A test between two hypotheses and a possible third way for the control of prehension

Abstract We used an obstacle avoidance task to test two opposing accounts of how the nervous system controls prehension. The visuomotor account supposes that the system independently controls the grip formation and transport phase of prehensile
of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Abstract We used an obstacle avoidance task to test twoopposing accounts of how the nervous system controlsprehension. The visuomotor account supposes that thesystem independently controls the grip formation andtransport phase of prehensile movements. In contrast, thedigit channel hypothesis suggests that the system con-trols the thumb and finger more or less independently.Our data strongly favoured the traditional visuomotorchannel hypothesis and demonstrated that the time takento grasp an object in the presence of obstacles was wellpredicted by a Fitts' law relationship. We suggest a “third-way”hypothesis in order to retain the advantages of thedigit channel hypothesis within the visuomotor framework.The third-way hypothesis suggests that the nervoussystem selects a single digit to transport to the object.We speculate that the actual digit selected might dependupon attention and the nature of the prehension task. Thishypothesis is able to account for most of the empiricalfindings unearthed by researchers investigating the con-trol of prehension. Key words Prehension · Precision grip · Motorprogramming · Visual cues · Obstacle avoidance · Human Introduction We may be within a megaparsec (where a megaparsec is3.1 × 10 24 cm) of calculating the distance to neighbouringgalaxies (Paczynski 1999), but we are a long way fromunderstanding basic human behaviour. Imagine reachingout to pick up a wine glass – an everyday task for a ma- jority of Europeans. The task depends upon the nervoussystem accurately gauging the distance of the wine glassfrom the body, judging the properties of the glass (e.g.size, weight) and generating the appropriate motor com-mands in response to this information. We are still manymegaparsecs from understanding how the system solvesthese various problems. Our lack of understanding canbe illustrated by the controversy that exists with regardto just one aspect of that task – namely, what commandsdoes the nervous system send to the hand to ensure thatthe digits travel the correct distance and open wideenough to grasp the wine glass? Two different hypothe-ses have been proposed to account for the nature of thecommands. The first (traditional) hypothesis postulatesthat the visuomotor transformations related to reachingand grasping are controlled independently (i.e. a separatetransport and grip formation component exist). This hy-pothesis is known as the visuomotor channel hypothesisand was srcinally proposed by Jeannerod (1988). An al-ternative model suggests that separate visuomotor chan-nels exist for the finger and the thumb. This hypothesiswas proposed by Smeets and Brenner (1999), and wewill refer to it as the digit channel hypothesis. In the dig-it channel hypothesis, Smeets and Brenner (1999) aban-don grip as a variable within grasping. Instead, they re-gard grasping as simply moving the fingers and thumb topositions on the surface of an object of interest. One im-portant feature of Smeets and Brenner's model is that thedigits approach surfaces perpendicularly. This featurehas empirical support (Smeets and Brenner 1995), andthis constraint allows Smeets and Brenner (1999) to gen-erate realistic digit trajectories for grasping movements.These authors (Smeets and Brenner 1999) provided anextensive review of extant studies on reaching and grasp-ing to show that their model can account for observedhuman prehensile behaviour.Smeets and Brenner (1999) have highlighted variousadvantages that the digit channel hypothesis holds overthe visuomotor channel hypothesis. First, since they claimthat both the finger and thumb are transported, it avoidsthe problem of deciding which anatomical part of thehand is controlled in the transport phase (an inherent dif-ficulty with the visuomotor channel hypothesis). Second, M. Mon-Williams ( ✉ ) · R.D. McIntoshSchool of Psychology, University of St Andrews, St Andrews,Fife KY16 9JU, Scotlande-mail: Tel.: +44-1334-462074Fax: +44-1334-463042Exp Brain Res (2000) 134:268–273DOI 10.1007/s002210000479 RESEARCH NOTE M. Mon-Williams · R.D. McIntosh  A test between two hypotheses and a possible third way  for the control of prehension Received: 28 October 1999 / Accepted: 10 May 2000 / Published online: 29 July 2000©Springer-Verlag 2000  269 the account allows the movement of the digits to be mod-elled using existing accounts of motor control such asthe minimum-jerk model. Third, the model takes accountof the contact points on an object when describing gripformation. Although the digit channel hypothesis has ad-vantages over the visuomotor channel hypothesis, it re-mains an empirical question as to which hypothesis bestcaptures the nature of prehension. Clearly, describing theopening and closing of the grasp aperture in terms of themovement of the two digits is a mere tautology. Smeetsand Brenner (1999) are not providing a tautology, how-ever, but are suggesting a different underlying neural or-ganisation to that traditionally envisaged. Differentiatingbetween the two hypotheses is thus difficult unless wecan gain an insight into the underlying organisation of the system. One possible way of testing the two hypothe-ses is to manipulate the manner in which a prehensilemovement can be carried out and then to observe the ef-fect of that manipulation on a known feature of thesystem's organisation. Arguably, Paul Fitts (1954) hasprovided the only lawful account of the system's behav-iour, and thus the best approach to testing the two hypo-theses might be usefully centred around Fitts' law. Itshould be noted that Fitts' law was developed to accountfor simple aiming movements and not prehension move-ments per se. Nonetheless, numerous studies have shownthat Fitts' law generalises to movement other than simplearm movements (Langolf et al. 1976). Fitts' law de-scribes the speed-accuracy trade-off in aiming tasks asfollows:  MT  = a + b log 2 (2  A  /  W  )(1)where MT is movement time, a and b are constants thatdepend upon the individual and the task,  A is the move-ment amplitude and W  is the target width. Log 2 (2  A  /  W  )is referred to as the index of difficulty (ID). Once more,imagine reaching out to pick up a wine glass, but thistime picture it located between a bottle of Claret on theright and a glass of water on the left. It is clear that thesystem must now take into account the obstacles presentwhen grasping the wine glass and modify its commandsaccordingly. The manner in which the commands aremodified will be different depending upon whether thesystem is controlling two digits or two components of the prehensile movement.We tested the two hypotheses by exploring the extentto which the different hypotheses could account for thetime taken to grasp a target object located next to an ob-stacle. According to the visuomotor channel hypothesis,the nervous system is trying to match the distance be-tween the digits with the size of the target (“target gripaperture” in Fig.1). The size of the target grip aperture isdictated by the gap between the two obstacles (under theassumption that the system is trying to ensure that the in-ter-digit distance is smaller than the gap between the ob-stacles and larger than the object). The visuomotor chan-nel hypothesis thus predicts that MTs will be governedby the width of the “grip aperture” (i.e. the distance be-tween the two obstacles). In contrast, the digit channelhypothesis suggests that MT should be predicted by thewidth of the aperture for the thumb and the width of theaperture for the index finger. In this situation, the Fitts'law relationship becomes more complicated. Fitts' lawstates that MT may be predicted from ID using Eq.1(with particular values of  a and b ) for a particular personperforming a particular aiming task. It does not state thatchanging the aiming task (or the conditions under whichit is executed) will allow Eq.1 to predict MT from IDwith the same values of  a and b . Indeed, it is known thatthe values of  a and b can vary from task to task and areconsequently rather sensitive to task parameters. In gen-eral, therefore, Fitts' law can be written:  MT  =  A ( T  ) +  B ( T  )  ID (2)where  A and  B are functions of the task context (T), tak-ing particular constant values ( a and b ) for particulartask conditions. In bi-digit aiming, the task context is de-termined by the IDs of the two targets and so, for thethumb, Eq.2 becomes:  MT  i =  A i (  ID  R  ,ID  L ) +  B i (  ID  R  ,ID  L )  ID i (3)where i=R or L (R and L indicate the thumb and indexfinger, respectively). Equation3 implies that the MT of one digit can depend both upon the ID of its own targetand on what the other digit is doing. As it stands, Eq.3 isan empirical relationship which can be considered a gen-eral but completely unconstrained version of Fitts' lawthat, in principle, allows MT to be almost any functionof the IDs. In order to make Eq.3 into a meaningful ex-pression of Fitts' law, additional simplifications and con-straints are needed. Ultimately such constraints mustcome from fitting models of the form of Eq.3 to the ex-perimental data. The simplest form of Eq.3 that can use-fully be considered arises from the following set of sim-plifying assumptions: (1) The influence of a constanttask for one digit on the other digit is independent of theother digit's task. This means that  A L and  B L would de-pend only upon ID R , and  A R and  B R only on ID L ; (2)  B i is constant ( b i ); (3)  A R and  A L are linear functions withconstant coefficients:  A R (ID L )= c R + m R ID L ; (4) the con-stants b i and m i are equal. If these assumptions are inplace, then Eq.3 reduces to the simplest form possible,which is consistent with the finding of synchronousmovements of the two digits:  MT  i = c i + d  i (  ID  L +ID  R )/2(4)which is a Fitts' law relation stating that the MT of the in-dex finger is a linear function of the mean index of diffi-culty of the thumb and the index finger's target,(ID L +ID R )/2. If the two digits move synchronously(MT R =MT L ), then c i and d  i are the same for both thumband index finger. A previous experiment (see Wann et al.1998 for a discussion of the data currently submitted forpublication) has established that bimanual MTs are welldescribed by a Fitts' law relationship involving the meanID of the two targets if the attentional demands of the task are constant (i.e. MTs were predicted by Eq.4). In the cur-rent experiment, the index of difficulty of the thumb's “tar-  270 get aperture” was held constant and thus the attentionaldemands were constant for this digit. The digit channelhypothesis then predicts that the MT should be predictedby the meanID of the two targets (the target aperture forthe thumb and the target aperture for the index finger). Materials and methods Six people (four men, two women) participated in this study volun-tarily. All were members of the School of Psychology at the Univer-sity of St Andrews and had normal or corrected-to-normal vision.The age range was 21–28years and all subjects were naive to theaims of the experiment. The experimental task was to reach forwardand grasp a target object using a precision grip. An obstructing ob- ject was placed on each side of the target object (see Fig.1). Targetand obstructing objects were arranged on a smooth, flat, white tablesurface. The target object was a rectangular (6cm height × 3cmwidth × 2cm depth) block of plastic painted green. The obstacleswere unpainted grey plastic blocks, 20cm height × 3cm width × 1cm depth. Positions of three small infra-red-emitting diodes(IREDs) placed on the participant's reaching limb were recorded byan Optotrak movement recording system. The Optotrak camerasystem was positioned approximately 2m from the experimentalworkspace at a height 1.5m above the table surface. The system wasfactory-precalibrated to a static positional resolution of better than0.2mm (dynamic resolution at speeds characteristic of human armmovements in reaching to grasp tasks was not significantly differentfrom this). The three-dimensional (3D) coordinates of the IREDswere referenced to a coordinate system defined by three IREDs.With this set-up, the 3D coordinates of IREDs within the workspacecould be measured to within ±0.5mm of their positions as measuredwith a ruler. Three IREDs were attached to participants' right arms atthe wrist (styloid process of radius), distal phalanx of the index fin-ger and of the thumb as indicated in Fig.1. Positions of the IREDswere recorded at a sampling rate of 100Hz.Participants reached for the target object which was placed ei-ther 20cm or 30cm from the start position along the centreline,which was approximately along the participant's midline (Fig.1).The hand was initially positioned with the wrist in a relaxed neu-tral posture (neither flexed nor extended), with the fingers flexedand the thumb and index finger touching. The hand was positionedsuch that the point at which the thumb and index finger pads metwas above the start point defined as the junction of the T in Fig.1.Reaches were made under eight obstacle conditions. The gap be-tween the object and the obstacle on the side of the index fingervaried in each condition. The gaps were 2.1cm, 3.7cm, 5.6cmand 7.7cm when the object was at 30cm, and were 2cm,2.75cm, 3.6cm and 4.5cm when the object was at 20cm. Thegap between the object and the obstacle on the side of the thumbwas maintained at a constant index of difficulty (the gap was ap-proximately 4cm at 30cm and 3cm at 20cm). The gap sizeswere chosen simply to provide a range of indices of difficulty thatwould allow differentiation between the two hypotheses. We en-sured that the thumb gap had a fixed index of difficulty, in order toensure that the attentional demands for this digit remained con-stant over the trials. It will be noted that this arrangement resultsin asymmetric positions for obstacles on either side of the target. Ithas been established previously that equally spaced flanking ob- jects produce much less effect when located on the side contralat-eral to the reaching limb, i.e. when the flanking object is ipsilater-al to the fingers and contralateral to the thumb (Jackson et al.1995). Our stimulus configuration thus provided us with a goodtest between the two different hypotheses. We avoided having a noobstacle condition, as it is not possible to calculate an index of dif-ficulty for such a condition.Ten reach trials in each of the eight obstacle conditions werepresented in a randomised order (80 trials in total). Participantswere instructed to reach out and grasp the target, pick it up and thereplace it on the table, in the presence of the obstacles. They wereinstructed to grasp the target object on the lateral surfaces. Partici-pants were allowed a small number of practice trials with the obsta-cles randomly placed in one of the eight possible positions. Partici-pants were instructed to reach as quickly but as accurately as possi-ble. The participants were explicitly told to avoid touching the ob-stacle. In the event, no participant touched the obstacle during thepractice period. Following the practice period, the participant per-formed the blocks of experimental trials. Any trial during which aparticipant touched an obstacle was considered void and immedi-ately re-run; in the event this occurred only infrequently. Partici-pants were cued to start by one of the experimenters with the ver-bal signal “Go”. Data acquisition was initiated approximately si-multaneously with the experimenter's verbal start command. Datawere recorded for a period of 1.5s, which was always sufficient tocapture the whole movement to object contact. The raw  x ,  y and  z coordinates of each IRED were digitally filtered by a dual passthrough a 2nd-order Butterworth filter with a cutoff frequency of 10Hz. Following this procedure, the tangential speed of the IREDswas computed and from this the onset and offset of the reaching es-timate was estimated using a standard algorithm (see Jakobson andGoodale 1991 for details). We also inspected six other kinematicvariables in order to ensure that the prehensile movements were“normal”: (1) peak velocity, (2) peak acceleration, (3) time to peak velocity, (4) time to peak acceleration, (5) time spent decelerating(the time to peak velocity subtracted from the MT), (6) normalisedtime spent decelerating (deceleration time divided by the MT). Wecould make no a priori quantitative predictions regarding the effectof obstacles on these parameters (Fitts' law only describes MT),and thus the dependent measure was MT. Fig. 1 . Schematic of the experimental configuration. The hand be-longs to James Tresilian and was drawn by Anna Plooy from acaptured video image. In the posture shown, the distance from thetip of the index finger to the styloid process of the wrist is approx-imately 15cm. Participants reached to an object ( solid black square ) placed on the midline. Obstacles ( hollow rectangles ) wereplaced on the right and the left of the target. According to the vis-uomotor channel hypothesis, task difficulty should be dictated bythe distance reached and the total distance between the inner edgesof the two obstacles (the “grip aperture”). According to the digitchannel hypothesis, the task difficulty should be dictated by thedistance reached and the width of the gap for the thumb (thumbaperture) together with the width of the gap for the finger (indexfinger aperture). See text for details  271 Results Table1 provides the kinematic variables recorded duringthe experiment. It can be seen that the participantsshowed normal reaching and grasping responses. Fig-ure2 shows the predicted pattern of results for the twohypotheses. The left-hand column shows the predictedresults if the visuomotor channel hypothesis is correctwhen MT is plotted against the ID calculated from thevisuomotor channel hypothesis (Fig.2a) and when MT isplotted against the ID calculated from the digit channelhypothesis (Fig.2b). The right-hand column shows thepredicted results if the digit channel hypothesis is correctwhen MT is plotted against the ID calculated from thevisuomotor channel hypothesis (Fig.2c) and when MT isplotted against the ID calculated from the digit channelhypothesis (Fig.2d). Plotting the MT against the incor-rect ID produces distinct patterns in the relationship be-tween ID and MT – these patterns thus serve as usefulqualitative features for discriminating between the twohypotheses.Figure3 shows the actual MTs plotted against the IDcalculated from the visuomotor channel hypothesis, andthe ID calculated from the digit channel hypothesis. Itshould be noted that the MTs of the thumb and index fin-ger were synchronous so that the movement of the two Table 1 Summary of the seven kinematic variables with SDs acrossparticipantsMeanSDMovement time (ms)526.120171.220Deceleration time (ms)318.179116.980Normalised deceleration time (%)59.2900.030Peak acceleration (cm/s 2 )55.04025.670Peak velocity (mm/s)991.869185.794Time to peak acceleration (ms)299.997103.441Time to peak velocity (ms)207.94154.537 Fig. 2a–d Schematic qualita-tive behaviour of movementtime as a function of index of difficulty predicted by the vis-uomotor channel hypothesis( right-hand column ) and thedigit channel hypothesis ( left-hand column ). a and c show themovement time data plotted asa function of the index of diffi-culty predicted by the visuomo-tor channel hypothesis. b and d show the movement time dataplotted as a function of the in-dex of difficulty predicted bythe digit channel hypothesis.It will be noted that plottingmovement times against theincorrect indices of difficultycauses a distinctive pattern of results. These patterns thusserve as useful qualitative fea-tures for differentiating betweenthe two hypotheses. See textfor details  272 digits began and finished together. It is clear that the em-pirical data correspond to the left-hand column of Fig.2when Fig.2a and Fig.2b are compared with Fig.3a andFig.3b, respectively. The data plotted in Fig.2a werewell described by a linear fit apart from one point thatindicated a MT that was 45ms slower than that predictedfrom the linear fit (this point was excluded from the fitshown in Fig.3). This point corresponds to the conditionin which the index finger's obstacle was furthest awayand the object was at 30cm. The slower time can be ex-plained by the presence of a ceiling effect, where the dis-tance to be reached and the presence of an obstacle onthe side of the thumb place constraints on the speed of the movement regardless of the distance away of the ob-stacle on the side of the index finger. It should be notedthat this ceiling effect does not affect the conclusions inany way – in fact, the MT predicted from the linear fit inFig.3a would improve the quantitative similarity be-tween Fig.3b and Fig.2b. Nonetheless, the qualitativesimilarity between Fig.3b and Fig.2b is particularlystriking, with the empirical data closely following thepredicted pattern of results. The results thus stronglysupport the visuomotor channel hypothesis over the digitchannel hypothesis. Discussion Our results are highly suggestive: the empirical datastrongly support the visuomotor channel hypothesis. TheMT for reaching out and grasping the object was wellpredicted by the total width of the aperture between thetwo obstacles. In contrast, the mean ID of the target ap-erture for the finger and thumb did not predict MT. Itshould be noted that the ID for the thumb was held con-stant; it follows that the MT could not be predicted froma consideration of the ID of the target aperture for thethumb, finger or the mean ID. Notably this is not true forbimanual tasks where the limb movements are well de-scribed by a Fitts' law relationship involving the meanID of the two targets (see Wann et al. 1998 for a discus-sion of the data currently submitted for publication).In the Introduction we highlighted various advantagesthat the digit channel hypothesis holds over the visuomo-tor channel account. Our findings raise the question of whether the empirical data must force us to “abandon thebaby together with the somewhat dirty bathwater”? Wesuggest that this may not be necessary and would like topropose an alternative hypothesis that retains many of the digit channel features within the visuomotor channelframework. We will call the alternative account the “third-way” hypothesis. According to the digit channel hypoth-esis, both the thumb and the index finger are being trans-ported when carrying out a precision grip. Let us sup-pose, however, that the nervous system is concernedwith transporting just one digit to the object and that theactual digit selected depends upon the task in hand. Forexample, imagine reaching for the wine glass with thebottle of Claret on the thumb side but nothing on the oth-er side of the glass. In this situation it is likely that thegap between the glass and the bottle will be foveated. Wesuggest that the system would then programme a move-ment to transport the thumb to the foveated location. Al-ternatively, the bottle might be located on the index fin-ger side of the glass and this gap is then fixated. In thiscase the system might choose to programme a transporttrajectory for the index finger. Note that in both of these Fig. 3a,b Actual movement times (mean across six participants)plotted as a function of the index of difficulty predicted by the vis-uomotor channel hypothesis ( a ) and the index of difficulty predictedby the digit channel hypothesis ( b ). The point marked as an asterisk in a was not included in the linear regression analysis (least-squaresfits to the data) shown on the graph. The movement time for thisexperimental configuration was 45ms slower than that predictedby the linear fit. The slower movement time can be accounted forby the presence of a ceiling effect (see text for details). It shouldbe noted that the presence of a ceiling effect for this configurationdoes not alter the conclusions. Compare a and b with the left- andright-hand columns of Fig.2. It is clear that there is excellent quali-tative agreement between Fig.2a and b and a and b , respectively.The results thus favour the visuomotor channel hypothesis overthe digit channel hypothesis
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks