The type of visual information mediates eye and hand movement bias when aiming to a Müller–Lyer illusion

RESEARCH ARTICLE

Ann Lavrysen Æ Werner F. Helsen Æ Digby Elliott

Martinus J. Buekers Æ Peter Feys Æ Elke Heremans

The type of visual information mediates eye and hand movement biaswhen aiming to a Muller–Lyer illusion

Received: 28 October 2005 / Accepted: 31 March 2006 / Published online: 23 May 2006� Springer-Verlag 2006

Abstract Aiming bias typically influences perception butaction towards the illusory stimulus is often unaffected.Recent studies, however, have shown that the type ofinformation available is a predictor for the expression ofaction bias. In the present cyclical aiming experiment,the type of information (retinal and extra-retinal) wasmanipulated in order to investigate the differentialcontributions of different cues on both eye and handmovements. The results showed that a Muller–Lyerillusion caused very similar perturbation effects on handand eye-movement amplitudes and this bias was medi-ated by the type of information available on-line.Interestingly, the impact of the illusion on goal-directedmovement was smaller, when information about thefigure but not the hand was provided for on-line control.Saccadic information did not influence the size of theeffect of a Muller–Lyer illusion on hand movements.Furthermore, the illusions did not alter the eye–handcoordination pattern. The timing of saccade terminationwas strongly linked to hand movement kinematics. Thepresent results are not consistent with current dichoto-mous models of perception and action or movementplanning and on-line control. Rather, they suggest that

the type of information available for movement plan-ning mediates the size of the illusory effects. Overall, ithas been demonstrated that movement planning andcontrol processes are versatile operations, which havethe ability to adapt to the type of information available.

Keywords Context-induced illusion Æ Manual aiming ÆOn-line processing Æ Eye–hand coordination

Introduction

Over the last several years there has been considerabledebate about the way the central nervous system usesvision information in various perceptual and motortasks. Probably, the most influential position has beenMilner and Goodale’s (Goodale and Milner 1992; Mil-ner and Goodale 1995; see also Ungerleider and Mishkin1982) Perception Action Model (PAM) that posits adissociation between perceptual–cognitive judgmentsabout the visual characteristics of objects and the visualprocesses that are responsible for the control of limb andbody movements in 3D space. From a neuroanatomicalperspective, perceptual–cognitive judgments have beenassociated with the ventral pathway that extends fromprimary visual cortex to the inferior temporal areas ofthe brain. This visual stream is associated with theallocentric coding of space and plays an important rolein object recognition and identification. In contrast, thePAM holds that goal-directed movements are controlledby the dorsal stream. This visual pathway also originatesin primary visual cortex but terminates in the superiorparietal areas of the brain. The dorsal pathway is asso-ciated with body-referenced or egocentric coding of vi-sual space.

Support for the dissociation of function attributed tothese two visual pathways comes from a number ofanimal studies and clinical case studies involving braininjured humans (Milner and Goodale 1995). Forexample, a lesion to the ventral pathway results in severevisual agnosia (i.e., object and pattern recognition

A. Lavrysen Æ W. F. Helsen (&) Æ M. J. BuekersP. Feys Æ E. HeremansDepartment of Biomedical Kinesiology, Faculty of Kinesiologyand Rehabilitation Sciences, Motor Learning Laboratory,Katholieke Universiteit Leuven, Tervuursevest 101,3001 Leuven, BelgiumE-mail: [email protected]

A. Lavrysen (&)Faculty of Psychology and Educational Sciences,Center for Developmental Psychology,Katholieke Universiteit Leuven, Tiensevest 102,3000 Leuven, BelgiumE-mail: [email protected].: +32-16-326097Fax: +32-16-326144

D. ElliottDepartment of Kinesiology, McMaster University,Hamilton, ON, Canada

Exp Brain Res (2006) 174: 544–554DOI 10.1007/s00221-006-0484-9

deficits), while leaving visually guided reaching andaiming behaviors relatively intact (James et al. 2003).The reverse pattern of deficit is apparent following adorsal stream lesion (i.e., a double dissociation; Milneret al. 2003). The results of research designed to examinePAM with intact humans is less clear.

Most of the research on PAM with intact humans hasinvolved visual illusions such as the Ebbinghaus andMuller–Lyer illusions. This is because the model predictsthat perceptual–cognitive judgments associated withventral stream coding will be affected by the visualcontext associated with illusory configurations, whilegoal-directed movements will be immune to these ‘‘per-ceptual’’ biases. The notion is that action will be unaf-fected by illusory visual context because the dorsalstream codes visual–spatial information using an ego-centric or body-based frame of reference and not relativeto the action-irrelevant visual surround (Goodale andHaffenden 1998).

Although a number of studies have found thatgrasping and aiming movements are either unaffected orless affected by illusory context than perceptual judg-ments about the same stimuli (e.g., Aglioti et al. 1995;Haffenden and Goodale 1998), other researchers havereported that actions are not immune to illusory con-figurations (Elliott and Lee 1995; Franz et al. 2000;Franz 2001). Some of these discrepancies in the litera-ture appear to be due to the manner in which perceptualand motor biases are quantified (Franz et al. 2000).However, other differences appear to be due to thenature of the visual and nonvisual information availablefor limb control in a given experiment, and the extent towhich continuous, or at least recent, visual informationfrom the target and the limb is available for the on-linecontrol of action (e.g., Binsted and Elliott 1999b). Forexample, while Milner and Goodale (1995) have alwaystaken the position that the dorsal stream operates insomething close to real time, in an early version of PAMit was suggested that the more dependent ventral streamonly contributed to limb control if direct visual infor-mation from the limb and the target was eliminated fortwo or more seconds prior to movement initiation(Milner and Goodale 1995). Recently, Westwood andGoodale (2003) have revised the PAM position onmemory dependent limb control and have suggested thatthe complete dorsal-based control requires visual infor-mation from the limb and the target during the move-ment preparation period just prior to movementinitiation. This change in the model has helped reconcilethe findings of illusory bias in aiming and graspingexperiments in which vision is eliminated coincident withthe signal to begin the movement (i.e., at the beginningrather than the end of the reaction time interval).

Recently, Glover and Dixon (2001, 2002) and Glover(2002) have forwarded an alternative to the PAM thatattempts to explain illusory effects on action by makinga distinction between the types of visual coding impor-tant of movement planning and on-line control (see alsoWoodworth 1899; Elliott et al. 2001). Their Planning-

Control Model (PCM) posits that, while premovementmovement preparation processes are susceptible to theinfluence of visual illusions, on-line control processes areimmune to these same contextual influences. Thus,according to the PCM, the extent to which visual con-text affects movement bias will depend on the relativeimportance of movement planning processes and on-linecontrol for the particular type of action under consid-eration. For example, Glover (2004) holds that foraccuracy-constrained movements such as aiming to atarget of a fixed size (e.g., Fitts 1954), movement timeswill be planned prior to movement onset. Thus, thesemovement times will vary with the perceived rather thanthe real size of the target (e.g., van Donkelaar 1999). Inother situations, that require more on-line regulation,initial movement biases associated with movementplanning can be expected to diminish as the actualmovement unfolds (Glover 2004).

Although Glover’s (2004) model helps reconcile someof the discrepant findings in the literature, it is still atodds with several recent findings that indicate illusoryinformation available during movement planning andexecution have additive effects on final movement bias(e.g., Handlovsky et al. 2004; Mendoza et al. 2006).Once again, it appears that the impact of illusory con-figurations on limb control may have more to do withthe specific sources of information available to the par-ticipant in any given task (e.g., Binsted and Elliott1999b; de Grave et al. 2004) and/or the specific taskgoals (e.g., Smeets and Brenner 1999, 2001, 2004) than adichotomous set of visual coding processes (i.e., PAM orPCM). The goal of the present study was to examine therelative impact of different sources of visual as well asnonvisual sources of information on illusory bias. Thiswas done by independently manipulating visual infor-mation from the limb and the target during manualaiming to the vertices of Muller–Lyer configurations.Additionally, the contribution of extra-retinal informa-tion for the eyes was assessed. The contribution of eyemovements and eye position to illusory bias was deemedimportant because it is clear that the saccades to Muller–Lyer vertices are profoundly affected by configuration ofthe Muller–Lyer tails. Moreover, Binsted and colleagues(Binsted and Elliott 1999b; Binsted et al. 2001) havedemonstrated that eye movement biases have the po-tential to influence hand movements when precise visualinformation about the position of the hand and thetarget is not available.

To better understand the contribution of potentialcues, we need to go into detail about what informationfrom the hand, the target and the eyes is available innormal aiming. Firstly, retinal visual information fromthe target and feedback from the moving limb are em-ployed on-line to ensure the accuracy of manual goal-directed movements. Reasonable accuracy is, however,also possible in the absence of visual information fromthe limb or the target. In the first case, a calibration ofhand and eye position can take place based on veridicalaspects of the target (Prablanc et al. 1986). Visual

545

(retinal and eye position) information about the targetcan be compared to the position of the limb, while thelimb makes its way to the target. In the second case withconcealed targets, information about the target positionis not readily available and, thus, offline positionalinformation must be retrieved from memory (Elliott1990). This remembered target position can be com-pared with visual feedback from the ongoing limbmovement (e.g., Elliott and Madalena 1987).

Secondly, apart from the prevailing role of retinalvisual information from limb and target, there are othercues for planning and updating goal-directed move-ments. It has been demonstrated that limb movementsuse information from both static and dynamic charac-teristics of the oculomotor system (Abrams et al. 1990;Enright 1995). Moreover, the kinematics of open-looppointing movements are influenced by simultaneouslyproduced eye movements (van Donkelaar 1997, 1998;van Donkelaar and Staub 2000). Finally, vision of thestatic hand prior to the movement (Desmurget et al.1995) as well as of the moving limb (van Donkelaar1997) can provide valuable feedback for movementcorrections. In order to provide the movement controlsystems with the necessary information in a timelymanner, the eyes and hand must work together. Thestrong link between eye and hand movements, both intime and space, has received a lot of attention in recentstudies (see Starkes et al. 2002 and Bekkering and Sailer2002 for comprehensive overviews). As a result of thistight coupling, feedforward, (extra)retinal and proprio-ceptive information from both effectors and informationabout the position of the target can additionally be usedfor planning and control of actions. In this regard, it wassuggested that extra-retinal information might mediatethe size of illusion biases by providing additionalinformation about target location for on-line movementcorrection (see also Glover 2002).

Apart from their behavior in normal conditions, theexamination of eye movements and eye–hand coordi-nation with respect to visual illusions has received sub-stantial attention for more than a century. Although itremains undecided whether perceptual bias causes theperceived erroneous eye movements, or distorted eyemovements elicit perception bias (Coren 1986), this topichas strong potential to contribute to the understandingof mechanisms underlying the generation of illusorypercepts and their effects on movements. Typically, eyemovements mimic the distorted patterns of perception(e.g., Binsted and Elliott 1999a). Hand movements,however, often seem resistant to the effects of illusions.It can be argued that some of the results obtained on theresistance of limb action to perceptual illusions might infact be accounted for by veridical gaze signals. In thisregard, the comparison of saccade versus fixation con-ditions might be enlightening as to understand thecontribution of gaze signals to aiming performance.Recent technologies have enabled the simultaneousrecording of gaze position signals during experimentaltasks. In this regard, Soechting et al. (2001) made

combined measurements of eye and limb movementsexposed to a directional illusion induced by a movingbackground (i.e., the Duncker illusion). A strong cor-relation between the errors of gaze and manual pointingwas interpreted as if gaze position provided the targetsignal for limb movement. Bernardis et al. (2005) foundsmaller bias in pointing than in saccadic or perceptualjudgement tasks, and they argued that efference fromprimary saccade information was thus not used to drivepointing movements. However, they measured saccadesand pointing movements during separate trials. Perhapsthe most instructive work comes from Binsted and col-leagues (Binsted and Elliott 1999b; Binsted et al. 2001).They observed that gaze position was influenced,whereas manual movements were immune to the effectsof a Muller–Lyer illusion, except in trials with concur-rent eye movements and elimination of retinal targetinformation (Binsted and Elliott 1999b; Binsted et al.2001). Nevertheless, the strong correlation betweensaccadic and manual movements typically observed indiscrete aiming tasks (Helsen et al. 1998b) was notapparent in their task.

To summarize, the main goal of the present study wasto test whether the impact of a visual size illusion wasmediated by the type of visual information. Participantsmade continuous aiming movement to the ends of a linewith either inward or outward pointing Muller–Lyervertices. The type of information was manipulated byproviding incomplete vision of either hand or figure.Secondly, the involvement of extra-retinal informationwas studied by either allowing or restricting eye move-ments. Apart from studying the effects of a Muller–Lyerillusion on eye and hand movements, the temporalrelationship between eye and hand movements wasexamined to determine to what extent oculo-manualtiming was influenced by the illusion as well as toinvestigate the nature of a possible exchange of infor-mation between effector systems.

Method

Participants

Fifteen graduate and undergraduate students rangingfrom 20 to 32 years (mean age 23.2 years) from the K.U.Leuven took part in this study. Amongst them were 5males and 10 females. They all had normal or corrected-to-normal vision and were naive to the hypotheses beingtested. All participants were classified as strongly right-handed, as verified by an adapted version of Bryden’s(1977) handedness questionnaire (mean score =28.8 ± 3.3, where a maximum score of 30 representsextreme right hand preference). This study was con-ducted in accordance with the ethical standards laiddown in the 1964 Declaration of Helsinki, as has beenassessed and approved by the Committee for EthicalConsiderations in Human Experimentation of theKatholieke Universiteit Leuven.

546

Apparatus

Participants performed a cyclical aiming task, makingflexion and extension movements with their dominantright hand. They sat in an armchair with both armssupported by the chair arms in front of a 17 in. PCscreen. The movements were paced by an auditorymetronome at a frequency of 1 Hz, thus, requiring oneflexion and one extension movement every second.

Hand movement

The participants wore a right-hand orthosis that wasaligned to the anatomical axis of the wrist joint. Thefeatures of the orthosis restricted the hand movements ofthe participants in such a way that they could onlyperform flexion and extension in a horizontal plane.Wrist angular position was recorded by means of a non-ferromagnetic high-precision shaft encoder fixed to themovement axis of the orthosis with an accuracy of 0.09�.Direct vision of the forearm and hand was prevented bya cardboard panel. Instantaneous angular position waspresented as a cursor moving horizontally on a PCscreen in front of the participants (Fig. 1). A wristangular displacement of 40� was required to move acursor over a 20 cm horizontal line. Thus, although thiswas a computer-aiming task, very little spatial transla-tion was required. In fact, the task was very much likemoving the end of a hand held implement (e.g., a longstylus or a stick) back and forth between two targetpositions with wrist flexion–extension movements. Inthis case, the cursor behaved like the end of the imple-ment and responded to the wrist movements in real time.

Point of gaze (PG)

An Applied Sciences Laboratory head-mounted eyetracker (ASL-5000 model 501) was used to record theinstantaneous viewing point independent of headmovement or position (i.e., point of gaze or PG). Thiseye monitoring system is integrated with a head trackingsystem to measure the 3D head position while allowing

participants free head motion throughout the experi-ment. Eye and head coordinates were thus integrated,resulting in a 2D PG coordinate.

Stimulus presentation and sampling of both hand andeye movements was done with custom-made software ata frequency of 200 Hz, providing synchronized coordi-nates for wrist angular position and PG.

Task

The participants performed a unimanual cyclical aimingtask. The cyclical nature of the task used in the presentwork differentiates from previous work with pictorialillusions that has mainly focused on discrete movements.However, most every-day tasks involve cyclical move-ments. A recent imaging study (Schaal et al. 2004)showed in fact that discrete movement may in factsimply be a special case of rhythmic movement (Mialland Ivry 2004). Targets were the two ends of a 20 cmhorizontal line with either an inward or outwardpointing Muller–Lyer configuration (shaft length: 4 cm;fin angle with horizontal line: 45�). They were instructedto move the cursor from the left end of the line to theright and back in time with an external metronome(2 Hz). Participants were asked to follow as precisely aspossible the rhythm imposed by the auditory signal,aligning the reversal of the movement as accurately aspossible to the beep. Each trial consisted of 31 beeps andwas divided into two phases: a ‘viewing phase’ and a‘movement phase’1. Throughout the first 10 beeps (i.e.,viewing phase), participants were instructed only to lookat the presented figure. Then, the figure would disappearfor 1 s. During the remainder of the beeps, the partici-pants had to make hand movements according to thelength of the line (i.e., movement phase). The size andtype of the figure in the viewing and the movementphases was identical.

There were two illusion conditions (pointing-in ><and pointing-out <> Muller–Lyer configurations),three line sizes (small = 33�, medium = 40�, andlarge = 47�), two eye movement conditions (fixationand saccade) and two visual information conditions(cursor and figure). The Muller–Lyer configurationswere made up of different actual line lengths in an at-tempt to increase the perceptual complexity of judgingthe line sizes and, thus, to elicit perceptual bias. Eye-movement generation, and therefore the amount of(extra)retinal information, was varied by the instruc-tions. In the fixation condition, the participants wererequired to fixate a dot in the centre of the screen. Inthe saccade condition, they were asked to make aneye movement from one end of the line to the other intime with every hand movement. Finally, the type of

cardboard panel

armrests

Fig. 1 Test setup

1As movements were based on remembered target information inthe cursor condition, the presentation of the figure was essentialhere. That is, the viewing phase was used as to create an identicalpre-movement exposure for both vision conditions.

547

information was manipulated in such a way that, in bothconditions, the participants received incomplete butdifferent information. In the cursor condition, the figurehad to be recalled from memory, while in the figurecondition, there was no on-line visual reference for handposition.

Procedure

First, the participants were comfortably seated in anarmchair in front of a table, with their head about 50 cmaway from the PC screen. The right-hand orthosis wasfixed and both forearms placed in elbow-rests attachedto the chair. A cardboard panel was placed above theright forearm to prevent direct vision of the handmovement. Next, the encoder was calibrated and par-ticipants were presented with open-loop practice trials ofthe medium amplitude condition with a line withoutvertices. From then on, each trial consisted of theviewing and movement phases, as in the experiment it-self. Seven trials were given with vision of the line onlyand finally two trials with vision of the cursor only.After the practice trials, the ASL tracker was placed onthe participant’s head and calibrated. During theexperiment, participants were informed about the eyemovement condition that would follow. There was a30 s rest period between the randomized conditions andeach condition was presented twice.

Data analysis

Hand movement

As the hand movements were made in a continuousfashion, the hand movement time (MT) and amplitudewere measured between two consecutive reversal points(AMPr). In addition, the travelled distance at peakvelocity was measured (AMPpv). Within-subject vari-ability was measured at both time points in the handtrajectory (SD AMPpv and SD AMPr), as it can beinstructive with regard to the amount of feedback beingprocessed on-line (Khan et al. 2002). In order to evalu-ate the evolution of the illusion effect, the differencebetween pointing-in and pointing-out absolute ampli-tudes was evaluated at both time points (TTPV andreversal). In line with previous work (Starkes et al.2002), velocity data were obtained by differentiating thefiltered position data (1st order low-pass Butterworthfilter with a cut-off frequency of 20 Hz). Mean time topeak velocity was then calculated as a proportion ofhand MT (%TTPV). This variable provides an index ofthe temporal symmetry, with larger proportional timesspent after peak velocity being associated with a greaterdegree of on-line control (Elliott et al. 1999, 2001). Italso provides a reference point for evaluating eye–handtemporal coordination, as saccade arrival is linked withhand movement kinematics (Starkes et al. 2002).

Eye movement

The methods used to analyse the eye data were similar tothe ones described by Helsen et al. (1998b). An eyemovement or saccade is preceded and followed by astable PG or fixation. In the present experiment, thecriterion for this fixation was a standard deviation forPG smaller than 1� for minimum 100 ms. The start ofthe primary saccade was determined departing fromwhere gaze position crossed the screen midline, movingbackwards in time. The standard deviation of 20 con-secutive eye positions was calculated and the 20-pointtime frame was each time shifted one position back-wards in time, until the standard deviation was less than1�. The last of these 20 consecutive eye positions was thestart of the primary saccade. The end of the primarysaccade was determined in exactly the same way, as thefirst of 20 points with a standard deviation that first fellbelow 1�, this time moving forward in time. Saccadeamplitude was calculated between the start and the endof the primary saccade. As there were small differencesbetween the subjects’ distance to the screen, there wassome difference between the actual saccade amplitudewhen looking to the ends of the medium size line,varying between 20 and 23�. Therefore, and also forenabling rapid comparison with hand and saccademovement amplitude, the saccade amplitude was nor-malized to 40� for the medium line size condition.

Eye–hand coordination

The saccade termination time point was expressed as apercentage of hand movement time. Furthermore, theabsolute amount of time between the arrival of the eyesand TTPV was analysed to evaluate relative timing ofeye and hand movements.

Design

One global estimate for every variable in each conditionwas attained by averaging the values from all move-ments in this condition. Hand movement time andamplitude at movement reversal were analysed using atwo illusion (pointing-in and pointing-out), by three linesizes (small, medium, large), by two eye movementconditions (saccade and fixation), by two visual infor-mation (cursor and figure) repeated measures ANOVAdesign. The dynamic effect of the illusion (pointing-inminus pointing-out) on hand amplitude and amplitudevariability were examined using a two time point (PV,reversal), by three line sizes (small, medium, large), bytwo eye movement conditions (saccade and fixation), bytwo visual informations (cursor and figure) repeatedmeasures ANOVA design. For saccade amplitude andeye–hand coordination variables, each of the dependentvariables was analysed using a two illusion (pointing-inand pointing-out) by three line sizes (small, medium and

548

large), by two visual informations (cursor and figure)repeated measures ANOVA design. Two participantshad to be removed from the eye movement and the eye–hand coordination analyses, because they did notcorrectly follow the instructions for the various eyemovement conditions. Significant effects were identifiedthrough Tukey–Kramer post-hoc procedures. Becausewe had a number of dependent variables, all statisticaltests were completed with alpha set at P < 0.01.

Results

The eye movement by visual condition by line size cellmeans and between-subject standard deviations for handand eye amplitude are presented in Table 1.

Hand movement

No significant effects were found for hand movementtime (MT). On average, they made one movement every496 ± 10 ms, confirming that participants did obey thetask instructions to move to the beat.

The analysis for hand amplitude at peak velocity(AMPpv) revealed significant main effects for illusionF(1,14) = 75.76, P < 0.0001, eye movement conditionF(1,14) = 17.04, P = 0.001, and line size F(2,28) = 76.89,P < 0.0001, and a significant interaction effect for illu-sion by visual information F(1,14) = 9.4, P < 0.01. Theillusion effect showed that the distance travelled at themoment of hand peak velocity was shorter in thepointing-out (16 ± 3�) than in the pointing-in condition(19 ± 3�). The size of the horizontal line of the Muller–Lyer figures proportionally mediated the hand ampli-tude at time to peak velocity (Small: 16 ± 3�; medium:18 ± 3�; large: 19 ± 3�). The post-hoc analyses dem-onstrated that the differences between any of theseamplitudes was significant. When making saccades, thedistance travelled was shorter (17 ± 3�) than when fix-ating the central spot (18 ± 3�). Finally, the illusion by

visual information interaction showed that hand move-ments were more biased by the Muller–Lyer configura-tions in the condition where the hand position wasshown, as compared to the figure condition (see dottedbars in Fig. 2).

The hand amplitude at movement reversal (AMPr)analysis of variance revealed significant main effects forillusion F(1,14) = 105.53, P < 0.0001, eye movementcondition F(1,14) = 16.89, P < 0.01, visual InformationF(1,14) = 9.00 P < 0.01, and for line size F(2,28) =133.82, P < 0.0001. Furthermore, there were significantinteraction effects for illusion by visual informationF(1,14) = 16.95 P = 0.001, and visual information byline size F(2,28) = 6.79, P < 0.01. The main effect forillusion showed that participants overshot the averageamplitude of 40� in the pointing-in condition (43 ± 6�)while they moved too short in the pointing-out condition(37 ± 7�). The illusion by visual information interac-tion, as shown in the filled bars in Fig. 2, revealed thatthis difference was significantly greater when there wasvision of the cursor in the movement phase, compared tothe condition where only the figure was presented.

As for the visual information effect, the difference inamplitude between presenting the cursor or the figure inthe movement phase was significant, although in bothconditions the amplitude was close to the average targetamplitude of 40� (figure: 41 ± 7�; cursor: 39 ± 6�). Themain effect for eye movement condition showed thatparticipants made smaller movements in the saccadecondition (39 ± 7�), compared to the fixation condition(41 ± 7�). Smaller movements (37 ± 7�) were associ-ated with the small condition, larger ones (43 ± 6�) withthe large condition, and participants made 40 ± 6�movements in the medium condition. The interactioneffect for visual information by line size showed that thedifference between the three amplitudes was more pro-nounced in the cursor condition (small: 35 ± 6; med-ium: 39 ± 6; large: 43 ± 5) than in the figure condition(small: 38 ± 7; medium: 41 ± 7; large: 44 ± 7).

The evaluation of hand amplitude at two time points(PV and reversal) allowed for a dynamic evaluation of

Table 1 Eye and hand movement amplitude as a function of eyemovement, visual information and illusion condition

Eye movement Saccade Fixation

Visual information Figure Cursor Figure Cursor

Illusion >< <> >< <> >< <> >< <>

Hand amplitude at time to peak velocity (�)Mean 17.6 15.6 18.6 15.7 19.4 17.5 19.6 16.9SD 3.0 2.6 2.8 2.7 3.2 3.4 2.6 2.3Hand amplitude at reversal point (�)Mean 41.0 37.0 41.5 34.7 45.3 41.0 43.0 36.3SD 5.9 6.6 5.7 5.9 6.3 7.8 5.2 4.6Primary saccade amplitude (�)Mean 37.5 33.4 30.7 22.5 – – – –SD 5.6 5.8 6.9 6.7 – – – –

SD between-subject standard deviation

0

5

10

15

2025

30

35

4045

HA

ND

AM

PLIT

UD

E (

FIGURE CURSOR

At hand reversal

At peak velocity

Pointing-In ><

Pointing-Out <>

0

5

10

15

2025

30

35

4045

HA

ND

AM

PLIT

UD

E (

°)

FIGURE CURSOR

At hand reversal

At peak velocity

Pointing-In ><

Pointing-Out <>

Pointing-In ><

Pointing-Out <>

Fig. 2 The illusion effect on hand amplitude as a function of visualinformation at time to peak velocity (AMPpv; dotted bars) and atmovement reversal (AMPr; filled bars)

549

the illusion effect (difference between pointing-inand pointing-out conditions). This analysis revealedsignificant main effects for time point F(1,14) = 38.12,P < 0.001, and visual information F(1,14) = 20.67,P < 0.0005, and an interaction effect for eye movementcondition by visual information by line sizeF(2,28) = 6.86, P < 0.01. The absolute difference be-tween pointing-in and pointing-out conditions increasedfrom peak velocity (2 ± 3�) towards the point ofmovement reversal (5 ± 5�). This increase was signifi-cant for both the cursor t(14) = �7.85, P < 0.0001,and the figure conditions t(14) = �3.45, P < 0.01, asevaluated by paired t-tests. Overall, bias was smaller inthe figure (3 ± 5�) than in the cursor condition (5 ± 4�)for both time points (time to peak velocity:t(14) = �3.07, P < 0.01; reversal: t(14) = �4.12,P = 0.01). The three-way interaction involving eyemovement condition, visual information and line sizewas not relevant as to the scope of the present investi-gation and therefore will not be regarded into detailhere. More interestingly, the interaction effect for timepoint by visual information F(1,14) = 7.86, P = 0.014,approached conventional levels of significance. This ef-fect showed that the increase over time was more pro-nounced for the cursor than for the figure condition (seeFig. 3).

The within-subject variability was evaluated at timeto peak velocity and at movement reversal. The ANO-VA revealed significant main effects for time pointF(1,14) = 147.72, P < 0.001, visual informationF(1,14) = 17.57, P < 0.001, and line size F(2,28) = 6.06,P < 0.01. Amplitude variability decreased from peakvelocity (5.4 ± 1.4�) to movement reversal (2.9 ± 0.9�).The visual information effect revealed that variabilitywas greater in the figure (4.5 ± 1.7�) than in the cursorcondition (3.8 ± 1.7�). Finally, there was greater vari-ability in the large (4.4 ± 1.8�) as compared to the small(4 ± 1.6�) and medium (4 ± 1.7�) line size conditions.

No significant effects were found for time to peakvelocity (%TTPV). On average, peak velocity wasreached at 58 ± 7% of the hand movement time.

Eye movement

The primary saccade amplitude analysis revealed sig-nificant main effects for illusion F(1,12) = 95.81,P < 0.0001, visual information F(1,12) = 25.18,P < 0.01 and line size F(2,24) = 158.56, P < 0.0001,and a significant interaction effect was found for illusionby visual information F(1,12) = 16.61, P < 0.01. As wellas the hands, the eyes were deceived by the illusion, asamplitudes were larger in the pointing-in (34 ± 7�) thanin the pointing-out condition (28 ± 8�). Figure 4 showsthat the effect of the illusion was significantly greater inthe cursor condition, compared with the conditionwhere there was only visual information of the config-uration. The effect for visual information revealed largeramplitudes in the figure condition (35 ± 6�) comparedto the cursor condition (27 ± 8�). Logically, the effectfor line size showed that participants made smaller pri-mary saccades (26 ± 7�) in the small condition incomparison to the medium (31 ± 8�) and the largeconditions (36 ± 8�).

Eye–hand coordination

A typical example of the coordination of eye and handmovements is shown in Fig. 5. The primary saccadetermination as a proportion of hand movement time

Fig. 3 Hand amplitude as a function of illusion and visualinformation

0

5

10

15

20

25

30

35

PRIM

AR

Y S

AC

CA

DE

AM

PLIT

UD

E (

° )

FIGURE CURSOR

tinPointing-In ><

Pointing-Out <>

0

5

10

15

20

25

30

40

C

Fig. 4 Primary saccade amplitude as a function of illusion andvisual information

Fig. 5 A typical eye–hand coordination pattern showing the timingof eye and hand events as percentage of hand movement time.Characteristically, the saccade ends before hand peak velocity isreached

550

analysis yielded a significant interaction effect for illu-sion by visual information F(1,12) = 10.72, P < 0.01.The eyes arrived later in the pointing-in (47 ± 10%)than in the pointing-out condition (42 ± 10%) whenvision of the cursor was provided, and simultaneouslywhen the figure was presented (pointing-in: 45 ± 12%;pointing-out: 44 ± 11%).

No significant effect was found between the experi-mental conditions for the time between the primarysaccade end and the time to peak velocity of the hand.The eyes arrived at the target 68 ± 52 ms before thehand reached peak velocity.

Discussion

In the present cyclical aiming experiment, the type ofinformation (retinal and extra-retinal) was manipulatedin order to investigate the differential contributions ofdifferent cues on both eye and hand movements. Theresults confirmed that both eye and hand movementamplitudes were mediated by the line sizes. But mostimportantly, the movements of both effectors were sig-nificantly smaller when aiming to the outward pointingconfiguration as compared to the inward pointing ver-tices.

In both visual information conditions when either thefigure or the cursor was presented, there was incompleteinformation regarding limb and target position.According to the PAM based on the ventral–dorsalstream dichotomy, perception would be biased in bothsituations, but illusion effects on action should exhibitonly in absence of accurate visual information from thefigure (Milner and Goodale 1995). Although a mea-surement of perceptual judgement could be instructiveas to better understanding the mechanisms of the dis-sociation between vision for perception and for action,the scope of the present work was to evaluate the effectsof illusion induced biases on action. On the one hand, inthe figure condition, movements would be executedbased only on veridical information travelling throughthe dorsal stream and the action-irrelevant vertices ofthe figures would be ignored. Thus, no bias would beexhibited. On the other hand, in the condition where thecursor was presented, the memory-based representation(ventral stream) would cause greater differences in mo-tor outputs between the pointing-in and pointing-outconditions. So, bias was indeed present in the cursorcondition, but also in the figure condition. Therefore,the results are only partially in agreement with the tenetsof the PAM.2

As to the predictions of another dichotomous model,the PCM, bias would be corrected on-line when there

was opportunity for correction. This would materializeas a decrease of bias throughout the movement (i.e.,dynamic illusion effect; Glover 2002). The adaptation oftime to peak velocity to the presented information cangive a first indication of the occurrence of feedbackbased corrections (Elliott et al. 1999). However, thisanalysis did not yield significant results. Furthermore,amplitude variability was measured at different timepoints in order to evaluate the evolution of this measurethrough time (Khan et al. 2002). Generally, at time topeak velocity, hand movement variability was greaterwith larger hand amplitudes, as a result of both illusoryeffects and actual line size differences. Most importantly,the variability data showed a reduction from peakvelocity to movement reversal, irrespective of experi-mental condition. The PCM states that the illusion ef-fects after delays of more than 2 s reflect the influence ofplanning only, as control is ineffective in reducing theimpact of the illusion after delays. However, in ourexperiment the movement trials lasted far longer than2 s and the reduction in variability from peak velocity tomovement reversal suggests that participants did notsimply unfold preplanned movements, but engaged insome amount of on-line feedback processing (Elliottet al. 1999; Khan et al. 2002). Therefore, there was anoccasion for on-line corrections to diminish biasing er-ror towards the end of the movement, as predicted bythe PCM (Glover and Dixon 2001). On the contrary, thepresent results showed that the absolute difference be-tween pointing-in and pointing-out conditions even in-creased from peak velocity to the point of movementreversal (see also Heath et al. 2004; Meegan et al. 2004;Mendoza et al. 2006). When cursor and figure conditionsare regarded separately, the increase in difference be-tween both configurations was less pronounced for thefigure condition than for the cursor condition. Thus,with on-line target information available, a smallamount of bias correction was perceived. The possibilityfor movement corrections together with the increasingeffect of the illusion from peak velocity towards the endof the movement pose difficulties for the PCM.

Overall, the present data suggest that the type ofinformation available is important in planning andupdating the ongoing movement. It provides evidencefor the versatility and adaptability of the visuomotorsystem to specific task demands, allowing the participantto adopt different strategies according to the specificcues at hand (Mendoza et al. 2006). Perhaps a possibleexplanation of the difference in size between the illusioneffects in the figure and cursor conditions lies partlywithin some sort of exchange between ventral and dorsalstream processes. In the figure condition, the partici-pants had access to two different types of information.On the one hand, accurate on-line retinal informationabout target position was available to the dorsal streamfor parietal processing, whereas perceptual bias waselicited by ventral-stream processing of the Muller–Lyerconfiguration. Thus, a conflict between dorsal and ven-tral stream information might have caused a part of the

2Although every attempt was made to make our computer-aimingprotocol as direct and egocentric as possible (e.g., our aiming witha stick analogy), inconsistencies between our results and the pre-dictions of PAM could stem from translation processes inherent inour task.

551

biased information from the ventral stream to betransferred and implemented in the limb movementplan. As there was no confounding information avail-able from the dorsal stream in the cursor condition, theillusory interpretation was possibly brought out to thefullest here. Interactions between ventral and dorsalstreams might occur through direct communication be-tween both streams, which might be possible because ofthe extensive anatomical interconnections that havebeen shown to exist in monkeys (Merigan and Maunsell1993). Alternatively, recent work suggests that the ven-tral stream might influence pointing by bypassing thedorsal stream via the prefrontal cortex (Lee and vanDonkelaar 2002).

It has been previously demonstrated that saccadicinformation typically contributes to hand control. Inthis regard, hand movements have been found to berespectively greater or smaller, in line with the saccadeamplitude of the corresponding eye movement (vanDonkelaar 1997, 1998). Further, the second feedback-based component of an aiming movement is believed toprovide opportunity for on-line correction processes (seeElliott et al. 2001 for a review), even in the absence ofon-line vision (Goodale et al. 1986). In the presentexperiment, peak velocity was reached at 58% of thetotal hand movement time and the primary saccadeswere completed on average 68 ms before this event. Forthese reasons, it is argued that sufficient time andopportunity was available for the biased saccade infor-mation to be implemented in the hand control andcorrection processes shortly after peak velocity (Carlton1981). The results showed that, even when the figure waspresented during the movement phase, gaze positioncould not be accurately positioned towards the end ofthe lines and was moreover biased by the Muller–Lyerillusion. Therefore, the illusory information from thesaccades potentially could have contributed to handmovement inaccuracy. As there was indeed also an im-pact of the illusion on hand amplitude, at first sight, thepresent results are congruent to the notion that eyemovement information holds the capacity of influencingmanual movements. However, based on the fact thatsimilar illusory effects on hand movement exhibited inthe fixation condition, the perturbed eye movementswere perhaps not the only or main cause of the handillusion effects (Bernardis et al. 2005). Perhaps, not thesaccades per se, but rather the erroneous retinal infor-mation of the figure, available from the initial viewingphase, might have caused the perturbations. Interest-ingly, the results confirmed that the illusory bias on botheye and hand movements were similarly mediatedaccording to the type of information that was presentduring the movement phase. This would entail that thesame information was indeed used for the planning ofboth eye and hand movements (Prablanc et al. 1979;Gielen et al. 1984).

There was no significant influence of the illusoryeffects on eye–hand coordination, as expressed by theresults of the timing pattern for eye and hand termination.

As our group has suggested elsewhere (Helsen et al.1998a), the eye–hand coordination may be more tied tothe movement endpoint (i.e., the target) than to move-ment initiation. Therefore, it is argued that in particularthe termination pattern of eye and hand is of impor-tance, as the moment when the eyes are focused near thetargets creates the moment from when the feedbackloops and, thus, hand movement correction processescan be initiated. Overall, in the present study, primarysaccade termination was early enough for the hand touse this information for on-line guidance of the move-ment (Carlton 1981). Interestingly, the saccade termina-tion pattern was independent of the type of informationpresented. There was only a difference between the twoMuller–Lyer configurations in saccade terminationtiming when the cursor was presented. This interactioneffect of illusion by visual information occurred also inthe analyses of hand movement amplitude, saccadeduration and amplitude. This indicates that, when par-ticipants saw their hand position in the cursor condition,the hand and eye dynamics was altered in terms of theperceived difference in line size. Specifically, there was agreater impact of the illusion on both the hand and eyemovements in the cursor condition. Furthermore,shorter saccades understandably resulted in shortersaccade durations. As a result of these altered dynamics,also the termination pattern of both eye and hand wasearlier with the shorter movements. Interestingly, how-ever, the coordination pattern between saccades andhand movements was unaltered regardless of the changeddynamics. As shown by the analysis of time betweensaccade termination and hand time to peak velocity, thesaccades arrived consistently before the hand reachedpeak velocity. In conclusion, although the present resultsdid not show any positive or negative effect from makingconcurrent saccades in resisting to the illusion, the illu-sions did not alter nor affect the eye–hand coordinationpattern significantly. The timing of saccadic arrivalwas consistent and strongly linked to hand movementkinematics.

The temporal coordination pattern and spatial simi-larities between eye and hand movements seems toindicate that already the planning stages of both effec-tors may have been susceptive to and influenced by theperceived and not the actual size of the movement. Boththe spatial and temporal results provide arguments thateye and hand movements were generated based oncommon commands or at least common information.Sharing one programme for both eye and hand move-ments would be an efficient approach for reducing theamount of processing needed before movement start,especially under time pressure circumstances as in thepresent fast aiming task. In the literature, a distinction ismade between a common command for movement ini-tiation (Bizzi et al. 1971) and a shared spatial represen-tation of the target (Prablanc et al. 1979; Gielen et al.1984). In the present experiment, both saccades and limbmovements showed similar bias, thus it rather seems thatthe representation of the target was shared.

552

To summarize, the effect of a Muller–Lyer illusionwas similar on both eye and hand movement amplitudeand this bias was mediated by the type of informationavailable on-line. Additionally, the illusion effect in-creased consistently from peak velocity until movementreversal. Interestingly, saccadic information did notinfluence the size of the effect of a Muller–Lyer illusionon hand movements. Furthermore, the illusions did notalter the eye–hand coordination pattern. The timing ofsaccade completion was consistent and strongly linkedto hand movement kinematics. At the end, the presentresults cannot be reconciled fully with strict dissocia-tion theories between action/perception (PAM) andplanning/control (PCM). Alternatively, they suggestthat the type of information available for movementplanning and perhaps the conflicts and interactionsbetween the available information mediate the size ofthe illusory effects. Overall, it has been demonstratedthat movement planning and control processes areversatile operations, which have the ability to adapt todifferent situations.

Acknowledgments Werner Helsen and Ann Lavrysen acknowledgethe K.U. Leuven Research Council for the support of this researchproject. Peter Feys is post-doctoral fellow of the Fund for ScientificResearch-Flanders, Belgium. The authors wish to thank Ir. MarcBeirinckx and Ir. Paul Meugens for providing invaluable guidancein designing the research equipment and the electronics.

References

Abrams RA, Meyer DE, Kornblum S (1990) Eye–hand coordina-tion: oculomotor control in rapid aimed limb movements. J ExpPsychol Hum Percept Perform 16:248–267

Aglioti S, DeSouza JF, Goodale MA (1995) Size-contrast illusionsdeceive the eye but not the hand. Curr Biol 5:679–685

Bekkering H, Sailer U (2002) Commentary: coordination of eyeand hand in time and space. Prog Brain Res 140:365–373

Bernardis P, Knox P, Bruno N (2005) How does action resist visualillusion? Uncorrected oculomotor information does not ac-count for accurate pointing in peripersonal space. Exp BrainRes 162:133–144

Binsted G, Elliott D (1999a) The Muller–Lyer illusion as a per-turbation of the saccadic system. Hum Mov Sci 18:103–117

Binsted G, Elliott D (1999b) Ocular perturbations and retinal/extraretinal information: the coordination of saccadic andmanual movements. Exp Brain Res 127:193–206

Binsted G, Chua R, Helsen W, Elliott D (2001) Eye–hand coor-dination in goal-directed aiming. Hum Mov Sci 20:563–585

Bizzi E, Kalil RE, Tagliasco V (1971) Eye–head coordination inmonkeys: evidence for centrally patterned organization. Science173:452–454

Bryden MP (1977) Measuring handedness with questionnaires.Neuropsychologia 15:617–624

Carlton LG (1981) Processing visual feedback information formovement control. J Exp Psychol Hum Percept Perform7:1019–1030

Coren S (1986) An efferent component in the visual perception ofdirection and extent. Psychol Rev 93:391–410

van Donkelaar P (1997) Eye–hand interactions during goal-direc-ted pointing movements. Neuroreport 8:2139–2142

van Donkelaar P (1998) Saccade amplitude influences pointingmovement kinematics. Neuroreport 9:2015–2018

van Donkelaar P (1999) Pointing movements are affected by size-contrast illusions. Exp Brain Res 125:517–520

van Donkelaar P, Staub J (2000) Eye–hand coordination to visualversus remembered targets. Exp Brain Res 133:414–418

Desmurget M, Rossetti Y, Prablanc C, Stelmach GE, Jeannerod M(1995) Representation of hand position prior to movement andmotor variability. Can J Physiol Pharmacol 73:262–272

Elliott D (1990) Intermittent visual pickup and goal directedmovement: a review. Hum Mov Sci 9:531–548

Elliott D, Lee TD (1995) The role of target information on manual-aiming bias. Psychol Res 58:2–9

Elliott D, Madalena J (1987) The influence of premovement vi-sual information on manual aiming. Q J Exp Psychol A39:541–559

Elliott D, Binsted G, Heath M (1999) The control of goal-directedlimb movements: correcting errors in the trajectory. Hum MovSci 18:121–136

Elliott D, Helsen WF, Chua R (2001) A century later: Wood-worth’s (1899) two-component model of goal-directed aiming.Psychol Bull 127:342–357

Enright JT (1995) The non-visual impact of eye orientation on eye–hand coordination. Vision Res 35:1611–1618

Fitts PM (1954) The information capacity of the human motorsystem in controlling the amplitude of movement. J Exp Psy-chol 47:381–391

Franz VH (2001) Action does not resist visual illusions. TrendsCogn Sci 5:457–459

Franz VH, Gegenfurtner KR, Bulthoff HH, Fahle M (2000)Grasping visual illusions: no evidence for a dissociation be-tween perception and action. Psychol Sci 11:20–25

Gielen CC, van den Heuvel PJ, van Gisbergen JA (1984) Coordi-nation of fast eye and arm movements in a tracking task. ExpBrain Res 56:154–161

Glover S (2002) Visual illusions affect planning but not control.Trends Cogn Sci 6:288–292

Glover S (2004) Separate visual representations in the planning andcontrol of action (discussion 24–78). Behav Brain Sci 27:3–24

Glover S, Dixon P (2001) Motor adaptation to an optical illusion.Exp Brain Res 137:254–258

Glover S, Dixon P (2002) Dynamic effects of the Ebbinghausillusion in grasping: support for a planning/control model ofaction. Percept Psychophys 64:266–278

Goodale MA, Haffenden A (1998) Frames of reference for per-ception and action in the human visual system. Neurosci Bio-behav Rev 22:161–172

Goodale MA, Milner AD (1992) Separate visual pathways forperception and action. Trends Neurosci 15:20–25

Goodale MA, Pelisson D, Prablanc C (1986) Large adjustments invisually guided reaching do not depend on vision of the hand orperception of target displacement. Nature 320:748–750

de Grave DD, Brenner E, Smeets JB (2004) Illusions as a tool tostudy the coding of pointing movements. Exp Brain Res155:56–62

Haffenden AM, Goodale MA (1998) The effect of pictorial illusionon prehension and perception. J Cogn Neurosci 10:122–136

Handlovsky I, Hansen S, Lee TD, Elliott D (2004) The Ebbinghausillusion affects on-line movement control. Neurosci Lett366:308–311

Heath M, Rival C, Binsted G (2004) Can the motor system resolvea premovement bias in grip aperture? Online analysis ofgrasping the Muller–Lyer illusion. Exp Brain Res 158:378–384

Helsen W, Starkes JL, Elliott D, Ricker K (1998a) Sampling fre-quency and the study of eye–hand coordination in aiming.Behav Res Meth Ins C 30:617–623

Helsen W, Starkes JL, Elliott D, Ricker K (1998b) Temporal andspatial coupling of point of gaze and hand movements in aim-ing. J Mot Behav 30:249–259

James TW, Culham J, Humphrey GK, Milner AD, Goodale MA(2003) Ventral occipital lesions impair object recognition butnot object-directed grasping: an fMRI study. Brain 126:2463–2475

Khan MA, Elliott D, Chua R, Lyons J (2002) Optimal controlstrategies under different feedback schedules: kinematic evi-dence. J Mot Behav 34:45–57

553

Lee JH, van Donkelaar P (2002) Dorsal and ventral visual streamcontributions to perception–action interactions during point-ing. Exp Brain Res 143:440–446

Meegan DV, Glazebrook CM, Dhillon VP, Tremblay L, WelshTN, Elliott D (2004) The Muller–Lyer illusion affects theplanning and control of manual aiming movements. Exp BrainRes 155: 37–47

Mendoza J, Hansen S, Glazebrook CM, Keetch KM, Elliott D(2006) Visual illusions affect both movement planning and on-line control: a multiple cue position on bias and goal-directedaction. Hum Mov Sci DOI 10.1016/j.humov.2005.09.002 (inpress)

Merigan WH, Maunsell JH (1993) How parallel are the primatevisual pathways? Annu Rev Neurosci 16:369–402

Miall CR, Ivry R (2004) Moving to a different beat. Nat Neurosci7:1025–1026

Milner AD, Goodale MA (1995) The visual brain in action. OxfordUniversity Press, Oxford

Milner AD, Dijkerman HC, McIntosh RD, Rossetti Y, Pisella L(2003) Delayed reaching and grasping in patients with opticataxia. Prog Brain Res 142:225–242

Prablanc C, Echallier JF, Komilis E, Jeannerod M (1979) Optimalresponse of eye and hand motor systems in pointing at a visualtarget. I. Spatio-temporal characteristics of eye and handmovements and their relationships when varying the amount ofvisual information. Biol Cybern 35:113–124

Prablanc C, Pelisson D, Goodale MA (1986) Visual control ofreaching movements without vision of the limb. I. Role ofretinal feedback of target position in guiding the hand. ExpBrain Res 62:293–302

Schaal S, Sternad D, Osu R, Kawato M (2004) Rhythmic armmovement is not discrete. Nat Neurosci 7:1137–1144

Smeets JB, Brenner E (1999) A new view of grasping. MotorControl 3:237–271

Smeets JB, Brenner E (2001) Action beyond our grasp. TrendsCogn Sci 7:287

Smeets JB, Brenner E (2004) Curved movement paths andthe Hering illusion: positions or directions? Vis Cogn 11:255–274

Soechting JF, Engel KC, Flanders M (2001) The Duncker illusionand eye–hand coordination. J Neurophysiol 85:843–854

Starkes J, Helsen W, Elliott D (2002) A menage a trois: the eye, thehand and on-line processing. J Neurophysiol 20:217–224

Ungerleider LG, Mishkin M (1982) Two cortical visual systems. In:Ingle DJ, Goodale MA, Mansfield RJW (eds) Analysis of visualbehavior. MIT, Cambridge, MA, pp 549–586

Westwood DA, Goodale MA (2003) Perceptual illusion and thereal-time control of action. Spat Vis 16:243–254

Woodworth RS (1899) The accuracy of voluntary movement.Psychol Rev 3:1–119

554