Feedback-based training of grip force control in patients with brain damage

653

Feedback-Based Training of Grip Force Control in Patients With Brain Damage Giinter Kriz, MA, Joachim Hermsdtrfer, PhD, Christian Marquardt, MA, Norbert Mai, PhD, AID ABSTRACT. Kriz G, HermsdOrfer J, Marquardt C, Mai N. Feedback-based training of grip force control in patients with brain damage. Arch Phys Med Rehabil 1995;76:653-9. • Objective: Feedback-based training of grip force control in patients with various brain lesions was evaluated. Design: Patients were instructed to hold a force transducer in a precision grip and to track with their grip force a moving target, which was presented together with the feedback signal on a monitor. Training performance was evaluated during a maximum of 10 sessions. Before and after the training, performance in two transfer tasks, which differed in target characteristics from the training task, was examined. Patients: Ten patients with impaired grip force control, after brain lesions of different origin, were selected on the basis of a clinical examination of hand function. Main Outcome Measures: Tracking accuracy in training tasks and transfer tasks was evaluated by calculating the conventional root-mean-square error. Results: Nine out of the 10 patients reduced their tracking error considerably during a maximum of 10 subsequent sessions (t test, p < 0.05), and most of them reached normal or near-normal performance. In addition, they improved in both transfer tasks (t test, p < 0.05). Detailed analysis showed that impaired initial performance and improvement was not uniform among patients and could be attributed to individual aspects of force control. Conclusions: In view of these results, a feedback-based training of grip force may be a useful enrichment of motor therapy. © 1995 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation

Precise control of grip forces is an essential prerequisite for skilled manipulation of grasped objects. Measurements of the finger forces that are applied to grasped objects have shown that grip forces are precisely adapted to the external requirements: During lifting of objects, grip force increases in parallel with the vertical lifting force that counteracts gravity; 1 additional loads applied to grasped objects induce simultaneous changes of grip force; l unpredictable loads that tend to pull objects away from the fingers are responded to with short latency grip force adjustments; 2 and inertial loads that emerge from acceleration of weighty objects are accompanied with corresponding changes in grip force) Lesions within the central nervous system are frequently accompanied with impairments of hand function and cause clumsiness and losses of dexterity when patients grasp and manipulate objects. Pursuit tracking tasks are frequently used methods to analyze disturbances of hand function in neurological patients. 4-8 Typically, movements or forces produced by the patient are displayed graphically on a monitor, and patients are instructed to match a target that may move in different fashions on the monitor. During pursuit tracking, the motor output has to be adjusted continuously to external requirements. Therefore, performance during tracking should reflect manual abilities like the adjustment of grip

forces to changing external requirements. Tracking experiments have been able to distinguish between normal and abnormal performance and have been employed to analyze disease-specific deficits of performance. 48 Despite the significance of tracking studies in neurological research, investigations on the effects of a training based on a tracking task in neurological patients are very rare. The main goal of this study was to determine whether patients who initially yield clear deteriorations in a grip force tracking task change their performance during frequent repetition of this task. For this reason, 10 patients with different brain lesions were trained in the control of grip forces produced in precision grip. Grip forces applied to rigid objects are not accompanied with finger movements, and therefore, are not open to direct visual control. In addition, the value of electromyographic recordings of single hand muscles as a feedback during precision handling may be limited because virtually all agonistic and antagonistic muscles are active during tasks like this 9 and EMG-activity patterns vary substantially between and within subjects.I° In order to display a patient's performance, the grip force was therefore measured directly using an adequate force sensor and the obtained signal was visualized on a feedback monitor. METHODS

From EKN Entwicklungsgruppe Klinische Neuropsychologie (Mr. Kriz, Dr. HermsdiSrfer, Mr. Marquardt), St~idtisches Krankenhaus Mfinchen-Bogenhansen, Miinchen, Germany; and Ludwig Maximilians-Universit~it (Dr. Mai), Neurologische Klinik, M~inchen, Germany. Submitted for publication August 15, 1994. Accepted in revised form February 14, 1995. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated. Reprint requests to G(inter Kriz, MA, EKN Entwicklungsgruppe Klinische Neuropsychologie, St~dtisches Krankenhaus Mtinchen-Bogenhausen, Dachauer Strasse 164, 80992 Miinchen, Germany. © 1995 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation 0003-9993/95/7607-319253.00/0

Subjects

Ten patients with motor disorders after brain lesions of different causes (table) and 17 control subjects (age 22 to 42 years; 6 women, 11 men) participated. The patients were selected on the basis of a clinical examination of hand function. All patients who showed an impairment in grip force control were included if they were able to hold the force transducer appropriately in a precision grip between thumb and index finger. Only the performance of the impaired hand (respectively the dominant hand in patients with bilateral Arch Phys Med Rehabil Vol 76, July 1995

654

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz Subjects

Code

Age

(Female/Male)

Etiology

(Yes/No)

(Yes/No)

Examined

Months Since Lesion

TRI5 TR31 TRI8

22 24 32

F F F

N N Y

N N Y

R L L

19 28 10

10 10 10

TR32

53

F

N

Y

L

5

6

TR35 TR17 TR12

22 23 38

M M M

Y N N

N N N

R R R

26 19 24

6 10 10

TR22

41

M

Y

Y

L

6

10

TR28 TR27

41 42

M M

Traumatic brain injury Viral encephalitis Intracerebral bleeding caused by ruptured aneurysm of ACA Infarction in the territory of the middle cerebral artery Traumatic brain injury Traumatic brain injury Infarction in the territory of the middle cerebral artery Intracerebral bleeding of the putamen-claustrum-type Cerebral abscess Intracerebral bleeding caused by vascular malformation

N N

Y Y

R R

4 6

6 10

Gender

defcits) was tested and trained in each patient (see table). All control subjects were right handed and were tested only on the right hand because a former study 7 demonstrated no hand differences in tasks very similar to that used here. Subjects were informed of the purpose and procedures of the study in advance and gave their informed consent.

Apparatus and Procedures The subjects were seated in front of a table facing a monitor. A light-weight force transducer was held between the pulps of thumb and index finger at a distance of 60mm between both fingers, all other fingers being flexed. During measurement, the forearm was resting comfortably on the table. All tasks required slow increases and decreases of grip force. For visual feedback, a vertical bar was displayed on the monitor. Its length corresponded to the actual force applied, and the subject was instructed to align the top of the vertical bar to a variable horizontal target line. The signal from the force transducer was amplified and fed into a personal computer by means of an A/D converter. Sampling frequency was 100Hz, and accuracy of force measurement was 0. IN (Newton). Details of the force measurement system used were described in Hermsdt~rfer and associates. H The following tasks were used: Training task. According to a trapezoidal target function, the training task " T r a p e z " required grip force increases and decreases with a constant rate of 2.5N/see separated by short periods of constant target force lasting 1.2 seconds. The lower target force level was 2.5N; the upper level was 7.5N. Trial duration was 20 seconds including three periods of grip force increase and decrease. Transfer tasks. In addition to the training, two transfer tasks were examined. The task " S i n u s " required the tracking of a sinusoidal target that moved continuously at a frequency of 0.2Hz. In the second task " R a n d o m " subjects were instructed to follow pseudorandom movements of the target. The target function was the product of three nonharmonic sinusoids with a maximum frequency component of 0.33Hz. Minimum and maximum target forces (2.5 and 7.5N) and trial duration (20 seconds) of both transfer tasks were the same as in the training task. Arch Phys Med Rehabil Vol 76, July 1995

Impaired Sensibility

Paresis

Hand

Number of Sessions

The patients were trained with the task " T r a p e z " over a maximum of 10 subsequent weekly sessions. In patients who reached normal performance after six sessions, the training was terminated. Sessions usually lasted 30 minutes, each consisting of 36 trials. Patients were allowed brief breaks every 12 trials. Before the first and after the last training session, patients' performance in the two transfer tasks "Sinus" and " R a n d o m " was recorded during 5 consecutive trials. The control subjects accomplished 5 trials of the tasks " S i n u s " and " R a n d o m " and 36 trials of the task " T r a p e z " in a single session. Patients and control subjects were familiarized with the measurement system before starting the measurements.

Data Analysis In the data analysis, stored time series of grip force were smoothed and the first derivative (force rate) was calculated digitally using kernel estimates. Kernel estimates provide nonparametric estimation of regression functions by moving weighted averages of a predefined number of data points. 12 Force traces and their derivates were displayed on a computer screen, and particular parameters were determined interactively. The force traces of the training task were segmented into the three different periods of force increase (F+), force decrease ( F - ) , and constant force. The first and second period of F - and the second and third period of F + were used for further evaluation. From the transfer tasks " S i n u s " and " R a n d o m , " the last 15 seconds were evaluated. Tracking accuracy was analyzed by calculating the root-mean-square error (RMS). 13 In addition, the peak force rate in each analyzed F + period was determined. Mean values and standard deviations of RMS and peak force rates were calculated for each session of 36 trials. Students t test was used for all statistical analyses.. The acceptable level for statistical significance was set at p < .05. RESULTS

Force Traces Figure 1 shows representative force traces and corresponding force rates from single trials. The control subject

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz

A

655

NORMAL CONTROL

F/v VF _]k~l~L, _.~L..

2.5N

~,~L._. ll 0 N/s 5s

B

PATIENTS FIRST SESSION

LAST SESSION

TR1 7

F

VF Fig 1--Grip forces during single trials of the tracking task "Trapez." Dotted lines indicate the target force, solid lines the actual

TR18

grip force (F) and grip force rate (VF), respectively; the horizontal lines in the force rate diagrams indicate zero force rate. (A) Typical single trial of a control subject. (B) Single trials of patient TR17 and TR18 in the middle of the first and last session of the training.

VF

J.a .,4

(fig 1A) was able to track the target with only small deviations of actual grip force from the target force. The corresponding force rate trace (VF) provides additional information about tracking performance of control subjects in this task. Perfect tracking of the target would have required to change the force with a constant rate of + / - 2.5N/sec in

the force increase and decrease periods interrupted by periods with a 0 force rate. However, the actual performance of the control subjects was characterized by frequent changes in force rate during each period, with amplitudes that fluctuated around the required force rate levels. In control subjects, no significant change in tracking performance during the 36 Arch Phys Med Rehabil Vol 76, July 1995

656

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz

trials of one session was found, when the mean tracking error in both periods (F+ and F - ) of the first and last 6 trials was compared (t test for pairs; p > 0.10). When tested within subjects, significant improvement was found only in three control subjects in F+ and only in one control subject in F - (t test for groups; p < 0.05). Thus, the range of performance of the control subjects was used as an indicator for normal performance. Force traces of two patients recorded during the first and last training session are displayed in figure lB. In contrast to the performance of the control subject, the patients' tracking patterns in the first session were characterized by high amplitudes indicating overshooting corrective actions. Patient TR17 produced abrupt force steps that frequently overshot the target force. The corresponding force rates exhibit extremely high amplitudes in the first session. Patient TR18 had pronounced difficulties in the periods of force increase but a nearly normal performance in force decrease. In the last session of the training, both patients' tracking performance differed drastically: the irregular velocity undulation pattern of patient TR17 had changed into a smoother pattern, both patients produced smoother force increases and decreases where force was much better fitted to the target, and the tracking patterns looked similar to that of control subjects. T r a c k i n g E r r o r (RMS)

Tracking performance of three patients across all sessions estimated with the tracking error RMS is illustrated in figure 2. Patient TR17 (fig 2A) reduced the mean tracking error as well as the performance variability (which is indicated by the standard deviation) profoundly during training. In the F+ periods, he improved continuously across all sessions; in the F - periods, the main improvement took place during the first sessions. At the end of the training, he almost reached the normal range in force increase and decrease. Figure 2B shows the performance of one patient (TR22) who did not improve during the training. In this patient the mean tracking error in F - periods even increased during the training. Figure 2C shows a patient (TR35) who started the training with a much lower error level than the two patients illustrated earlier. He reached the range of the control subjects already in the second session in force decrease and in the sixth session in force increase. Although high variability was evident among patients in the initial. RMS (fig 3), all patients, except one (patient TR22), improved during the training. Apart from this solitary exception, all patients reduced their tracking error, at least in force increase or decrease, and many of them reached the normal range at the end of the training. The highest reduction in RMS was found in patient TR17 who started the training with a high error level. However, the error level at training onset was no predictor for the extent of improvement. This is apparent in the two other patients with a high error level at training onset, TR22 did not improve and TR31 improved only slightly during the training, performing clearly out of the normal range at the end of training. Compared with the aforementioned three patients, the other seven patients started the training with a moderate level of RMS. In both periods (F+, F - ) , six of these patients reduced the tracking error. As a result of the error reduction, three of the six Arch Phys Med Rehabil Vo176, J u l y 1995

Force Increase (F+)

A z

Force Decrease (F-)

2.5 2.0 1.5

1.0 0.5

B z

2.5 2.0 1.5

¢¢

1.0

ttttttf tt \

t" ~

2.5

/

1.5 1.0 0.5

0.5

C 0.8 t z

0.6

,,-

0.4

~

~

0.8 0.6

\

0.4 0.2

0.2 i

i

i

i

,

i

i

i

,

1 2 3 4 5 6 7 8 9 1 0 SESSION

i

,

i

i

i

i

i

J

i

i

i

1 2 3 4 5 6 7 8 910 SESSION

Fig 2--Tracking accuracy of three patients during the training analyzed with the root-mean-square error (RMS). Periods of force increase (F+, left column) and force decrease ( F - , right column) were analyzed separately. Symbols represent the mean error of 36 trials tested in one session; vertical lines indicate corresponding standard deviations. Dotted areas display the range of performance of controls subjects (F+: RMS range 0.26N to 0.37N, overall mean 0.31N; F - : RMS range 0.34N to 0.52N, overall mean 0.42N). Training in patient TR35 was terminated after six sessions because he reached normal performance in both periods (F+ and F - ; see methods). Note the different scale in C.

patients reached the range of the control subjects in force increase; four of the six in force decrease. Three patients exhibited a normal tracking error in F - in the first training session. Although in most patients, error levels in the F+ and F periods coincided, there are dissociations in some patients. Patient TR18 had a high error in F+ and a very low error in F - at training onset; he improved clearly in the F+ periods. Patient TR15 had a higher error at training onset in F compared with F+ and improved only in F - . Force Rates

The analysis of the force rates (fig 4) during the periods of force increase (F+) demonstrated that all patients reduced the peak rates of the intermittent force adjustments during tracking the target (fig 1). The reduction was significant in all patients except TR22 and TR31. As a result, seven patients reached the normal range of peak force rates. On the average, patients who had high tracking errors in the first session showed high force rates, and lower rates were found in patients with lower tracking errors as indicated by the arrangement of patients on the abscissa in figure 4. Although the tracking error (RMS) was clearly out of the normal range

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz

Force increase (F+)

I TR22 TR17

0.8

Force decrease (F-) 2.0

2.0

1.25

l TR22 TR31 TR17 1.25

TR31

TR18

0.8

Z o~ n0.6

0.4 Fig 3 - - R M S of all patients in the first and last training session. Symbols represent the mean error of 36 trails tested in one session. Dotted areas display the range of performance of controls

subjects.

657

TR32 TR35 TR12 TR27 TR28

0.6

~,~

TR12 TR27 TR15 TR35 TR32 TR28

A

TR15

0.4

TR18 ii~iiiiiiiiii!!-2~iiiii;:_-i~i~iiii~i-~iiiii!!i~-=:-~iiiiiiiiii~ii~ -?'~iiii}-~-;}iii.~=---?iiii-£--?iii?--=--iiiiii----i? ================================================ I

I

I

I

First

Last

First

Last

Session

in all but one patient, five patients showed normal or nearnormal peak force rates at training onset. In order to compare the patients' reduction in RMS with the reduction in peak force rates, the ratio of the RMS in the first and last session was correlated with the corresponding ratio of peak force rates. The calculated correlation (r = .61) indicated an association between RMS reduction and peak force rate reduction (p < .05). Transfer

Figure 5A illustrates the performance of one patient (TR22) in the two transfer tasks before the training. The tracking performance in both tasks was clearly impaired corresponding to his performance in the training task at training onset, which was characterized by a high tracking error (fig 3). Figure 5B shows that all the patients who improved in the training task (fig 3) also reduced their tracking error in both transfer tasks. Seven of the patients reached the normal range in sinusoidal and two of them in random tracking. No improvement in the transfer tasks was observed only in patient TR22 who also did not reduce the RMS in the training task (see fig 3).

Session DISCUSSION

The effects of a simple training procedure involving isometric grip force tracking was tested in a group of neurological patients with impaired grip force control. The results are clear cut. Nine out of 10 patients improved in this task, and most of them reached a normal or nearly normal performance. The tracking tasks required precise control of low forces at a rather small rate of change. The slow ramp and hold pattern of the target in this training task induced a strategy that has been designated as position tracking, 14"~5characterized by frequent corrective movements. The analysis demonstrated that control subjects also showed a considerable tracking error, which was higher during force decrease ( F - ) compared with force increase (F+). Nevertheless statistical comparison of the first and last trials of the session yielded no tendency of the tracking error to decrease, indicating that the controls reached their limit of performance almost in the first trials. In comparison with control subjects, the initial performance of the patients was clearly impaired, it should be noted, however, that the patients showed substantial variability in the tracking error as well as in other aspects of their Arch Phys Med Rehabil Vol

76,July1995

658

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz

50

•

First session

]

Last session

One parameter that could have induced the decrease in tracking error is the force rate at which grip force adjustments were initiated. The finding that initial peak force rates decreased during the training in most patients provided evidence for this hypothesis. However, the percentage change in peak force rate accounted only for 37% of the variance (r = .61) in percentage change in tracking error. Inspection of individual data showed a tight coupling of tracking error and peak force rate reduction only in patients with very high initial force rates. Thus, a high tracking error may be attributed to excessive force rates in some patients, but not in others. It could be assumed that peak rate reduction significantly contributed to task improvement in the former group, whereas it was only a side effect in the latter. Remarkably, none of the patients showed peak force rates lower than the rates observed in controls (fig 4) despite mild to moderate paresis in some cases. In contrast to the training task, where dynamic (F+ and F - ) and static periods of grip force control alternated, the target moved continuously in the two transfer tasks. The signal was predictable in sinusoidal tracking as was the training task. In random tracking the signal moved unpredictably and consisted of distinct periods with very low and high force rates (fig 5) therein resembling the training task. Analysis demonstrated a strong linkage between training and transfer task: Relative error levels at training onset coincided especially between the tasks "Trapez" and "Random," all patients who improved during the training in the "Trapez" showed a clear transfer, and the reverse was true in the only patient who did not benefit from the training. In view of the differences between the tasks, one could therefore suppose that performance was based on similar sensorimotor capabilities. Thus, it was concluded that the training effect was not

4o

30 = o

_~ 2o

10 TR22 TR17 TR31 TR18 TR32 TR35 TR12 TR27 TR28 TR15

Patient Fig 4 - - P e a k force rates during force increase (F+) in the first and last session of the training (means and SD). Bars are arranged according to the patients' initial error in condition F + of the training task. The dotted area illustrates the range of control subjects (range 7.3N/s to 10.9N/s overall mean 8.5N/s).

performance. A particular dissociation was found between the tracking errors in the force increase (F+) and decrease ( F - ) periods. Some of the patients showed selective impairment in the F + periods. In other patients, the error during F - tended to be higher compared with F+ periods. Despite the different patterns of impairment, however, 9 out of 10 patients reduced their initial tracking error (during a maximum of 10 subsequent weekly sessions) regardless of the periods in which they were more impaired.

RANDOM

SINUS

A •,"

/

-

.,... " ".

;... ~

t

..-'. . \

2.5 N] ~

.,..

. .

"...

~

/

",

5e

5.5

5.5 -

z

a: ~

,.

2.5 N] 5s

B

~";,...,.,"

2.0

•

Before Training

[]

After Training

2.0-


1.5

g 1.0

1.0

~o 0.5

0.5

¢i:

i

i

i

i

i

i

i

i

i

R i ~ i i i T 22 TR17 TR31 TR18 TR32 TR35 TR12 TR27 TR28 TR15

TR22 TR17 TR31 TR18 TR32 TR35 TR12 TR27 TR28 TR15

Patient

Patient

Arch Phys Med Rehabil

Vol 76, July 1995

Fig 5 - - P e r f o r m a n c e of the patients in the transfer tasks. (A) Exemplary force traces of one patient (TR22) in the tasks "Rand o m " and "Sinus." Dotted lines indicate the target force, solid lines the patient's grip force. (B) RMS of both transfer tasks tested in five trials before and after the training (means and SD). Bars are arranged according to the patients' initial error in condition F + of the training task. The dotted areas represent the range of normal controls.

FEEDBACK-BASED GRIP FORCE TRAINING, Kriz

limited to the specific characteristics of the task that was trained but must have involved improvement in some more basic aspects of force control. Patients were selected from a larger group based in their impairments of hand function. From this criteria, a patient group with very heterogeneous etiologies resulted (table 1), and it is therefore not possible to relate training performance to lesion sites. The only patient who did not improve had a lesion mainly localized in the basal ganglia, which extended to the sensorimotor cortex. This single case performance may lead to some speculations that basal ganglia lesions may induce a negative prognosis for an improvement in this task. This hypothesis would be consistent with the observed association between basal ganglia lesions and impaired motor learning reported by other authors. 16'17 Alternatively the exaggerated pressure to achieve, which was particularly apparent in this patient, might have been a possible reason for the patient's failure in the training. Of particular note, improvement was not limited to patients with impaired sensibility of the hand (table 1). Although at a first glance, visual feedback of grip forces seems to be a well-suited training method in patients with losses of proprioceptive input, this study clearly showed that patients with lesions predominantly affecting the efferent motor pathways can benefit from the training also. In conclusion, the results show that precise registration and analysis of patients' performance in an obviously simple tracking task shows a variety of performance patterns that may be differently affected in different patients. However, regardless of the individual pattern of impairments, all but one patient succeeded in improving their tracking performance and transferring regained capabilities to other tasks. Control of grip force is a fundamental aspect of grasping and handling objects. If too little force is applied to an object, it may slip through the fingers, yet too much force may hamper skilled manipulation or even destroy a fragile object. The ability to control such forces is obviously difficult to test in real-life conditions. Patients report that objects fall out of their hands. However, this problem may be a rare event. The opposite, too much grip force, is more frequently observed in patients with damage to sensory motor areas. If this force is applied to rigid objects, the extent of force is

659

difficult to evaluate. Only direct measurement provides valid information about a particular capacity to control grip force. In view of these results, the feedback-based training of grip force may be a useful enrichment of motor therapy. The technique is easily applicable in a clinical setting and provides the possibility for evaluating performance continuously during the training. References

1. Johansson RS, Westling G. Roles of glabrous skin receptors and sensorimotor memory in automatic control of precision grip when lifting rougher or more slippery objects. Exper Brain Res 1984;56:550-64. 2. Johansson RS, Riso R, H~iger C, B~ickstrrm L. Somatosensory control of precision grip during unpredictable pulling loads I. Changes in load force amplitude. Exper Brain Res 1992; 89:181-91. 3. Flanagan JR, Wing AM. Modulation of grip force with load force during point-to-point ann movements. Exper Brain Res 1993; 95:13143. 4. Mortice BL, Becker WJ, Hoffer JA, Lee RG. Manual tracking performance in patients with cerebellar incoordination: effects of mechanical loading. J Can Sci Neurol 1990; 17:275-85. 5. Beppu H, Suda M, Tanaka R. Analysis of cerebellar motor disorders by visually-guided elbow tracking movement. Brain 1984; 107:787809. 6. Beppu H, Nagaoka M, Tanaka R. Analysis of cerebellar motor disorders by visually-guided elbow tracking movement. Brain 1987; 110:1-18. 7. Mai N, Bolsinger P, Avarello M, Diener HC, Dichgans J. Control of isometric finger force in patients with cerebellar disease. Brain 1988; 111:973-98. 8. Nagaoka M, Tanaka R. Contribution of kinesthesia on human visuomotor elbow tracking movements. Neurosci Letters 1981; 26:245-9. 9. Smith AM, Bourbonnais D. Neuronal activity in cerebellar cortex related to control of prehensile force. J Neurophysiol 1981;45:286-303. 10. Chao EYS, An KN, Cooney WP, Linscheid RL. Biomechanics of the hand. Singapore: World Scientific Publishing, 1989:147-62. 11. Hermsdrrfer J, Mai N, Marquardt C. Evaluation of precision grip using pneumatically controlled loads. J Neurosci Meth 1992;45:117-26. 12. Marquardt C, Mai N. A Computational procedure for movement analysis in handwriting. J Neurosci Meth 1994;52:39-45. 13. Poulton EC. Tracking skill and manual control. New York, London: Academic Press, 1974. 14. Day BL, Marsden CD. Two strategies for learning a visually guided motor task. Percep Motor skills 1982;55:1003-16. 15. Miall RC, Weir DJ, Stein JF. Visuomotor tracking with delayed visual feedback. Neuroscience 1985; 16:511-20. 16. Harrington DL, Haaland KY, Yeo RA, Marder E. Procedural memory in Parkinson's disease: impaired motor but not visuoperceptual learning. J Clin Exper Neuropsychol 1990; 12:323-39. 17. Soliveri P, Brown RG, Jahanshahi M, Marsden CD. Effect of practice on performance of a skilled motor task in patients with Parkinson's disease. J Neurol Neurosurg Psychiatry 1992;55:454-60.

Arch Phys Med Rehabil Vo! 76, July 1995