeeds. In these tasks the target moved unpredictablyin both directions and control was not mirrorreversed. We found that there were small between ses-sions improvements for these tasks particularly forthose in which the average target speed was low.However the improvements were far less than those


shown by the patients and controls in the experi-mental tasks. For example in the mirror tracking taskcontrols and Parkinson's disease patients improvedby 18%. In a slow unpredictable task (which was eas-ier than in the mirror tracking task in terms of time ontarget) between sessions improvement was 10% andin fast unpredictable tasks (which was more difficultthan the mirror tracking task) improvement was only6%. We would conclude that there was evidence fromour data that patients with Parkinson's disease canlearn a new skill to a level at which it can be per-formed to some extent automatically.

The nature of the temporary improvement inperformance (Set)The major abnormality shown by the Parkinson's dis-ease patients on these tasks was a lack of anyimprovement in performance during the first 30 sec-onds of practice in both the first and the second ses-sion of work. We have argued above that thisimprovement reflects some temporary learning pro-cess that might be labelled "set". What is it then thatnormal people learn so rapidly at the beginning of apractice session, but then forget during the rest?One possibility is that "set" reflects the adaptation

of the subject to non-specific features of the task. Itwould be desirable that these aspects of performancebe acquired only temporarily since they may bemodified from one session to the next. For examplethe precise way in which limbs and muscles have to becontrolled to produce the required movement of thejoystick will depend on how the joystick is held, theposition of the elbow and so on. At the very beginningof practice the subject will have to adjust his motorprograms to allow for these variations. In some cir-cumstances these variations can be extreme. Forexample holding the steering wheel of a car in the 6o'clock position with one hand requires very differentmovements from the 10 to 2 position with two hands.Nevertheless a driver can adjust to these variationswith ease.A contrasting possibility is that the adjustment is in

relation to features specific to the task being learned.For example in a choice reaction time task therelations between stimulus and response are usuallyarbitrary, for example lifting the left index fingerwhen the red light goes on and the right finger whenthe green light goes on. To perform this task the sub-ject must temporarily "attach" his left finger liftingprogram to the appearance of a red light. Such tem-porary attachments are commonly required in real lifesituations. They can be temporary since the same taskmay only be performed once. If it is performedrepeatedly then special programs will develop andperformance becomes skilled. The same mechanismcould apply to the continuous tracking tasks used in



j.com/

J Neurol N


ownloaded from


Skill learning in Parkinson's disease

this experiment. In the mirror reversed task for exam-ple the subject must attach his program for movingthe joystick to the left to the stimulus of the targetmoving to the right. In the semi predictable task hemust attach his program for moving the stick to posi-tion X2 to the stimulus of the target being in positionXl. It seems plausible that these temporary attach-ments would be made at the very beginning of prac-tice, if the necessary knowledge were available andthat the rapid improvement of performance at thistime would reflect the success of these attachments.These two possibilities can easily be distinguished

since, if set reflects adaptation to non specific featuresof the task then it should occur even when there isnothing new in the task to be learned. We were able toexamine this in the new control data collected fromthe tasks in which the target movements were unpre-dictable and the joystick control was not mirrorreversed. The results were very clear cut. In such tasksthere was little or no increase at the very beginning ofa practice session. Thus the rapid post-rest upswingwe found in the experimental tasks does not reflectadaptation of the subject to non-specific features ofthe task. Thus the appearance of post-rest upswingseems to be associated with the novel features thathave to be learned in a task. We suggest that upswingreflects the success with which the subject can makenovel use of existing motor programs on the basis ofthe knowledge he has about the task requirements.This can be seen as a coping strategy to be used whilenew motor programs are developed that are specificto the novel task. The observation that upswingoccurs even for the component in the task for whichthere are no novel features to be learned suggests that,once the appropriate set has been developed it can bemaintained with little attention.On this account patients with Parkinson's disease,

while they can develop new motor programs for anovel task, have difficulty in deploying already exist-ing programs in novel situations.

Implicationsfor models ofParkinsonian deficit(1) Flowers' has suggested that Parkinson's diseasepatients have a specific difficulty with "open loop"control of movement. In other words they cannotanticipate the movement of a target, but must followit. We modify this proposal by suggesting that whileParkinson's disease patients cannot use theirknowledge of the target movement to produce antici-patory movements, they can eventually learn the skillof making anticipatory movements in a particulartask. Thus Parkinson's disease patients will be exces-sively reliant on feedback in the early stages of per-formance in a novel task. This account may explainwhy some studies do find evidence of open loop con-trol in Parkinson's disease patients3 5 while others do

667

not.l(2) Bloxham et al3 suggest that the problem resides

in the initiation of preprogrammed movements. Wewould now modify this formulation and suggest thatinitiation difficulties are particularly likely when themovement involves the deployment of a motor pro-gram in novel circumstances. This would be the casewhen the movement is elicited by a stimulus differentfrom the one for which it was developed.

(3) Wing and Miller4 suggest that the difficulty liesin using an internal trigger to initiate a prepro-grammed movement. This would be consistent withour account ifan internal trigger is involved whenevera program has to be initiated in a novel manner. Thatis, not only in situations where there is no specificexternal stimulus, but also when the program is to beinitiated by a stimulus other than the one for which itwas originally developed. In this latter case therewould presumably have to be some intermediate linkbetween the stimulus and the program, hence theinternal nature of the trigger.

(4) Marsden2 proposes that the problem lies in the"automatic execution of learned motor plans". Webelieve our results show that Parkinson's diseasepatients can perform skilled movements auto-matically. However Marsden (personal commu-nication) implies something more complex by his for-mulation, that is the ability to link together smoothlya number of different motor programs. This would beconsistent with our formulation if the sequencerequired was a novel one. However, we would expectthat patients would have little difficulty with a muchpractised sequence.

Some wider aspects ofsetWe have suggested that the development of motor setmay be defined as the modification of an existingmotor program for use in a new situation. In the fore-warned or cued reaction time task the advantageresulting from prior information occurs because thesubject can prepare the movement (that is, modify theprogram) in advance of the stimulus. Woodworth"called this preparation the development of muscularset. Thus the observation by Evarts et al'2 and Blox-ham et al3 that Parkinson's disease patients fail togain advantage from advance information in a choicereaction time task can also be explained in terms of afailure to develop an appropriate motor set. Denny-Brown and Yanagisawa'3 have suggested that thebasal ganglia are particularly concerned in activating"set" or "pump priming" for a certain act, that is thepreparation of the mechanism preparatory to a motorperformance orientated to the environment.

Set in the sense discussed above need not berestricted to motor behaviour, but could also apply tothinking and problem solving. The poor performance



j.com/

J Neurol N


ownloaded from


668

of Parkinson's disease patients on the Wisconsin cardsorting test14 could be seen as due to a problem withset in this sense. Flowers and Robertson'5 demon-strated poor performance in Parkinson's diseasepatients on a simplified form of the Wisconsin taskwhich they suggest reflects more directly the difficultythese patients have in maintaining a mental set. In thistask the subject has to continually modify the strategyhe uses for solving a problem by attending to differentdimensions of stimuli on alternate trials. Thus themental set required for this task is clearly similar tothe motor set we have suggested is involved in track-ing. Both kinds of set seem to be impaired in patientswith Parkinson's disease.

This study was conducted under the auspices andaccording to the rules of the Ethical Committee ofNorthwick Park Hospital. We are grateful to JulieKeenan for collecting the additional control data.

References

'Flowers K. Lack of prediction in the motor behaviour ofParkinsonism. Brain 1978;101:35-52.

2 Marsden CD. The mysterious motor function of the basalganglia: The Robert Wartenberg lecture. Neurology(NY) 1982;32: 14-39.

3Bloxham CA, Mindel TA, Frith CD. Initiation and exe-cution of predictable and unpredictable movements inParkinson's disease. Brain 1984;107:371-84.

'Wing A, Miller E. Basal ganglia lesions and psychological


analyses of voluntary movement control. CIBA Sym-posium 1984;107.

'Day BL, Dick JPR, Marsden CD. Patients with Parkin-son's disease can employ a predictive motor strategy. JNeurol Neurosurg Psychiatry 1984;47:1299-306.

6 Eysenck HJ, Frith CD. Reminiscence, Motivation and Per-sonality. New York: Plenum Press, 1977.

Fitts P. Factors in complex skill training. In: Glaser R, ed.Training Research and Education. University of Pitts-burgh Press, 1962.

5Frith CD, Lang R. Learning and reminiscence as a func-tion of target predictability in a two-dimensional track-ing task. Q J Exp Psychol 1979;31: 103-9.

Scott KN. The effects of advance information on initiationand execution of movements in Parkinson's disease.Unpublished research dissertation for the BPS Diplomain Clinical Psychology, 1983.

Bronstein AM, Kennard C. Predictive ocular motor con-trol in Parkinson's disease. Brain (in press).

Woodworth RS. Experimental Psychology. New York:Holt, 1938.

1Evarts EV, Teravainen H, Calne DB. Reaction time inParkinson's disease. Brain 1981;104:167-86.

"Denny-Brown D, Yanagisawa N. The role of the basalganglia in the initiation of movement. In: Yahr MD, ed.The Basal Ganglia. New York: Raven Press,1976:115-49.

14Bowen FP. Behavioural alterations in patients with basalganglia lesions. In: Yahr MD, ed. The Basal Ganglia.New York: Raven Press, 1976:169-80.

"Flowers KA, Robertson C. The effect of Parkinson's Dis-ease on the ability to maintain a mental set. J NeurolNeurosurg Psychiatry 1985;48:517-29.



j.com/

J Neurol N


ownloaded from


Documents

Impairments in the of manual skill in patients · skill and (b) achieve a degree ofautomaticity in the performance ofthat skill. Weinvestigated two kinds of learning using continuous