Journal of Neuroscience Methods 100 (2000) 3–12

A system for the study of visuomotor coordination duringreaching for moving targets

Thomas Schenk a,*, Jens Philipp a, Andreas Haußler a, Alexa Hauck b,Joachim Hermsdorfer c, Norbert Mai a

a Neurologische Forschung, Klinikum Großhadern, Ludwig-Maximilians-Uni6ersitat Munchen, Marchioninistrasse 23, 81366 Munich, Germanyb Lehrstuhl fur Prozessrechner, Technische Uni6ersitat Munchen, Munich, Germany

c Entwicklungsgruppe Klinische Neuropsychologie, Stadtisches Krankenhaus Munchen-Bogenhausen, Munich, Germany

Received 13 April 1999; received in revised form 5 April 2000; accepted 7 April 2000

Abstract

Prehensile behavior is a popular task in current research on human motor control. Most studies on reaching used stationarytarget objects and, therefore, most models do not address the challenges the motor system must respond to when reaching formoving objects. The machines used in earlier studies to produce object motion offered a limited range of trajectories and restrictedcontrol over various movement parameters. We have developed a device that allows a great variety of object trajectories alonga flat-table surface and gives the experimenter full control over all movement parameters. A linear positioning system is used tomove a sled beneath the table surface. Magnetic coupling transfers the sled’s movement to the target object on the tabletop. Thisarrangement allows fast movements of the object (up to 5 m/s) and at the same time protects subjects from any harm due to themoving parts. The system is connected to LC shutter glasses, a 3-D movement registration device, and a switch that detects theonset of hand motion. This allows the selective withdrawal of vision during the reaching task or the introduction of changes inthe object motion depending on the subject’s reactions. © 2000 Elsevier Science B.V. All rights reserved.

Keywords: Prehensile movements; Moving objects; Motor control; Motor learning; Reaching; Grasping

www.elsevier.com/locate/jneumeth

1. Introduction

Prehension is a popular task for studying primatesensorimotor control. It consists of two parts: bringingthe hand to the object (reaching) and grasping theobject. Reaching and grasping involve different neuralstructures and have different functional requirements(Jeannerod, 1996). Nevertheless, the two prehensilecomponents must be well-coordinated. A great numberof experimental studies have been conducted to explorethe processes and neural mechanisms that allow coordi-nation of the prehensile act (for reviews see Jeannerod(1996) and Smeets and Brenner (1999)). With the excep-tion of a few studies (e.g. Carnahan and McFadyen(1996), von Hofsten et al. (1998), Zaal et al. (1998)),most have examined prehension using stationary ob-

jects. Therefore, most models of the prehensile act donot address the additional challenges (e.g. temporalconstraints, uncertainty with respect to grasping posi-tion) that arise with moving target objects (Arbib, 1981;Hoff and Arbib, 1993; Smeets and Brenner, 1999; seehowever: Zaal et al. (1998)). However, even infants areable to reach for moving objects (von Hofsten et al.,1998) and, therefore, a model that cannot explain thisability is clearly unsatisfactory.

Another aspect of prehension that has received lessattention than it deserves is the finding that someneurological patients are much better when reaching formoving than for stationary objects (Majsak et al., 1998;Schenk and Mai, 1999). This finding suggests that thetwo tasks involve different motor pathways (Glicksteinand Stein, 1991). Exploring such aspects of objectmotion which lead to performance benefits will notonly further our understanding of the neural control ofmovements. It may offer new options for the treatmentof such movement disorders.

* Corresponding author. Tel.: +49-89-70954828; fax: +49-89-70954801.

E-mail address: tschenk@nefo.med.uni-muenchen.de (T. Schenk).

0165-0270/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.PII: S 0 165 -0270 (00 )00222 -3

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–124

The study of prehensile movements toward movingobjects might also allow new insights on visuomotorlearning, particularly if unnatural motion trajectoriesare used which the subjects have not yet encountered.

The study of reaching movements toward movingobjects requires a machine that moves the target objectin a controlled and reproducible way. Ball-projectionmachines have been used in many studies on one-handed catching (see review by Savelsbergh and Whit-ing (1996)). They have the advantage of producing the3-D movement trajectories that are typical for manyball games. However, to study the issues mentionedabove ball-projection machines have a number of dis-advantages. First, the reproducibility of the trajectoriesare poor. Second, the experimenter can set only onecontrol parameter (i.e. initial impulse), and it deter-mines all features of the trajectory. However, if we areinterested in identifying the parameters of the objectmotion which determine the timing of the reach andgrasp components of the prehensile movements it isdesirable to have independent control of more than onecontrol parameter, for example, of acceleration andpeak velocity. Moreover, to study the time withinwhich visual information can be incorporated into themotor program it would be useful to introduce suddenalterations of the object motion trajectory at differenttime intervals. Third, the ball, which is sometimesmoved at considerable speed toward the subject couldhurt a handicapped patient. For these reasons ball-pro-jection machines should not be employed in experi-ments with patients.

von Hofsten et al. (1998) used a computer-controlledplane plotter and replaced the pen with a small magnet.The plotting area was covered with a sheet of alu-minum on top of which a second magnet held the targetobject. A similar system was also used by Zaal et al.(1998). These devices produce reproducible object tra-jectories and can be safely used with infants and pa-tients, but it is difficult to produce non-lineartrajectories and to control the dynamic parameters ofthe movement. Moreover, the fastest object speed is 0.6m/s, which is just about half of the maximal speed atwhich healthy adult subjects can reliably reach formoving objects (see Schenk and Mai (1999)). To iden-tify patients whose reaching for a moving target isimpaired and, therefore, localize the neural structuresinvolved in this task we must be able to test the fullperformance range of the subjects.

In this paper we describe a new system that producesobject motion. Like von Hofsten, we also use pairs ofmotors and linear axes that are covered by a metalplate and employ a pair of magnets to transfer themotion of the axes to the top of the plate. However,our system differs in some important respects. First,our system can produce a multitude of different 2-Dmotion paths not just linear ones. Second, the system

offers full control over all dynamic parameters, thus agreat variety of different velocity profiles can be pro-grammed. Third, at any point along the motion trajec-tory a change of parameters can be triggered by a signalwhich, for example, might come from a switch thatindicates the onset of hand motion. Fourth, the systemcan produce object motion with a peak velocity of upto 5.6 m/s.

Our system is coupled to LC shutter glasses and adevice for 3-D movement registration. It complies withhigh safety standards to allow the study of visuomotorcoordination in healthy subjects and patients with vi-sual or motor deficits.

2. Description of the system

2.1. System o6er6iew

The setup integrates several devices under the controlof a Pentium personal computer (PC1). Fig. 1 presentsan overview of the system:1. Servo object controller (SOC). This machine is used

to move an object within the horizonal plane and isdescribed in Section 2.2.

2. LC shutter glasses (Plato system, Translucent Tech-nologies). The LC shutter glasses can be switchedbetween an opaque (no vision) and a translucentstate (full vision) within 3–5 ms. They are used todetermine when vision is available to the subjectduring the trial.

3. Hand switch. An electronic switch is fixed on thebottom of a metal plate (8×8 cm). At the beginningof the trial the hand rests on this plate. As soon asthe hand moves the switch is released and a signal issent to the computer (PC1).

4. 3-D movement registration. The device for the 3-Dregistration of hand and arm movements (CMS70,Zebris) uses ultrasonic signals emitted by activemarkers placed on relevant points of hand and armto determine the 3-D position of these points (spa-tial accuracy: 0.1 mm; maximum sampling rate: 200Hz). The software (WinData 2.11, Zebris) to controlthe recording and the storage of 3-D position dataruns on a second Pentium personal computer (PC2).

5. User interface (GET–US). The user interface devel-oped to run experiments with the integrated setup isdescribed in Section 2.3.

Control of the integrated setup uses three modules.The first module is the user interface (GET–US), whichoperates within a Microsoft Windows environment, hasa graphical interface, and allows interactive control ofthe whole setup (the integrated system, IS). All pro-cesses requiring real time data handling are under thecontrol of two plug-in cards, which have their owndedicated processors and represent modules 2 and 3 of

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–12 5

the control system. The real time system controlling themotors of the servo object controller (SOC) consists ofa multiprocessor-based system on a PC card for theISA bus (AT6250, Parker, Compumotor) and comeswith custom software (system 6000), that provides high-level commands for the specification of position, veloc-ity, and acceleration parameters. A second plug-in card(timer card) handles the information flow to and fromthe other devices (LC shutter glasses, CMS70, handbutton). This timer card is based on a PC ISA bus card(Phytec) with a Siemens 80537A microcontroller (64kbyte RAM). The card was customized and its outputchannels optically isolated. The program for the micro-controller is written in assembler and communicateswith the user program via shared RAM on the timercard. This arrangement permits deterministic time con-trol and measurements within 1 ms. The timer card hasoutput connections to the AT6250 (to start the execu-tion of programs for the SOC), the LC shutter glasses,the CMS70 system (start/stop of 3-D movement regis-tration and information on the present status of thevarious devices) and input connections from the handswitch. Fig. 1 shows a schematic presentation of theconnections within the entire setup.

To prepare experiments with this setup trials have tobe defined within the user-interface (GET–US). Foreach trial the experimenter must determine which tra-jectory will be used for the object motion and whetherthe state of the LC shutter glasses will change duringthe course of the trial. When the experiment is beingrun and the trials are being executed those aspects ofthe specification concerning the control of the objectmovement are being translated into the system 6000code and sent to the AT6250 card where the program isstored until execution. All other aspects of the trialspecification are sent to the timer card and translatedinto assembler code for the Siemens 80537A microcon-troller. The timer card executes the commands for theshutter glasses, triggers the start/stop of 3-D movementregistration and initiates the program for the objectmotion. In addition, the status of all devices (see Table1) is read 2000 times per second and transmitted to theuser-interface, where the actual status of each device isvisualized. The same information is also sent to PC2,where it is translated into a binary code and stored inthe file for the data coming from the 3-D movementregistration system. In the off-line analysis of the 3-Dmovement recordings specific aspects of the hand and

Fig. 1. Connections of the integrated system.

Table 1Status information, sources, and targets

Device TargetsOn-status Off-status

Timer card; GET–USRecording is offHand movement is being recordedCMS70Object is movingSOC No object motion Timer card; GET–US, CMS70Translucent (full vision) Timer card; GET–US, CMS70LC glasses: left eye Opaque (no vision)

Timer card; GET–US, CMS70Opaque (no vision)LC glasses: right eye Translucent (full vision)Hand is on the switchHand switch Hand is off the switch Timer card; GET–US, CMS70

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–126

Fig. 2. Mechanical construction of the servo object controller (SOC).

arm movement can thus be easily related to events inthe course of the trial (e.g. start of the object motion orthe closing of the LC shutter glasses). Information onthe status of the hand switch can also be used by thetimer card to open or close the LC shutter glasses orstart another program for the SOC. This circuit is,therefore, ideal for realizing perturbation experiments.

2.2. The ser6o object controller (SOC)

2.2.1. ConstructionSOC is a computer-controlled positioning system

that can be used to move a target object along a varietyof trajectories. The positioning system is depicted inFig. 2. It consists of three linear guidance systems (Y1,Y2, X), two motors and a servo control system. Cog-belts within the linear guidance system (ELZ, Lineart-echnik-Korb) transfer the motion of the motor to thatof a sled, which runs on top of the linear axes. Y1 andY2 are mounted in parallel and connected by a shaft;thus, one servo motor drives both axes. The third linearguidance system (X) is mounted on top of this con-struction and fixed to the sleds of Y1 and Y2. The sledon X (SX) runs along an axis orthogonal to Y1 and Y2and is driven by a separate motor. By combining themovement of Y1/Y2 and X, it is possible to positionSX at any point within a horizontal area (1×1 m). Themotors are brushless servo motors (Kollmorgen H3480802), each of which can provide 925 W of continuouspower and achieve a peak torque of 10.5 Nm. Themotors’ power lines are connected to servo amplifiers(Kollmorgen ROL 20012) which use a closed-loop con-

trol system based on Hall sensors to achieve a giventorque on the shaft of its motor. The servo controlsystem computes the required target torque value toachieve a given position measured by encodersmounted on the shaft. Complex motion control appli-cations can be implemented with the 6000 series com-mand language (Compumotor).

The positioning system is mounted on a table whichis built from aluminum frames. The entire constructionis enclosed by wooden panels and covered by a steelplate (diameter: 1.5 mm) to prevent contact between themoving parts and the subjects (Fig. 3). A magneticcoupling transfers the motion of the sled to the targetobject on top of the steel plate (see Fig. 4). One magnet

Fig. 3. The fully assembled system (SOC).

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–12 7

Fig. 4. Construction of the magnetic coupling.

sits on the SX, the other is embedded in a circularplastic disk (weight: 144 g), which moves along the topof the steel plate and carries the target object (theobject carrier). Teflon sheets between the magnets andthe steel plate reduce friction. Both magnets are circularand of the same size and material, but the arrangementof their poles is complementary. This setup provides astrong, restricted, and self-centered magnetic field thatpermits object acceleration up to 20 m/s2. A peakvelocity of 5.29 m/s can thus be achieved on a diagonalpath crossing the usable area of the steel plate. Forcesabove 3 N will disconnect the magnet on the sled andthe object carrier; thus the risk of injury is negligible.

2.2.2. PerformanceTo assess the accuracy and reliability of SOC its

performance was measured using the 3-D movementregistration system (CMS70) with one marker on theobject carrier (sampling rate: 200 Hz). A diagonal path(length 0.9 m) and trapezoid velocity profiles werechosen. Thus, each velocity profile was composed ofthree sections: a constant acceleration phase (acc), aconstant velocity phase (con), and a constant decelera-tion phase (dec). Slow (acceleration/deceleration=1.25m/s2, peak velocity=0.86 m/s), moderate (acceleration/deceleration=5.0 m/s2, peak velocity=1.73 m/s), andfast (acceleration/deceleration=15.0 m/s2, peak veloc-ity=3.0 m/s) trajectories were used. Each type of tra-jectory was repeated ten times. The mean and S.D. forthe average acceleration/deceleration value in acc and

dec and the average velocity in con were calculated. Tocompute the relative error the difference between targetvalue and the average value was set in relation to thetarget value and expressed in percent. The results areshown in the top part of Table 2. The S.D.s were verylow, indicating the high reliability of the system. Therelative error was low for the constant velocity phase(B0.3%) but clearly higher for the acceleration anddeceleration phases (up to 8%). This error increasedwith higher acceleration values. However, the very lowS.D.s show that these errors are very similar acrossreplications and suggest that they may be compensatedfor by changing the target value by an appropriateamount.

In addition, absolute errors for the time course ofposition and velocity were calculated. The maximumand the mean of these errors along the three differentsections of the trajectory were computed and averagedacross ten replications of each type of trajectory. Ascan be seen in the second and third parts of Table 2 andin Fig. 5 errors in the time course of position were small(maximum: 0.04 m). It should be noted that the posi-tional errors occurred along the programmed path.Deviations from the target path were not detected.Errors in the time course of velocity were greater forfast trajectories. Again, the very low S.D.s for the timecourse of position and velocity indicate that despite theoccurrence of errors the trajectories are reproducible. Insummary, small errors occur with respect to the averagevelocity and acceleration and with respect to the time

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–128

course of position and velocity. No errors occur withrespect to the given path. All errors are, however,virtually constant. Thus, they might either be compen-sated for by using different target values or be consid-ered in the assessment of the corresponding reachingperformance.

2.3. The software

The software program is written in Borland Delphiand is devised as a compiler, which reads a file contain-ing a list of commands whose execution can then beinteractively controlled. The command language has

Table 2Performance of SOC

v-Profil 3v-Profil 1 v-Profil 2

Dec Acc Con Dec Acc ConAcc DecCon

Acceleration/6elocity a6eraged o6er time−5.0000a1.7300b5.0000a−1.2500a −15.000a0.8600b1.2500aTarget 3.0000b15.000a

−1.1724a 5.1183a 1.7340b −4.7952aMean 16.217a1.2971a 3.0200b −14.620a0.8609b

(0.1090)(S.D.) (0.0230) (0.0002) (0.0302) (0.0244)(0.0006)(0.0598)(0.0110)(0.0011)2.4492 0.2317 4.0958 8.1150 0.6670 2.52806.21043.7657Rel. error (%) 0.1063

Time course of position0.0387 0.0232 0.0140Max error (m) 0.01290.0403 0.0151 0.0202 0.02090.0402

(0.0035) (0.0002) (0.0003) (0.0003)(S.D.) (0.0003)(0.0035) (0.0002) (0.0004)(0.0036)0.01120.01820.0072Mean error (m) 0.0197 0.00890.0398 0.0192 0.0108 0.0132

(0.0036)(S.D.) (0.0005)(0.0018) (0.0001) (0.0003) (0.0004)(0.0004)(0.0003)(0.0013)

Time course of 6elocity0.0534 0.1361 0.0440Max error (m/s) 0.05710.0775 0.3236 0.1880 0.27420.0369

(0.0057) (0.0039) (0.0041) (0.0008) (0.0003) (0.0003) (0.0112) (0.0017) (0.0018)(S.D.)0.0403 0.0798 0.0255 0.0249Mean error (m/s) 0.14990.0593 0.0784 0.12420.0080

(0.0057) (0.0009)(S.D.) (0.0022)(0.0004)(0.0003)(0.0002)(0.0059)(0.0002) (0.0001)

a Values are in m/s2.b Values are in m/s.

Fig. 5. Performance of SOC. Time course of position and velocity for three different velocity profiles. The parameters for the three velocity profilesare presented in Table 2. The three upper graphs show the time course of the distance covered during the movement. The three lower graphs depictthe velocity course. The continuous line represents the target course; the broken line gives the actual movement of the SOC which the CMS70system measured.

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–12 9

been hierarchically structured. The experimental ses-sion, itself defined by a sequence of trials, correspondsto the upper layer of this hierarchy. The next layer isthen represented by the trial, which is defined by acombination of three parameter sections running underthe headings ‘geometry’, ‘velocity profile’, and ‘periph-ery’ and constituting the last layer. The path of theobject motion is defined in the section geometry. Theuser can choose between a linear path, a circular path,or any combination of the two and must then specifyall relevant parameters (i.e. starting point, end point,and radius). In the section velocity profile the course ofthe velocity along the path, defined in the correspond-ing geometry section, is fixed. Since all trajectories startand end with velocity=0, each velocity profile will startwith an acceleration phase and end with a decelerationphase; in between those any combination of furtheracceleration, deceleration, or constant-velocity phasesmay occur. The command language offers a variety ofpossibilities for specifying each phase of the velocityprofile. For example, one can fix the value of accelera-tion and the duration of this phase; the peak velocityand the travelled distance are then computed by thesystem. Conversely, one can specify the accelerationand the velocity, and the system computes the durationof the acceleration phase and the distance travelledduring that time. Similarly, a phase of constant velocitycan be specified with respect to its length or its dura-tion. Thus, a multitude of different profiles can bedefined, the only restriction being that the temporalintegration of the velocity profile (i.e. the distancecovered) corresponds to the length of those paths withwhich the velocity profile will be combined.

The temporal sequence of events is determined in thesection periphery. Such events are the start and stop ofthe 3-D movement registration, the opening and closingof the LC shutter glasses, and the start of the objectmotion as defined in the geometry and velocity profilesections. The timing of these events may be specifiedwith respect to the beginning of the trial or with respectto a change in the status of the hand switch. It is thuspossible, for example, to determine that the glasses willbe closed at the very moment the hand starts to move.Each path (section: geometry), velocity profile (section:velocity profile), and event sequence (section: periphery)is assigned a name by which it can be referred to in thedefinition of a trial (section: trials). Each new combina-tion of a path, a velocity profile, and an event sequencedefines a new type of trial that is also given a uniquename. The trials are listed by their names at the end ofthe program (section: session) in the order in whichthey are to be executed. However, this order may bechanged interactively. The whole program is writtenwith a text editor and stored in a file (called: CTR-File).An example of such a program is presented in Fig. 6.

Prior to experimentation, the CTR-File is read by theinterpreter, a syntax check is performed, and kind andlocation of errors are displayed. After correction of allerrors the program can be executed trial by trial. Alltrials are listed in their predefined order in one windowof the graphical interface. By default, the interpreterwill step through the trials in the order in which theyare listed; however, this order can be changed by theuser. The trial to be executed next is highlighted, and itsparameters are shown in a second window. In addition,the path and the velocity profile are displayed (Fig. 7).To execute the trial, the user must click a GO button.Prior to and during the trial, the status of all relevantdevices (3-D movement registration, shutter glasses,SOC, hand switch) is displayed graphically. This formof control, which combines prestored programs with aninteractive mode of program execution, offers the possi-bility of standardized experiments that can be adaptedto individual subjects and different experimentalsituations.

3. Illustration of the application

The possibilities of the system are illustrated by anexperiment in which a subject is expected to reach-and-grasp with her right hand for an object that is movingaway. For this movement, a trapezoid velocity profilewith one acceleration (a=5 m/s2), one constant veloc-ity (6=1.0 m/s), and one deceleration (a= −5 m/s2)phase was used. In most trials (standard trials) theobject moved along a straight path within the sagittalplane of the subject. However, in some of the trials(perturbation trials) a leftward movement was superim-posed, which started 100 ms after the subject lifted herhand from the hand switch and started her arm move-ment. This leftward motion began with an accelerationphase (10 m/s2); after 250 ms the peak velocity of 2.5m/s was reached. After a further 120 ms the decelera-tion phase (−10 m/s2) was entered. Subjects did notknow beforehand which of the two kinds of trialswould be presented. The arm movements were recordedwith the CMS70 using one marker on the wrist. Toregister the movement of the object another marker wasfixed to the object. The 3-D position signals weresampled with a frequency of 100 Hz. Fig. 8B showsmovement traces for a perturbation trial. In the acceler-ation course of the hand a pronounced negative peak(indicating a leftward turn) could be seen 120 ms afterthe perturbation occurred. Such a pronounced peakwas not observed in the standard trial (Fig. 8A) and,thus, can be interpreted as an early response to theperturbation, which allowed the subject to catch upwith the object and grasp it before the object was out ofher reach. The results show that fast reaching move-

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–1210

Fig. 6. An example of a program used to define an experimental session. The program defines two types of trials (diagonal–simple,diagonal–complex) both of which use the same path (geometry diagonal) and the same sequence of events (periphery standard) but two differentvelocity profiles (simple and complex). The path and velocity profiles used in the two trials are graphically presented in Fig. 7. The commandsare in bold typeface. Comments are separated by semi-colons and in a plain typeface.

ments can be modified on the fly. This confirms thatreaching movements toward moving objects are notunder exclusive feedforward control and illustrates theefficacy of the visuomotor interface which is able toproduce fast compensatory responses. This paradigmcould also be used to study the effect of perturbationon the coupling of hand opening and arm transport andthe role of visual feedback in the programming of anadequate compensatory response. Another example for

the application of the system is presented in Schenk andMai (1999).

4. Conclusions

The new machine that we developed to study reach-ing for moving objects in normal subjects and neuro-logical patients not only meets high safety standards

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–12 11

Fig. 7. Illustration of the trials defined in the sample program (cf. Fig. 6). Path (A) and velocity profile (B) for the trial diagonal–simple; path(C) and velocity profile (D) for the trial diagonal–complex. Open circles indicate a change in a parameter of the velocity profile.

Fig. 8. Illustration of the application. The upper left graph presents the path of the object (continuous line) and the hand (broken line). The lowerleft graph shows the time course for the horizontal movement of the object (continuous line) and hand (broken line), and the horizontalacceleration of the hand (dotted line). (A) Typical results of a standard trial. There is no horizontal movement of the object, and only a slightdeviation from the horizontal zero line in the hand movement. A small negative peak in the acceleration profile indicates the return of the handtowards the object’s path. (B) Typical results of a perturbation trial; 100 ms after the hand rises from the hand switch (H) the object deviates tothe left (P). In the acceleration profile of the hand a negative peak (indicating a leftward turn) can be seen 120 ms after the perturbation occurred(R). This peak is twice as high as the corresponding peak in the standard trial and can be interpreted as the subject’s response to the perturbationin the motion path of the target object.

T. Schenk et al. / Journal of Neuroscience Methods 100 (2000) 3–1212

but allows an unusual flexibility in the choice of kindsof object motion. Moreover, it incorporates devices thatpermit the manipulation of visibility, the programmingof perturbation experiments, and the 3-D recording ofhand and arm movements. Measurements of its perfor-mance confirm that the machine produces differentkinds of object motion with a high degree of accuracyand reliability. The application of the system was illus-trated in an experiment that shows just one of manypossible ways to use the system for studying relevantaspects of motor control in prehensile behavior.

Acknowledgements

This research was supported by the DeutscheForschungsgemeinschaft, SFB 462 (‘Sensomotorik’).

References

Arbib MA. Perceptual structures and distributed motor control. In:Brooks VB, editor. Handbook of Physiology: Section 1. TheNervous System: Motor Control, vol. 2. Bethesda, MD: Ameri-can Physiological Society, 1981:1449–80.

Carnahan H, McFadyen BJ. Visuomotor control when reachingtoward and grasping moving targets. Acta Psychol (Amst)1996;92:17–32.

Glickstein M, Stein J. Paradoxical movement in Parkinson’s dis-ease. Trends Neurosci 1991;14:480–2.

Hoff B, Arbib MA. Models of trajectory formation and temporalinteraction of reach and grasp. J Mot Behav 1993;25:175–92.

von Hofsten C, Vishton P, Spelke ES, Feng Q, Rosander K. Pre-dictive action in infancy: tracking and reaching for moving ob-jects. Cognition 1998;67:255–85.

Jeannerod M. Reaching and grasping. Parallel specification ofvisuomotor channels. In: Heuer N, Keele SW (Eds.), Handbookof Perception and Action. London: Academic Press, 1996:405–60.

Majsak MJ, Kaminski T, Gentile AM, Flanagan JR. The reachingmovements of patients with Parkinson’s disease under self-deter-mined maximal speed and visually cued conditions. Brain1998;121:755–66.

Savelsbergh GJP, Whiting HTA. Catching: a motor learning anddevelopmental perspective. In: Heuer H, Keele SW, editors.Motor Skills. London: Academic Press, 1996:461–502.

Schenk T, Mai N. Time constraints improve reaching movementsin an ataxic patient. Exp Brain Res 1999;128:214–8.

Smeets J, Brenner E. A new view on grasping. Mot Control1999;3:237–71.

Zaal FT, Bootsma RJ, van Wieringen PC. Coordination in prehen-sion. Information-based coupling of reaching and grasping. ExpBrain Res 1998;119:427–35.

.