Visual control of reaching movements without vision of the limb

Exp Brain Res (1986) 62:293--302

nResearch �9 Springer-Verlag 1986

Visual control of reaching movements without vision of the limb

I. Role of retinal feedback of target position in guiding the hand

C. Prablanc 1, D. P61isson 1, and M.A. Goodale ~

I Laboratoire de Neuropsychologie Exp6rimentale, INSERM - Unit6 94, 16, Avenue du Doyen L6pine, F-69500 Bron, France 2 Department of Psychology, University of Western Ontario, London, Ontario N6A 5C2, Canada

Summary. The spatial and temporal organization of hand and eye movements were studied in normal human subjects as they pointed toward small visual targets. The experiment was designed to assess the role of information about target position in correcting the trajectory of the hand when view of the hand was not available. To accomplish this, the duration of target presentation was systematically varied across blocks of trials. The results of this experiment showed that pointing movements were about 3 times more accurate when the target was present throughout the entire pointing movement, than when the target disappeared shortly after the hand movement had begun. These data indicate that pointing movements made without view of the limb are not purely preprogrammed but instead, are corrected during their execution. These modifications to the motor program are smoothly integrated into the ongoing movement and must depend upon comparing visual information about the position of the target with non- visual information about the position of the limb. The source of this non-visual information was not directly established in the present experiment but presumably must be derived from kinesthetic reaffer- ences and/or efference copy.

Key words: Retinal feedback - Limb movement - Trajectory control - Man

Introduction

From the recent studies on goal directed arm movements, a considerable number of results have been

Offprint requests to: C. Prablanc (address see above)

interpreted in the light of the well known speed- accuracy trade off. Fitts (1954) interpreted the mathematical relationship between the duration of a movement and its index of difficulty (ratio of required displacement to target size) as a function of the limited capacity of channels within the CNS to transmit information. At the same time he assumed that the error of the initial motor program is corrected "on line" by comparing the position of the target with the position of thc limb, the retina playing the same role as a feedback comparator in a ser- vomechanism. This interpretation, however, was challenged by Klapp (1975) who questioned whether Fitts' law reflects the time necessary for the corrections based on feedback or the time necessary for the optimization of the programming itself. He showed that for small movements an increase in accuracy could be obtained by lengthening the reaction time, presumably because the increased time permitted a better selection of motor commands to the muscles. Rabitt and Rodgers (1977) have also shown that amendments to an ongoing response could be based on information derived from previous experience and extrapolation from previous trials.

Following Fitts' idea, a number of authors (Keele and Posner 1968; Beggs and Howarth 1972; Carlton 1981; Crossman and Goodeve 1963) have suggested (or assumed) that the corrections that occur during the course of a reaching movement depend largely on comparing the position of a seen hand with that of a seen target. According to this view, the initial movement of the limb is ballistic and only later does dynamic visual feedback correct the movement trajectory (Paillard 1980; Beaubaton and Hay 1982). However, in an earlier experiment (Prablanc et al. 1979), we showed that when subjects were required to point toward a small visual target without any visual feedback from their moving hand, the relationship between movement duration and accuracy

294

stimulation

feedback eye position

,f

"C'

' ,/' / Or

1.,' 1 I _ T. LLT.o _l -

- r ~ - .

hand position

\

\

E'

Fig. 1. Experimental apparatus showing the matrix of electroluminescent diodes (LEDs) on upper surface, the semi-reflecting mirror on middle surface and the recording lower surface. Each of the 8 LEDs used l~r5ng between C and E can be selected by a random code generated by the computer. The subject with his head fixed by a bite-board sees the target E' (virtual image of the lit LED E) on the pointing surface through the semi-reflecting mirror. Starting from the central position C', he can point without vision of his hand, the light between the 2 lower surfaces being turned off by an electronic shutter (within 5 ms) at onset of hand movement. Horizontal eye movements are recorded by electro-oculographic technique, and hand position by a thimble attached to the finger; when the thimble is on the surface (fed by a current) it transmits its coordinates. Eye and hand data acquisition are performed on line; each elementary response is displayed onto an interactive terminal for some seconds, allowing the experimenter to nullify artefacts. The left hand of the figure summarizes the parameters recorded on the oculomotor and manual response (from top to bottom) following a target step 0T:TE, TED, 0E and TH, THD, OH are respectively the latency, duration and position of the response, ATE is corrective saccade latency. The box "feedback" shows that a new stimulation can be triggered by eye velocity signal when it exceeds a threshold level (see Part II)

nevertheless conformed to Fitts' law. This result would suggest that visual feedback about the relative positions of the hand and target is not the only source of information about the accuracy of the reaching movement, and that information derived from pro- prioception and/or efference copy of the motor commands can also play a role in modifying an ongoing movement. Such information must have been compared to some internal representation of the position of the target.

Most of the recent investigations on pointing movements at visual targets have paid little attention to the nature of the representation of the target. In the case of repetitive movements made between two targets in the same location, a "stored" internal representation of target position relative to the body can be used. On other occasions, foveal or non- foveal vision of the target might be used as the movement is executed. In either case, it is often

implicitly assumed that the information about the location of the goal contains no error component , and that errors occur at the level of the programming and of the execution of the movement. In our earlier work (Prablanc et al. 1979), we showed instead that subjects are less accurate in pointing in the dark at a peripheral visible target when they are not allowed to foveate the target. This result suggests that an increase in accuracy of locating the target is due either to the sharpness of the foveai image, or to the alignment of the gaze onl~o the target, or to both.

The present experiment was designed to extend these observations of pointing movements made without view of the hand, to examine the effect of manipulating the time over which the target was visible, and to clarify the nature of the information that the target provided. In every case, subject was pointing toward a small visual target as soon as it was presented in his peripheral visual field. In no case,

295

TARGET

EYE

HAND

Q 500

0 o

I o

I i

,

f / 2 ~ ~ 0 ~ 0

EYE POSITION ERROR

. . . . . . . . . . 1[HAND POSITION ERROR

t

i i i 800 ~000 IZO0 l~ (ms)

| [ 120ms �9 [ I r / f

t

f

| . . . . J . . . . . . . .

f

t

|

/ J

_ _ i n . _ _ HEADPHONE

Fig. 2. The four experimental conditions. In all these, vision of the hand is allowed in the central position and is cut off at onset of hand movement (thick arrows). The three traces represent target, eyes and hand position, respectively. The steps of the target position followed by motor responses which are recorded are always from the center position to a randomly selected peripheral one, therefore the backward steps are not shown. Errors are indicated by the arrows for oculomotor and manual responses. The vertical dotted lines indicate the onset of hand movement (also the end of target presentation) in condition 1, the end of the main saccadic response in condition 2, the onset of hand movement and the end of the main saccade (signaled to the subject by a click at the headphone) in condition 4. In the three first conditions the subjects have to respond as quickly as possible with both their eyes and hand; visual stimulus is cut off at the onset of hand movement in condition 1, or 120 ms after the end of the main saccade in condition 2, or stimulus is present for 2.5 s, i.e. throughout the motor responses in condition 3. Instead, in condition 4 the subjects have to wait for a click at the headphone, triggered by the end of their main oculomotor response, before pointing to the target which vanishes at the onset of the hand movement (as in condition 1)

could the subject see his hand as it moved toward the target. What we manipulated was the time the target remained visible during the movement and the way in which the subject was instructed to move his hand and eyes. Several questions were asked. First, would subjects be more accurate when the target remained visible than when it disappeared after the reach had begun? If they were, then this would suggest that some sort of reafferent control of movement was taking place even when visual information about the position of the hand was not available. Second, if they did do better when the target remained visible, was the information about the position of the target that they used to modify their movement derived from gaze position alone or from gaze position plus a retinal error signal? In other words, would a foveal anchoring on the target favor pointing accuracy much more than a foveal landing followed shortly by disappearance of the target.

Methods

Apparatus

The apparatus we used is illustrated schematically in Fig. l (for details, see Prablanc et al. 1979). An array of red light-emitting diodes (LEDs) was located above a half-reflecting mirror. Below the mirror was another surface, on which the virtual image of an LED was projected. A subject was seated in front of the apparatus with the index finger of his right hand resting on the lower surface. The subject's head was immobilized with a bite bar and horizontal eye movements were monitored by electrooculography (with auto- matic drift control and with a low-pass filter from DC to 30 Hz). The illumination of the surface beneath the mirror (through which the subject could see his hand) could bc eliminated within 5 ms by means of an electronic shutter.While the subjccrs view of his hand would disappear, he would still see the virtual image of the LED in depth on the surface below the mirror. The subject 's fingertip was covered with a metal thimble. When this thimble contacted the conductive surface on which the subject made his pointing movements, the coordinates of that position were recorded. The spatial location and the duration of the target along a fl'onto-

296

parallel line were controlled by an on-line computer (PDP8) that monitored and recorded the movement of the hand and eyes.

Experimental control and data acquisition

The computer-based system for controlling the experiment and recording data has been described in detail in an earlier publica- tion (Echallier et al. 1978). Position signals from the eyes and hand were sampled at 200 Hz for 1.5 s following target presentation. These signals were stored on digital tape with an on-line computation of the following parameters: latency and duration of the initial saccade, eye position at the end of that saccade, latency and duration of the hand movement, and the position of the index finger at the end of the hand movement. Logic signals generated at the beginning and end of the cyc and hand movements could be used to control onset and offset of illumination of the hand and the duration of the target.

Procedure

Ten subjects with normal vision were tested. After their EOG signal had stabilized, they wcrc instructed to keep their gaze and the index finger of their right hand on a central target located about 54 cm in front of them on their body axis. When this target jumped from its central location to a randomly selected position in their right periphery, they were required "to look and point to the target as quickly and as accurately as possible". With the exception of one testing condition described below, no mention was made of the sequence of eye and hand responses they might makc. As soon as their finger left the surface on which it was resting, their entire hand and forearm vanished from view. After completing their movement, subjects had to wait for the target to reappear at the center, before returning their gaze and hand to the central position. Illumination of their hand was restored only at the onset of the hand return movement when the target was already back to its central position. Subjects therefore had no visual information about the accuracy of their pointing movcment to the peripheral target, even though they could accurately return their finger to the central target under direct visual control.

Peripheral targets were always presented on the right sidc along a frontoparallel line at distances of 20, 24, 30, 34, 40, 44, 50, and 54 cm from the central target with an equal probability of occurrence. As the fronto-parallcl line of targets lay at a distance d = 54 cm from the subjects' eyes, the resulting ocular angle from the midsagittal plane was equal to arctg (target eccentricity/ distance), corresponding respectively to 20.3 deg for 20 cm, 29 deg for 30 cm, 36.5 dcg for 40 cm and 42.8 deg for 50 cm. Although thc main parameters described refer to hand pointing accuracy, it should be kept in mind that target position in cm from the midsagittal plane corresponds to a slightly smaller number in degrces of visual angle. Only responses to the 20, 30, 40, and 50 cm targets were analyzed. The other targets were used to prevent the subjects from constructing an intcrnal represcntation of the target positions, something that might have occurred had we used only 4 target positions. When asked at the end of the experiment, how many targets had been presentcd, most subjects answered 5 or 6.

Experimental design

Each of the 10 subjects was tested under four different conditions (see Fig. 2). In all conditions, the same initial cues were available, namely, normal view of the hand and target prior to the hand movement. As soon as the finger lifted from the surface, however, view of the hand disappeared. Each condition consisted of 80 trials

in which each of the eight different targets were presented 10 times in random order.

The same instructions were given in conditions 1, 2 and 3 below, in order to obtain the same close coupling between the onset of e.m.g, activity in eye and hand, as that had been already observed by Biguer et al. (1982) in a similar task.

In Condition 1, when the index finger left the central position, vision of both the hand and the target disappeared.

In Condition 2, the target remained in view slightly longer, disappearing 120 ms after the end of the first saccade. This delay permitted the subject to make a corrective saccade so that his gaze could be aligned with the target position (even though the target itself would no longer be visible).

In Condition 3, the target remained illuminated throughout the entire pointing movement. In this condition, both accurate gaze position information and a retinal error signal were available to the subject.

In Condition 4, subjects were asked to delay their hand movement until they had completed their first eyc movement toward the target. A brief tone signalled the end of their first saccade, initiating a pointing movement to the target. As soon as their finger moved from the central position, view of both their hand and thc targct disappeared. Thus, in this condition, subjects had an opportunity to preprogram their hand movement on the basis of accurate foveal information about the position of the target.

Half the subjects (Group I) werc tested in the order, Condition 1, 4, 2, and 3. The other half (Group II) were tested in the reverse order.

Results

F o r e a c h o f t h e 10 s u b j e c t s , t he l a t e n c i e s o f t h e e y e

and the h a n d , t h e d u r a t i o n o f t he h a n d m o v e m e n t ,

and the e r r o r o f t hc f inal p o s i t i o n o f t h e f i n g e r w e r e

a v e r a g e d ac ross t h e 10 t r ia ls fo r t he 20, 30, 40, a n d

5 0 c m t a rge t s u n d e r e a c h o f t h e f o u r c o n d i t i o n s ( see

T a b l e 1). T h e s e d a t a w e r e s u b j e c t e d to an analys is o f

va r i ance . T h e des ign o f this ana lys i s was 4 ( C o n d i -

t ion) • 4 ( t a rge t d i s p l a c e m e n t ) • 2 ( o r d e r o f

t es t ing) . T h e resu l t s o f this ana lys i s a r e s u m m a r i z e d

in T a b l e 2.

Latencies

A s T a b l e 1 i nd i ca t e s , t h e l a t ency o f t h e first s a c c a d e

i n c r e a s e d s l ight ly b u t s ign i f i can t ly in all c o n d i t i o n s

as the d i s p l a c e m e n t o f t h e t a r g e t i n c r e a s e d (F 3,

1 2 8 = 6 . 8 , p < 0 . 0 0 1 ) , w i th i nd iv idua l m e a n s v a r y i n g b e t w e e n 233 ms ( S D = 21) a n d 302 ms ( S D = 34). A

s imi la r r e l a t i o n s h i p b e t w e e n t h e l a t e n c y and

a m p l i t u d e o f e y e m o v e m e n t s has b e e n o b s e r v e d in a

n u m b e r o f d i f f e r e n t s tud ies ( B a r t z 1962; W h i t e e t al.

1962; B e c k e r and F u c h s 1969). T h e e f fec t s o f c o n d i -

t ion and o r d e r o f t e s t i ng on e y e m o v e m e n t l a t ency d id no t r e a c h s ign i f i cance n o r d id t he i r i n t e r a c t i o n s . In a qu i t e d i f f e r e n t s t udy o n e y e - h e a d l a t enc i e s , Z a n g e m e i s t e r and S t a r k (1981) w h i l e t h e y f o u n d a

297

Table I. Average results of the 10 subjects in the four conditions for the four main parameters computed: eye latency (TE), hand latency (TH), hand movement duration (THD) and constant error (EH). Means and their standard dcviation (in brackets) computed for cach condition are shown in the last raw

Conditions 1 2 Parameters TE TH THD EH TE TH THD EH

20 243 380 365 -1.5 250 363 352 -1.4 30 252 382 413 -3.2 250 361 407 -2.7

Amplit(cM) 40 2711 405 452 -4.2 261 363 455 -3.3 50 288 430 493 -4.7 267 390 472 -2.5

Mean (SD) 263 399 431 -3.4 257 369 421 -2.5 (33) (92) (112) (2.54) (36) (51) (85) (2.2)

Conditions 3 4 Parameters TE TH THD EH TE TH THD Et I

20 236 341 395 -0.6 25l 544 433 -1.3 30 245 342 447 -1.8 257 583 484 -2.4

Amplit(cM) 40 248 351 493 -2.1 263 648 515 -2.6 50 264 348 535 -1.4 288 674 557 -2.5

Mean (SD) 248 345 467 -1.5 265 612 497 -2.2 (34) (35) (142) (2.07) (32) (121) (129) (1.94)

800

700

6OO

E 500

LO 4OO

3OO

2O0

I00

---/,r

ENDOF ~ e HAND MOVEMENT

o ~ e /

_ T I END OF SACCADE

I I I I I~ 20 30 40 50

TARGET AMPLITUDE ( cm )

Fig. 3. Temporal parameters of the two motor responses in condition 1 plotted against target amplitude. Time 0 is the instant of target jump, and the parameters are the latency and the timc of completion of the oculomotor main response and of the manual response, respectively, Vertical bars represent the standard deviation of the means of the lateneies. It is interesting to note the parallel between the latencies of eye and hand as a function of target amplitude. Note also that the zero intercept of the difference between the end of saecade and the hand latency lies near 40 cm of target amplitude

large inf luence of ins t ruc t ion on head la tency, found very litt le on eye la tency.

Since hand m o v e m e n t s in Cond i t i on 4 occur red only af ter the subjec t s had fovea ted the ta rge t , the hand la tency measu res f rom this cond i t ion were not

included in the analysis . T h e only s ignif icant fac tor in this analysis was cond i t ion (F 2, 96 = 6.01, p < 0.001) with la tencies dec reas ing f rom C o n d i t i o n 1 to 3 (see Tab le 1). T h e r e were no signif icant in te rac t ions .

The d i f fe rence b e t w e e n eye and hand la tency var ied f rom 136 ms in Cond i t i on 1 to 98 ms in Cond i t ion 3. These va lues are qui te s imi lar to the i sometr ic con t rac t ion t ime for h a n d m o v e m e n t r eco rded in an ana logous task (Biguer et al. 1982) and thus suggest tha t the c o m m a n d signals to the eye and to the g roup of muscles r e spons ib le for hand m o v e m e n t a re g e n e r a t e d in para l le l .

As Fig. 3 i l lus t ra tes , the ze ro in te rcep t for the end of the first saccade and the beg inn ing of the hand m o v e m e n t occur red at an ave rage ta rge t d isp lacemen t of 38 cm in Cond i t i on 1. This m e a n t tha t for target d i sp lacemen t s la rger than this va lue , the hand began to move be fo re the first saccade was com- p le ted , and subsequen t ly the sub jec t had no d i rec t visual in fo rmat ion abou t the pos i t ion of the t a rge t o the r than what had been de r ived f rom the or iginal pe r iphe ra l re t inal image. F o r sma l l e r t a rge t d isp lace- ments , the init ial saccade was c o m p l e t e d be fo re the hand had begun to move . Thus , for a 20 cm ta rge t , for example , a b o u t 70 ms of t ime for upda t ing in format ion a b o u t t a rge t pos i t ion was avai lab le be fore the hand began to move . Since hand la tencics were smal le r in Cond i t i ons 2 and 3, the ze ro in te rcep t for the end of the first saccade and the beg inn ing of the hand m o v e m e n t in these two cond i t ions occu r red at smal ler average ta rge t d i sp l acemen t s (33 and 28 cm, respec t ive ly) .

298

Table 2. Results of variance analysis for the same four parameters depicted in Table 1 are shown for the three factors studied: group of subjects (blocks), target amplitude and experimental condition. In this table, an estimatc of probability is shown when the effect reaches statistical significance (p < 0.05). The numbers of df for the computation of F arc shown in the top row for eye latency, hand duration and hand accuracy. They are different for hand latency becausc variance analysis had been conducted only with 3 levels of condition for this parameter (see text)

Factors Blocks Targets Conditions Interactions FA FB FC FAB FAC FBC FABC

Parameters (F1,128) (F3,128) (F3,128) (F3,1~) (F3,128) F(9,128) F(9,128)

Eye F=2.65 F=6.76 F=2.02 0.52 1.05 0.29 0.23 latency (TE) p<0.001

Hand F(1,96) F(3,96) F(2,96) F(3,96) F(2,96) F(6,96) F(6,96) latency (TH) =1.23 =1,09 =6.01 0.38 1.4 0.25 0.02

p<0.001

Hand F=66.78 F=7.99 F=3.21 0.18 0.32 0.02 0.03 duration (THD) p<0.001 p<0.001 p<0.05

Hand F= 12.62 F=5.37 F=4.97 0.1 0.2 0.45 0.08 accuracy (EH) p<0.001 p<0.005 p<0.005

E

Z 0 50C

r r

O I-- 4 0 0 Z UJ

UJ

~ 5 0 0

-g n,- o

w

z

if) z o

v

0 c r

klJ kJ . J o n

@

o . . . . . . . . . . .

~ ' - r I I I I l L -

|

,•,.,________- - o . . . . 7 ~ - - ' ~ - ~

. /L 1 I I I D_

Q ~ CONDITION [ �9 CONDITION 2 �9 CONDITION 3 L~ o CONDITION 4 I

- - . / t I 2o

I I I 30 4 0 ,50

I A R G E T A M P t . I T U D E ( c m )

Duration of hand movement

As Tables 1 and 2 illustrate, this parameter was affected by condition, target displacement, and testing order. The effects of testing condition and target displacement are illustrated in Fig. 4 where the duration of the hand movement is plotted as a function of target displacement for each of the four conditions. Among the first three conditions, it was Condition 3, in which the target was present for the entire pointing movement, that showed the longest movement durations. The difference between Condi- tion 3 and the other two conditions was significant for pointing movements directed at 40 and 50 cm targets. Paradoxically, however, Condition 4, in which subjects were required to foveate the target before initiating a pointing movement, produced the slowest hand movements, even though the maximum amount of information about target position was available before the movement took place and none was available during the movement itself. The significant

Fig. 4A--C. Relationship of the three main parameters of hand pointing versus target amplitude: duration (top), constant error (middle) and standard-deviation of constant error or variable error (bottom) for the four conditions. Note that duration is linearly related to target amplitude. Whereas the relationship of variable error versus target amplitude is poorly affected by the conditions, constant error shows a strong trend to decrease from condition 1 to 3, i.e. with increased target duration. Standard deviations of duration of movement are pooled in Table 1

299

i f? Z �9

klA

~ 2 123

Z

C) ..o

..,"

�9 ....... ~...;;,-c~ ....

. , �9

s ~ S

, s " I I I I I I I I I I 0 lO 20 30 40 5o 60 7o eo 90 ioo

M E A N V E L O C I T Y ( c m / s )

Fig. 5. Relationship between variable error (standard deviation) and average spccd (mean amplitude versus mean duration ratio for a given target amplitude) for the manual responses in the four conditions (labelled in right-hand of the curves) and for the overall means of the results (open triangles). These relationships are essentially linear, almost for thc ovcrall means, with their slope and zero intercept being only slightly affected by the conditions. The general relationship (through the open triangles) is really linear and can be measured directly on the figure: SD = 0.003 0H/THD

effect of target displacement on the duration of the hand movement (F 3, 128 = 7.99, p < 0.001) is illustrated in Fig. 4; that effect is present in situations where some emphasis is put on accuracy (Fitts 1954; Prablanc et al. 1979) by contrast with very fast movements either in man or animal (Freund and Biidingen 1978; Abend et al. 1982).

A large effect of testing order (i.e., the difference between Groups I and II) surprisingly appeared (F 1, 128 = 67; p < 0.001). The first group was always faster than the second under all conditions and at all target displacements. It must be emphasized, however, that there was no interaction between testing order and any other factor. In other words, the different conditions of testing and target displacements had the same effect on both groups of subjects.

Pointing accuracy

In all conditions, subjects had a strong tendency to undershoot the target, overshooting the target only 17% of the time. These constant errors (and thus for practical purposes, the absolute errors as well) were significantly affected by Conditions (F 3, 128 = 4.9, p<0.005). As Fig. 4 illustrates, accuracy increased from Condition 1 to 3, i.e., with increased duration

of target presentation. It is interesting to note that beyond 30 cm of target displacement, accuracy in Condition 2 becomes significantly greater than accuracy in Condition 1. Indeed, it is only for targets beyond 30 cm that Condition 2 offers an advantage over Condition 1 in that subjects have time to foveate the target before it disappears.

The size of the constant errors tended to increase as a function of target displacement (F 3, 128 = 5.37, p < 0.005) up to 40 cm after which it showed no change except in Condition 1 where errors increased from 20 to 50 cm (as can be seen in Fig. 4). It is worth noting that it was only in Condition 1 that the retinal images of targets at 40 and 50 cm (i.e. 36.5 and 42.8 deg of visual angle) would have been entirely peripheral. This suggests that the size of constant errors (linearly related to retinal eccentricity) may be proportional to retinal acuity, the latter decreasing almost linearly with target eccentricity (Kroon et al. 1980).

Accuracy also differed as a function of testing order (F 1, 128 = 12.6, p < 0.001). Subjects who were first tested with Condition I were less accurate over all conditions than subjects first tested with Condition 3, but as Fitts' law would predict, they were also faster.

300

Speed accuracy trade-off

Although this phenomenon was clearly evident in the different performances of the two groups of subjects who experienced different orders of testing, it was also analyzed more systematically by plotting variable error (standard deviation of the constant error) as a function of the average velocity of the pointing movements for each of the four conditions and for the overall performance (see Fig. 5). For all conditions, variable error shows a nearly linear relationship to average velocity and only slight differences in the slope can be seen. The overall relationship between variable error and velocity is very linear indeed.

Discussion

The results of pointing accuracy in this study clearly demonstrate that subjects can modify their pointing movements toward visual targets quite efficiently even when they have no view of their moving hand. Thus, "open loop" movements (open loop with respect to visual feedback about the relative positions of the target and the moving limb) are far from being uncorrected or ballistic. Indeed, the prime effect of target duration on pointing accuracy suggests that visual information about target position is somehow used to control the movement during its execution.

The existence of a reafferent control process that does not involve vision of the limb was expected. Our earlier work (Prablanc et al. 1979) demonstrated in a very similar task, that accuracy was better when the target was permanently illuminated than when it was extinguished at the onset of the oculomotor response, even when view of the hand was never available. Nevertheless, two other studies had produced very different results showing no or little influence of target duration upon pointing accuracy (Festinger and Canon 1965; Conti and Beaubaton 1976). The tasks employed in these investigations, however, were not very similar to those used in our own work. In the Conti and Beaubaton study, subjects were required to point toward a single target located at eye level from a lower resting position and thereby fixated the target (or the position of the target) throughout the experiment, even when the target was extinguished. The movements were thus initiated on the basis of either stored internal information about target position or on the basis of actual foveal information. In the second study (Festinger and Canon 1965), the manual response, which was not a pointing movement but a lever rotation toward the target position, occurred either immediately after

the target vanished or was delayed for 3 s. It is worth noting, however, that rotation of the lever was more accurate in this situation when the target was continuously present, although this difference did not reach significance.

If subjects are able to correct their trajectory during execution of the movement, then the question arises as to the source of the original error that requires correction. There are two possibilities. Either the initial motor programming contains some error due to an initiation on the basis of a non-foveal information about the position of the target, or errors occur at the level of movement execution (or both). If the first explanation is correct, then one might expect an improvement in accuracy if subjects are allowed to obtain foveal information before execut- ing the pointing movement. This is exactly what happened in Condition 4 of the present experiment when subjects were asked to foveate the target before pointing to it. Their performance in this Condition was much more accurate than it was in Condition 1 in which the commands to move the eye and hand were nearly simultaneous, even though in both cases, view of both the target and the hand was not available during the movement itself. Neverthe- less, Condition 4 did not produce the most accurate pointing movements. Subjects were most accurate during Condition 3, in which visual information about the position of the target was available throughout the pointing movement. This suggests that movements initiated on the basis of foveal visual information are not necessarily the most accurate if no opportunity for correcting them is provided during their execution. It appears that both initial programming and execution errors can lead to less than accurate pointing movements.

A finer analysis of error led us to consider separately the signed error (or constant error) and its standard deviation (or variable error). The second of these shows a strong linear relationship to the average velocity of the movements, a relationship that was first described by Schmidt et al. (1979) in very fast (less than 200 ms) pointing movements and formed the basis of their "impulse variability model". According to Schmidt et al. this model held only for uncorrected, ballistic movements. More recent observations by Wright and Meyer (1983) have led to the extension of this model to visually guided movements where the time constraints are very stringent. In our study, however, the duration of the movement was neither very short nor a variable controlled by the subjects. Nevertheless, a similar relationship between variable error and average velocity was observed, suggesting that this may be a characteristic feature of the motor system involved in making

301

pointing movements. Moreover, the fact that variable error appeared to be much less affected by the different conditions of the present experiment than constant error suggests that increasing the amount of target information available during the execution of the movement corrects for an initial bias in the motor program rather than a variable error. Furthermore, a linear relationship between constant error and target eccentricity (in degrees) was observed when the retinal feedback was removed as soon as the hand began to move (Condition 1). It suggests that the constant error may reflect either a perceptual bias in localizing the target or the tendency of the involved control systems to generate undershoots towards peripheral objects. This would be an efficient strategy for correcting the movement if more accurate information about target position is available later; i.e., it is better to underestimate rather than overestimate the amplitude of a required movement. This undershoot can then be corrected as more accurate foveal or parafoveal visual information becomes available (as it did in Conditions 2 and 3). In contrast to constant error, variable error would appear to be more related to the inherent noise in the motor system itself such as the muscular noise alluded to by Schmidt et al. (1979), and to be more sensitive to the movement extent and its velocity than to the experimental conditions.

The fact that on-going movements can bc controlled without visual feedback from the moving hand implies a mechanism of comparison between visual information about target position and extra-visual information about hand position. Therefore, at least two stcps in the controlling mechanism are necessary: accurate visual localization of the target yielding an internal representation of target position and a comparison between this representation and information about the position of the moving hand. The increase in accuracy between Conditions 2 (where the target vanished 120 ms after the first saccade) and 3 (where it remained in view for the entire movement) indicates that better visual target localization was available during the on-going movement in Condition 3 even though accurate oculomotor responses could occur in both conditions (Prablanc et al. 1978). This result confirms earlier suggestions by Prablanc et al. in 1979, and the recent observations of Mather and Fisk (1985) that pointing movements are not guided solely on the basis of an extra-retinal signal about eye position. While this conclusion is exactly opposite to that of Hansen and Skavenski (1977), the results of their study are somewhat questionable since the subjects' arm movements were well-practiced, and therefore based on the exact knowledge of the location of the single target. The

results of the present study suggest that both retinal and extra-retinal signals are necessary to encode target position accurately, at least when its location is randomly changed from trial to trial. The difference in accuracy between Conditions 2 and 3 must result from the necessity for the mechanism that summates the retinal and extra-retinal signals to be fed continuously by both signals throughout the pointing movement in order to update the internal representation of target position.

How is this internal target position information (updated after each saccade toward the target) used to fine-tune the on-going manual movement? This process requires that a comparison be made, either continuously or discretely, betwcen the position of the target and the position of the moving hand. Information about the position of the hand is presumably based on kinaesthetic feedback, efference copy, or on some combination of the two.

In conclusion, we have demonstrated that even when view of the moving limb is unavailable, correction of a pointing movement can take place as the finger is moved toward the target. The modifications that occur during the execution of the movement must therefore depend upon comparisons between visual information about target position and non- visual information about hand position. We have also shown that such a correction takes very little extra time. This is a strong argument against the existence of any extensive reprogramming of the motor control signals, and suggests that such corrections are part of the normal sequence of events that occurs when subjects reach toward visual targets. To reinforce this argument, it would be helpful to demonstrate that such on-line corrections can occur when much larger adjustments are necessary, and that such corrections fit well within the time-frame of normal pointing movements without inflexion point in the movement trajectory. It would also be important to know whether subjects are able to discriminate trials on which they make corrective adjustments in their movements from those in which they do not. In other words, what is the subjective phenomenology of such modifications in the trajectory of a pointing movement? The next study is designed to examine these questions (P61isson et al. 1986).

References

Abend W, Bizzi E, Morasso P (1982) Human arm trajectory formation. Brain 105:331-348

Bartz AE (1962) Eye movement latency, duration and response time as a function of angular displacement. J Exp Psychol 64: 318-324

302

Beaubaton D, Hay L (1982) Integration of visual cues in rapid goal directed movements. Behav Brain Res 5:92-93

Beckcr W, Fuchs AF (1969) Further properties of the human saccadic system: eye movements and correction saccades with and without visual fixation points. Vision Res 9:1247-1258

Beggs WDA, Howarth CI (1972) The movement of the hand toward a target. O J Exp Psychol 24:448-453

Biguer B, Jeannerod M, Prablanc C (1982) The coordination of eye, head and arm movements during reaching at a single visual target. Exp Brain Res 46:301-304

Carlton LG (1981) Processing visual feedback information for movement control. J Exp Psychol 7:1019-1030

Conti P, Beanbaton D (1976) Utilisation des informations visuclles dans le contr61e du mouvement: 6rude de la pr6cision des pointages chez l'homme. Trav Hum 39:19-32

Crossman ERFW, Goodeve PJ (1963) Feedback control of hand- movement and Fitts' law. Published (1983) Q J Exp Psychol 35A: 251-278

Echallier JF, Pernier J, Prablanc C (1978) Syst~me de pilotage de stimulations ct de recueil de param~tres relatifs ~ des r6ponses motrices. Int J Bio Med Comput 9:341-352

Festinger L, Canon LK (1965) Information about spatial location based on knowledge about efference. Psychol Rev 72: 373-384

Fitts PM (1954) The information capacity of the human motor system in controlling the amplitude of movement. J Exp Psychol 47:381-391

Freund JH, Biidingen HJ (1978) The relationship between speed and amplitude of the fastest voluntary contractions of human arm muscle. Exp Brain Res 31:1-12

Hansen RM, Skavenski AA (1977) Accuracy of eye position information for motor control. Vision Res 17:919-926

Keele SW, Posner RMI (1968) Processing of visual feedback in rapid movements. J Exp Psychol 77:155-158

Klapp ST (1975) Feedback versus motor programming in the control of aimed movements. J Exp Psychol 104:147-153

Kroon JN, Rij'sdijk JP, Van dcr Wildt GJ (1980) Peripheral contrast sensitivity for sine-wave gratings and single periods. Vision Res 20:243-252

Mather JA, Fisk JD (1985) Orienting to targets by looking and pointing. I. Parallels and interactions in ocular and manual performance. Q J Exp Psychol 37A: 315-338

Paillard J (1980) The multi channeling of visual cues and the organization of a visually guided response. In: Stelmach GE, Requin J (eds) Tutorials in motor behavior. North-Holland Publishing Co, Amsterdam

P61isson D, Prablanc C, Goodale MA, Jeannerod M (1986) Visual control of reaching movement without vision of the limb. II. Evidence of fast unconscious processes correcting the trajectory of the hand to the final position of a double-step stimulus. Exp Brain Res 62:303-311

Prablanc C, Echallier JF, Komilis E, Jeanncrod M (1979) Optimal response of eye and hand motor systems in pointing. I. Spatio- temporal characteristics of eye and hand movements and their relationships when va~ing the amount of visual information. Biol Cybern 35:113-124

Rabitt P, Rodgers B (1977) What does a man do after he makes an error? An analysis of response programming. Q J Psychol 29: 727-743

Schmidt RA, Zelaznik H, Hawkins B, Frank JS, Quinn JT (1979) Motor-output variability: a theory for the accuracy of rapid motor acts. Psychol Rev 86:415-451

White CT, Eason C, Barlett NR (1962) Latency and duration of eye movements in the horizontal plane. J Opt Soc Am 52: 210-213

Woodworth RS (1899) The accuracy of voluntary movement. Psychol Rev Monogr Suppl 3:3

Wright CE, Meyer DE (1983) Conditions for a linear speed- accuracy trade-off in aimed movements. Q J Exp Psychol 35A: 279-296

Zangemeister WH, Stark L (1981) Active head rotations and eye- head coordination. Ann NY Acad Sci 374:540-559

Received April 25, 1985 / Accepted November 19, 1985

Visual control of reaching movements without vision of the limb

Documents

Transcript of Visual control of reaching movements without vision of the limb