Consciousness

www.elsevier.com/locate/concog

Consciousness and Cognition 15 (2006) 269–284

andCognition

Dissociating size representation for action and for consciousjudgment: Grasping visual illusions without apparent obstacles

Elisabeth Stottinger, Josef Perner *

Department of Psychology, University of Salzburg, Hellbrunnerstr. 34, A-5020 Salzburg, Austria

Received 12 May 2005Available online 9 September 2005

Abstract

Visual illusions provide important evidence for the co-existence of unconscious and conscious represen-tations. Objects surrounded by other figures (e.g., a disc surrounded by smaller or larger rings, Ebbinghaus/Titchener illusion) are consciously perceived as different in size, while the visuo-motor system supposedlyuses an unconscious representation of the discs� true size for grip size scaling. Recent evidence suggestsother factors than represented size, e.g., surrounding rings conceived as obstacles, affect grip size. Use ofthe diagonal illusion avoids visual obstacles in the path of the reaching hand. Results support the dual rep-resentation theory. Grip size scaling follows actual size independent of illusory effects, which clearly biasconscious perception in direct comparisons of lengths (Experiment 1) and in finger-thumb span indicationsof perceived length (Experiment 2).� 2005 Elsevier Inc. All rights reserved.

Keywords: Visual illusions; Perceptual judgement; Knowledge for action; Implicit knowledge; Conscious perception;Obstacles avoidance

1. Introduction

The theory that it is possible to have visually gained knowledge without conscious awarenesshas been greatly helped by the finding from blindsight patients (e.g., Perenin & Rossetti, 1996;

1053-8100/$ - see front matter � 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.concog.2005.07.004

* Corresponding author.E-mail address: Josef.perner@sbg.ac.at (J. Perner).

270 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

Poppel, Held, & Frost, 1973; Weiskrantz, Sanders, & Marshall, 1974; Weiskrantz, 1998). Despitecomplete lack of conscious perception of stimuli in the �blind� part of their visual field they can,nevertheless, fairly accurately point to the stimulus or accurately act in other ways in relation to it.Support for the existence of similar dissociations in neurologically healthy people comes fromexperiments with illusory stimuli. Early demonstrations relied on motion illusions using the in-duced Roelofs Effect investigating eye gaze (Wong & Mack, 1981) and manual action (Bridgeman,Gemmer, Forsman, & Huemer, 2000; Bridgeman, Kirch, & Sperling, 1981). More recently, Agl-ioti, DeSouza, and Goodale (1995) used the Ebbinghaus/Titchener circles illusion to show a cor-responding dissociation. When asked to indicate which of two discs of the same size was larger,the one surrounded by smaller rings was judged larger than the one surrounded by larger rings—in keeping with the known effect of this illusion. However, the grip size scaling (thumb fingeropening on the way to grasping a disc) was about the same, indicating that the sensory-motor sys-tem allegedly uses a different—more accurate—representation (knowledge) of the discs� true sizesthan conscious perception. This interpretation of the original finding by Aglioti et al. (1995) has,however, come in for much criticism.

We concentrate here on the question whether the results of this experiment or any of its latervariations requires for their explanation the existence of two contradictory representations of real-ity (dual representations model)—one of which is available to conscious awareness and the otherone is not—or whether they can be explained on the basis of a single, conscious representation(single representation model). For answering this question the following aspects of the experimen-tal setup and model of relevant mental processes are critical. In the basic setup a particular stim-ulus object is presented and one of its properties (e.g., its size) is assessed by two different responsemodes (e.g., a verbal judgment and a grasping action) which yield contradictory response modeinformation about the stimulus property. The dual representations model assumes that the con-tradictory information in the different response modes comes from different representations. How-ever, the single representation model can also explain this contradictory response modeinformation if one of the following two possibilities arises:

1. The mapping from stimulus to internal representation is not constant over the two tasks inwhich each response mode is used. Pavani, Boscagli, Benvenuti, Rabuffetti, and Farne (1999)and, in particular, Franz, Gegenfurtner, Bulthoff, and Fahle (2000) have capitalized on thispossibility by observing that the size of the illusory effect is much stronger when the two discsin each display, one disc surrounded by smaller the other disc by larger rings, are directlycompared than when a disc in one of these displays is compared to a plain disc (withoutany surrounding rings). Participants in Agliotti�s experiment had to judge size by comparingboth displays, resulting in relatively large differences between represented and actual size. Incontrast, participants had to grasp only one of the two discs, which might have restrictedtheir attention to only that particular disc, attenuating the illusory effect and, consequently,resulting in seemingly more accurate grip size differences. Hence, on each occasion only a sin-gle representation of each disc�s size might be formed, but due to the attentional differences(attend to both displays for judgment but only to one when grasping) that representation ismore accurate in the grasping than in the judgment task. Consequently, the single represen-tation theory can adequately explain the difference in manually and verbally expressed size.

2. Extraneous reasons for acting may bias the action response. When asked to give a consciousjudgment of the size of a disc then the best estimate available is the size as internally represented.

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 271

In contrast, when grasping the disc there may be different valid reasons for a particular finger-thumb span (Jacob & Jeannerod, 2003; Smeets & Brenner, 1999). The actual size of the disc willonly be one of them. Another good reason is the space between the disc and surrounding obsta-cles. If the disc is presented on its own, one can afford a fairly large maximum grip span thatleisurely narrows in on the true size of the disc as the hand approaches. If, however, there is bare-ly space to fit one�s fingers between the disc and surrounding objects then one will adjust one�sgrip size to a minimum above the size of the disc. In fact, Haffenden and Goodale (2000) andHaffenden, Schiff, and Goodale (2001) found that the visuo-motor system seems to make exact-ly such adjustments—even when the surrounding obstacles are only two-dimensional marks onthe supporting surface (and strictly speaking are not real obstacles). Haffenden and Goodale(2000) used an additional ‘‘new’’ Ebbinghaus figure where the gap between target and sur-rounding small circles was the same as between target and surrounding large circles. They foundthat the critical variable for grasping was not the perceived size of the target but the distancebetween target and context circles.

However, Franz, Bulthoff, and Fahle (2003) did a similar experiment with another additionalEbbinghaus figure (the same gap between target and surrounding large circles as between targetand surrounding small circles). They found a systematically increasing grip aperture with perceivedsize—independent of the gap between target and context circles. In their experiment participantshad to grasp the target circle of the four context conditions with their dominant hand. As soonas participants started their grasping movements, shutter glasses suppressed their vision preventingany kind of visual feedback throughout the entire movement. Participants had four seconds to car-ry out this movement, which creates a problem of interpretation pointed out by Millner and Goo-dale (1995): the unconscious representation for guiding movements does not persist for longer than2 s. If the delay between seeing an object and movement execution toward this object is longer than2 s, movements are based on the longer lasting conscious perceptual representation, as found byGoodale, Jakobson, and Keillor (1994) with the visual form agnosia patient D.F. See also Bridg-eman, Peery, and Anand (1997), Marcel (1993). Hence, if participants in the experiment by Franzet al. (2003) did take close to 2 s for their movement with shutter glasses, it is fairly likely that theseauthors found an influence of the illusion on grasping because the motor representation had de-cayed and participants were forced to base their grasping on the illusory perceptual representation.

In any case, the Ebbinghaus Illusion is an inadequate method for deciding whether grasp aper-tures depends on the size of the objects to be grasped, because the object is surrounded by contextcircles (obstacles) and the gap between target and small and large context circles can never bematched completely. It remains an irresolvable question as to how to control the distance betweentarget disc and context circles: is it to be controlled at the narrowest point or in terms of the areabetween two context circles and the target object, etc.

Other illusory stimuli that have been used in this area suffer from similar problems. Severalstudies (for example Daprati & Gentilucci, 1997; Otto-de Haart, Carey, & Milne, 1999) usedthe Muller-Lyer illusion, where a stick was to be grasped that was made to look longer or shorterby placing it on arrow shaped background markings that were either inward or outward pointing(as in the standard Muller-Lyer illusion). One side effect of these background markings is that thewings of the background arrows may have unpredictable effects on the visuo-motor system. Onepossibility is that the wings of the inward pointing arrows (which make the stick look longer)make the system try to place the fingers cautiously within the V-shaped wings at the end. This

272 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

may reduce the maximum grip size and give the wrong impression that grasping is based on amore accurate representation.

Ironically, but we just cannot tell, the background markings might also have the opposite effect.Since the wings of inward pointing arrows extend beyond the ends of the stick they may give thevisuo-motor system the signal to provide for a potentially larger object to grasp (not the stick itselfbut the entire object of stick and wings). This may increase grip size from what it would be if thestick were presented alone. In this case, the danger is not that the background markings producethe impression that grasping is based on a more accurate internal representation of the stick�slength but that they cover up the existence of such an unconscious representation and make usthink grasping is subject to the illusion when it is not. Indeed, many of these studies report sig-nificant grip size differences in line with the illusory effects. In fact, Franz, Fahle, Bulthoff, andGegenfurtner (2001) report stronger illusion effects on grasping the Muller-Lyer illusion thanon perception. Again, if these extraneous reasons for adjusting grip size are not controlled for,we can conclude very little from these data about the existence of different representations or justa single representation for different response modalities.

Several studies (for example Jackson & Shaw, 2000; Westwood & Dubrowski, 2000) usedgrasping in conjunction with the Ponzo illusion, where same length objects in their vertical exten-sion were positioned at different points along four slanted horizontal lines that converged towardsthe centre giving the visual impression of four picture rails along a receding wall. This makes theobject look longer when positioned where the lines are closer together (same retinal image project-ed from something seemingly more distant) than when positioned where the lines are farther apart(same retinal image projected from something seemingly closer in space). Unfortunately, the con-vergence of the lines also means that, depending on where the end of the object happens to be inrelation to one of the lines, the line may create the tendency to narrow the grip to make it go be-tween the line and the object. Depending on whether it happens to affect the more ‘‘distant’’ objector the ‘‘closer’’ object than it works to either enhance or, respectively, diminish the effect of theillusion in a rather uncontrolled way. Similar concerns also apply to the parallel lines illusion usedby Franz et al. (2001).

The important lesson from these deliberations is that, unless we can neutralise such extraneousreasons (like potential obstacles) for adjusting one�s action parameters (e.g., maximum grip size),then these parameters tell us little about the existence of a dual (unconscious) representation.

To avoid the three problems outlined above we used new material and procedures. To avoid (atleast minimize) the interference of extraneous reasons for adjusting action parameters we used thediagonal illusion as shown in Fig. 1, where the diagonal of the larger parallelogram (left) is actu-ally shorter than the diagonal of the smaller parallelogram, but is consistently judged as longer.Participants are asked to grasp each red diagonal (as if it were a three-dimensional object) with

Fig. 1. Stimuli for the ‘‘against illusion’’ condition of Experiment 1. In the experiment the diagonals were printed inred.

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 273

their respective hand. Important for our efforts to limit the effect of extraneous reasons on gripsize is the fact that there are no obstructing background contours at either end of the diagonal,which could influence grip scaling. Of course, there are differences. For instance, the corner ofthe parallelogram over which the hand has to grasp is larger for the shorter diagonal (on left) thanfor the longer diagonal (on right). Similarly, the entire object surrounding the shorter diagonal islarger than the object surrounding the longer diagonal. Both these features could plausibly lead togreater grip sizes. However, if such an influence exists it would work in the direction of the illu-sion, i.e., that the shorter diagonal that looks longer would yield a larger grip size than the longerdiagonal that looks shorter. Hence, if we find grip size differences in line with the diagonals� truelength then this cannot be attributed to these features of the surrounding figure.

To counter the critique by Franz et al. (2000) of the original study by Aglioti et al. (1995) thatresponse mode (judgment versus action) is confounded with paying attention to both displays ver-sus only one display, we asked volunteers not only to compare the length of the two diagonals fortheir conscious judgment but also grasp both diagonals simultaneously.

To minimize arguments about the scaling of illusion we manipulated our stimuli so that wewould get a cross-over interaction between judgments of relative length and the difference betweengrip sizes. For that reason we made the diagonal of the larger parallelogram in the critical testcondition (‘‘length differences against the illusion’’) 3 mm shorter than the other diagonal, which,nevertheless, left the shorter diagonal as being judged longer.

To ensure that grasping is only based on the motor representation we avoid any delay betweenseeing and grasping the object. We want the participants to grasp the objects at a natural speedunder as normal perceptual conditions as possible. We should also point out that Vishton, Rea,Cutting, and Nunez (1999) used two-dimensional figures in their grasping condition before. Par-ticipants had no problems grasping these two-dimensional lines as if picking up a very thin object.Furthermore, as Kwok and Braddick (2003) found, grasping kinematics (grip aperture and veloc-ity) were similar for two- and three-dimensional objects. The only difference was that the meangrip apertures for two-dimensional objects were smaller than for three-dimensional objects. More-over, two-dimensional stimuli have certain advantages, e.g., they do not provide tactile feedback,which could lead to learned adjustment to the true size in the course of repeated grasping.

In sum, we anticipate that our critical condition will show that the shorter diagonal of the largerparallelogram will be judged longer than the actually longer diagonal embedded in the smallerparallelogram, but that grip sizes will significantly be smaller for the shorter diagonal than forthe longer one. In that case, we will have the most convincing evidence to date, that consciousjudgment and visuo-motor action use different representations of stimulus properties.

2. Experiment 1

2.1. Method

2.1.1. ParticipantsSixteen volunteers were tested but due to technical problems the data of two persons could not

be used. An additional person showed the strange ‘‘grasping’’ behaviour of not opening her gripat all except just before the target. Hence, there were no useable grip sizes at the point of measure-

274 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

ment. The final sample, therefore, comprised 6 women and 7 men with a mean age of 24.9(SD = 1.80) years. All of them but one said to be right handed (one dominant left handed and12 dominant right handed participants).

2.1.2. Apparatus and procedureVolunteers were seated with their hands on pads on a table underneath a cover (33.8 · 14.8 cm),

so that they could not see their hands for about 60% of the way when they reached forward to-wards the stimuli, which were placed 19.5 cm from the table�s edge (see Fig. 2). The hand move-ments were filmed with a Sony Handycam DCR-PC 1E video camera from above and forward, sothat the hands were clearly visible to the camera before they became visible to the volunteers. Thestill pictures taken at this point (about 70% of the way from resting position to final grasp of thediagonal) in the reaching movement were analysed. Both index fingers and thumbs had a smallorange sticker with a black cross on it so that the finger-thumb span at this point could be moreeasily measured from the video record. For this measurement the video was digitalized with a res-olution of 24 pictures per second. The program ‘‘MainActor for Windows v3.6’’ extracted the sin-gle pictures of the video stream (24/s.). From this set of pictures the relevant ones were selected.Inspection of the data showed that the grasping movements did not show the typical maximumgrip aperture at about 70% along the reaching trajectory. Rather, it seemed as if volunteers tendedto continuously open their grip almost to the end of their movement trajectory and then quicklyclosed it on the target. The most likely reason, therefore, seems to be the low resolution of ourcamera (24 pictures/s). The mean duration for one grasping movement was about 0.5 s(SD = .12) which means that we have about 12 pictures of each grasping. As Kwok and Braddick(2003) showed, velocity of the movement is very high at the point of maximal grip aperture. Fromthis it follows that we had too few pictures at this point (2–3). Therefore, it is very likely theparticipants did show a maximal grip aperture but due to our low resolution failed to detect it.Nevertheless, the pattern of grip aperture was very similar to the pattern of grip aperture towardstwo- and three-dimensional objects in the experiment of Kwok and Braddick (2003).

Hence, we decided to measure grip size at a standard point after about 70% of the movement(when the thumbs of both hands had reached a certain reference point at 60% in the reaching

Fig. 2. Schematic representation of the experimental setup.

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 275

movement; at this point the index fingers had reached 80% along the reaching trajectory; thismeans the hand as a whole had reached on average 70% of the movement.). As a consequencethe average measured grip size (of the dominant hand) was only 5.88 cm even though the lengthof the diagonal was over 6.5 cm. The still pictures of the hand at this position were selected andtransferred into the graphic program ‘‘Paint’’ in which each point of the pictures was defined bytwo coordinates. The grip aperture was determined by computing the norm-vector between thetwo reference points on thumb and index finger. For purposes of data analysis the distancesbetween picture points used by the program were scaled back to actual distances of the environment.

For grasping trials the cover was adjusted in height to prevent subjects from seeing the first 60%of their hand movements. Additionally, subjects were instructed to keep their sitting posture sothat the cover just kept a green line on the table covered, which marked the 60% of their move-ment trajectory. They were asked to grasp the red diagonals of the two stimulus parallelograms asquickly as possible at the signal ‘‘go.’’ Between trials the stimuli were exchanged. At this point it isworth considering that measurement took place as soon as participants were able to see both—dumb and index fingers of the grasping hand. Because the movement at this point is very fast theyhad less than 50 ms for adjusting their grip aperture using this visual feedback. Normally, the useof visual feedback during grasping does not take place under 100 ms. (Paulignan, MacKenzie,Marteniuk, & Jeannerod, 1991).

For judgment trials, volunteers were asked to indicate which of the two diagonals was the long-er or whether they were the same length. If they opted for ‘‘same length’’ they were given a forcedchoice to decide which diagonal might be longer.

2.1.3. MaterialsThe two plotted parallelograms were 4 cm apart. The height of the parallelogram was 4.5 cm.

The angle between base and sides was 62�. The base lines of the parallelograms were adjusted, sothat the lengths of the diagonals were as prescribed by the three different variants shown in Table1. Thus, the base lines of the large parallelogram were either 7.1 or 7.5 cm, the base lines of thesmall parallelogram varied between 2.2 and 2.6 cm.

2.1.4. DesignHalf the volunteers were given all judgment tasks followed by all grasping tasks while the other

half of volunteers were given these tasks in the reverse order. For judgments each of the six condi-tions was presented once in random order. For grasping trials each left-right variant was used fourtimes in the ‘‘with’’ and ‘‘neutral’’ condition. In the critical ‘‘against’’ condition, where actual lengthdifferences between diagonals went against the direction of the illusion, they were presented eighttimes each (see Table 1). They were presented in four blocks of 8 trials. In each block every left-rightvariation of each condition appeared once and of the critical condition twice in random order (fixedonce for all participants). The order of the blocks was counterbalanced in a Latin Square design.

2.2. Results

2.2.1. JudgmentsTable 2 shows the number of trials in which the diagonal in the large parallelogram was judged

as longer. (‘‘Spontaneous judgement’’ refers to participants judging promptly one of the diagonals

Table 1Stimulus presentations in Experiment 1 and 2

Length difference condition Frequency of presentation Stimulus presentation and length ofdiagonals

Experiment 1 Experiment 2

Against illusion 8·

6.5 6.8 5.7 5.48·

6.8 6.5 5.4 5.7Neutral 4·

6.6 6.6 5.5 5.54·

6.6 6.6 5.5 5.5With illusion 4·

6.8 6.5 5.4 5.74·

6.5 6.8 5.7 5.4

276 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

as longer. ‘‘Forced choice’’ refers to participants judging the diagonals as being the same lengthand then they were asked to choose one as longer.’’) Data are collapsed for left-right presentationsas there was no bias apparent. As the means make very clear, even in the ‘‘against’’ illusion con-dition, where the diagonal in the small parallelogram was actually 3 mm longer than the diagonalin the larger parallelogram, almost all volunteers judged the diagonal of the larger parallelogramto be longer. Significance was tested with a sign test comparing the number of volunteers who hadopted more often for the larger parallelogram against those who had opted more often for thesmaller parallelogram as having the longer diagonal.

Table 2Number of spontaneous judgments and under forced choice

Length difference condition Diagonal judged ‘‘longer’’ Sign test

Forced choice Forced choice v2 p

Against 18 5 2 1 7.36 6.01Neutral 21 5 0 0 12.0 6.01With 24 2 0 0 12.0 6.01

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 277

2.2.2. GraspingOne obvious feature was that the dominant hand behaved quite differently from the non-dom-

inant hand. Mean grip size at the point of measurement was significantly smaller(F (1,12) = 10.19, p < .01) for the dominant hand (6.00 cm) than for the non-dominant hand(6.45 cm). Most importantly, grip size of the non-dominant hand showed no interpretable rela-tionship to the stimulus conditions. To illustrate, even when the actual difference in length wasenhanced by the illusion (‘‘with’’ illusion condition) the average grip size was non-significantlylarger (6.62 cm) for the longer diagonal than for the shorter (6.44 cm) diagonal: t (12) = 1.94,p > .05. We, therefore, analysed only the data from the dominant hand in detail.

An analysis of variance was carried out on the mean grip size for the four or eight (in the‘‘against’’ condition) trials of each condition with the factor task order (judgments first versusgrasping first) between participants and factors condition (‘‘against’’ vs. ‘‘neutral’’ vs. ‘‘with’’ illu-sion) and parallelogram (larger vs. smaller) as within participants factors. The analysis revealed asignificant interaction between parallelogram and condition (F (2,22) = 6.93, p < .01). The relevantmeans in Fig. 3 show that the interaction comes about because grip size differs significantly in thedirection of the true length of the diagonals. In the ‘‘with’’ illusion condition there is a just signif-icant difference (t (12) = 2.13, p = .05). There is a practically zero difference (0.02 cm) when diag-onals are the same length (t (12) = .44, p > .50). And, most importantly, grip size differs withactual length of diagonals even when it goes against the illusion (t (12) = 2.35, p < .05). Further-more, there was a significant main effect of the factor task order (F (1,11) = 7.98, p < .05). Subjectsshowed a larger grip aperture when they had to judge the length of the diagonals before they hadto grasp them. No obvious interpretation of this effect comes to mind.

Franz et al. (2001) pointed out another testable implication of the dual representation andthe single representation models. The single representation model implies that individual differ-ences in the susceptiblity to the illusion should be reflected in conscious judgments as well as ingrasping behaviour, resulting in a positive correlation between these two measures. In contrast,the dual representation model implies that there should be no such correlation since it positsthat there is no common representation subject to the illusion that influences both grasping

5,75

5,8

5,85

5,9

5,95

6

6,05

6,1

6,15

6,2

gri

p a

per

ture

large

small

p = .037

p = .667

p = .054

Parallelogram

against zeroCondition

with

Fig. 3. Mean grip size for dominant hand in Experiment 1.

278 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

and judgment. To compute this correlation we computed the following susceptibility score forconscious judgments. We gave participants on each judgment trial +2 for opting spontaneouslyfor the larger parallelogram as having the longer diagonal, +1 for doing so after forced choice,�1 for opting for the smaller parallelogram after forced choice, and �2 for doing so sponta-neously. The sum of this score over all six trials was the overall susceptibility score. For grasp-ing trials we simply computed the average difference in mean grip size between the large andthe small parallelograms for each length of diagonal. Importantly, this measure is based onthe means for each diagonal length and is, therefore, not biased by the fact that there wereeight trials for some and only four trials for other parallelograms. The mean susceptibility scorefor the dominant hand in the grasping condition was �0.025 cm (not significantly differentfrom zero: t (12) = .54, p > .50) and its correlation with the susceptibility measure for judgmentswas practically zero: r = .004, p = .990. Thus, there is no support for the single representationtheory but evidence that grasping with the dominant hand was not influenced by the illusion incontrast to the conscious judgements. This provides some support for the dual representationmodel.

2.3. Discussion

The results clearly show that all volunteers were strongly subject to the illusion in their con-scious judgment of the diagonals� relative length, even when there was an actual length differencein the opposite direction. In contrast, the results indicate that grip size scaling of the dominanthand is not (or hardly) at all influenced by the illusion. Grip size differences are significantly dif-ferent in direction of the diagonals� actual lengths even when the illusion works against this dif-ference. This result shows a clear dissociation of representations of object properties in verbaljudgment and action. The fact that judgment and action dissociated with our stimulus materialand procedure is important for two reasons.

(1) The material is (unlike the Ebbinghaus/Titchener circles and other traditional illusions) freeof �visual� obstacles for the grasping movement. This makes it difficult for the single repre-sentation model to defend the use of only one representation by claiming that the grip sizedifferences deviate from the conscious judgments due to extraneous reasons for adjustinggrip size (e.g., having to squeeze in fingers, etc.).

(2) We tried to equate the grasping and the judgment conditions for the need to consider bothillusory stimuli simultaneously, by letting volunteers grasp both diagonals at the same time.This attempt, however, partly failed. The larger grip sizes of the non-dominant hand and itsinsensitivity to the actual length of the diagonals or illusion effects alike suggests that despitethe need for grasping both diagonals volunteers only concentrated on their dominant hand.To improve on this deficiency we made a renewed effort to equate grasping and consciousjudgment conditions in our next experiment.

3. Experiment 2

One remaining problem in Experiment 1 was that instructions to grasp two objectssimultaneously do not ensure the same degree of simultaneous attention to both objects as the

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 279

instructions to compare the length of the two objects. To improve the comparability of attentionalfactors across the grasping and the judgment conditions we asked volunteers to provide consciouslength judgments for each diagonal by indicating the estimated length with their finger-thumbspan as used previously by Haffenden and Goodale (1998); Haffenden et al., 2001. Unlike a directlength comparison of two diagonals this measure provides independent length estimates in thesame way as simultaneous grasping.

Franz (2001) has argued, though, that manual estimates are a non-standard measure of percep-tion, which tends to inflate illusory differences. This inflation, however, cannot be due to a differencein the illusory effect, i.e., the mapping from stimulus to internal representation. It must occur in howthe internal representation is mapped onto the observable response. This kind of inflation, fortunate-ly, cannot produce the interaction we seek in the critical ‘‘against illusion’’ condition, i.e., that gripscaling during grasping indicates a difference in the direction of diagonals� actual lengths while lengthestimates given by finger-thumb spans yield a difference in the opposite direction.

We used the opportunity of a new study to introduce a slight modification of our stimulusmaterial, by making that part of each stimulus display, which is in the trajectory of the approach-ing hand, identical for both displays as shown in Fig. 4.

3.1. Method

3.1.1. ParticipantsNineteen participants were tested but due to a technical problem the data of one person could

not be used. The final sample, therefore, comprised 11 women and 7 men, with a mean age of24.83 (SD = 3.07) years. Thirteen participants said to be dominant right handed, five said to bedominant left handed. In every case participants� self-classification was confirmed by the HDThandedness test (Steingruber & Lienert, 1976).

3.1.2. Apparatus and procedureThe procedure for the grasping condition was the same as in the first experiment with the

exception of the use of the finger-thumb span for indicating consciously perceived length differ-ences. Participants were seated, like in the grasping condition, with their hands on pads under-neath a cover (see Fig. 2). However, the pads were moved 10 cm away from table edge, whichwas necessary for recording the hands on video at the same point between body and stimulus asin the grasping condition. Also, the cover was raised to make sure that the participants couldnot see their hands while indicating the length of the diagonals. As in the grasping condition,participants were asked to match the red diagonals with their index fingers and thumbs asquickly as possible at the signal ‘‘go.’’ They were instructed to do this simultaneously with bothhands. Pictures for measurement were selected when participants held their grip aperture steadyfor about one second. Otherwise this judgment condition was exactly the same as the graspingcondition.

3.1.3. MaterialsThe two plotted stimulus figures (see Fig. 4) were 4.5 cm apart. The base lines of the figures

varied between 4.0 and 4.2 cm, the height varied between 3.6 and 3.9 cm so that the lengths ofthe diagonals were as prescribed by three different variants shown in Table 1. The angle of the

Fig. 4. Stimuli for the ‘‘against illusion’’ condition of Experiment 2. In the experiment the diagonals were printed inred.

280 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

diagonals in both figures was kept constant at exactly 43� tilted towards the graspinghand.

3.1.4. DesignThe design was exactly the same as in Experiment 1 except that the manual judgment condition

replaced the verbal comparative length judgments. This made it possible to copy the design of thegrasping condition for the judgment condition.

3.2. Results

3.2.1. Indicating length (judgments)We analysed both hands separately to make a direct comparison with the grasping condition

possible. For the dominant hand an analysis of variance was carried out on the mean finger-thumb aperture over the four or eight (in the ‘‘against condition’’) trials of each condition withthe factor task order (matching first versus grasping first) between participants and factors condi-tion (‘‘against’’ vs. ‘‘neutral’’ vs. ‘‘with illusion’’) and figure (larger vs. smaller) as within partici-pants factors. The analysis revealed a significant main effect for factor figure (F (1,16) = 48.42,p < .001). Finger-thumb aperture was larger for the larger figure in all of the three conditions.Furthermore, there was a significant interaction between figure and condition (F (2,32) = 5.47,p < .01). Fig. 5A shows that the interaction comes about because finger-thumb aperture dependedon actual size difference (condition) and effects of the illusion (figure). In fact the influence of theillusion was so strong that even in the ‘‘against’’ condition it overpowered the actual lengthdifference in the opposite direction: t (17) = 2.11, p = .05.

The same analysis of variance for the non-dominant hand showed the same effects for figure(F (1,16) = 18.49, p < .01) and the interaction between figure and condition (F (2,32) = 5.67,p < .01) significant. As Fig. 5B shows the means look very similar to those from the dominanthand. However, the illusion was not strong enough make a significant difference in the ‘‘against’’condition (t(17) = .73, p = .48), only in the ‘‘neutral’’ condition (t(17) = 3.35, p < .01).

3.2.2. GraspingAgain, the grasping movements did not show the typical maximum grip aperture. For this

reason, we analysed the pictures at the point where the hands had gone through about 70% oftheir trajectory. As in the first experiment, the dominant hand behaved quite differently fromthe non-dominant hand. Although there was only a very small difference in grip size at the point

6,8

7

7,2

7,4

7,6

7,8

8

8,2

large

small

p = .004

p = .001

p = .000

Figure

against zeroCondition

with

6,8

7

7,2

7,4

7,6

7,8

8

8,2

dist

ance

bet

wee

n th

umb

and

inde

x fin

ger

p = .447

p = .004

p = .002

against zeroCondition

with

large

small

Figure

A

B

dist

ance

bet

wee

n th

umb

and

inde

x fin

ger

Fig. 5. (A) Mean finger-thumb aperture as indicator of perceived length in Experiment: dominant hand. (B) Meanfinger-thumb aperture as indicator of perceived length in Experiment: non-dominant hand

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 281

of measurement between dominant hand (5.62 cm) and non-dominant hand (5.50 cm), the gripsize of the non-dominant hand showed again no interpretable relationship with the stimulus con-ditions. To illustrate, even when the actually longer diagonal was made to look even longer in the‘‘with’’ illusion condition was the average grip size was smaller (5.67 cm) than for the shorter diag-onal (5.80 cm), though non-significantly so: t (17) = 1.49, p > .05. We, therefore, only analysed thedata from the dominant hand in detail.

The same analysis of variance was carried out as on the manual judgment data described above.The only significant effect was the interaction between figure and condition (F (2,30) = 11.51,p < .001). The relevant means in Fig. 6 show that the interaction comes about because grip sizediffers significantly in the direction of the true length of the diagonals. In the ‘‘with’’ illusion con-dition there is a highly significant difference (t (17) = 5.37, p < .001). There is practically no differ-ence when diagonals are the same length (t (16) = .66, p > .50). And, most importantly, grip sizediffers with actual length of diagonals even when it goes against the illusion (t (17) = 3.31, p < .01).

5,3

5,35

5,4

5,45

5,5

5,55

5,6

5,65

5,7

5,75

5,8

5,85

gri

p a

per

ture

p = .520p = .004

p = .000

large

small

Figure

against zeroCondition

with

Fig. 6. Mean grip size for dominant hand in Experiment 2.

282 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

3.2.3. Susceptibility to the illusionWe also computed scores for the susceptibility to the illusion in the grasping and in the judg-

ment condition. For grasping trials we used the same measure as in Experiment 1 and for judg-ment trials we computed the corresponding measure as for grasping. For the dominant hand thePearson correlation coefficient for these two susceptibility measures was non-significantly positive:r = .10, p = .71. This provides yet again no evidence for the single representation model. More-over, the mean susceptibility score in the grasping condition was .032 cm which is not significantlydifferent from zero (t (16) = .81, p = .43).1 In contrast, the mean susceptibility score for the judg-ment condition was .55 cm which is highly significantly different from zero (t (17) = 7.21,p < .001). Comparison of the two scores makes clear that susceptibility to illusion was much great-er for judgments than for grasping: t (16) = 6.04, p < .001. A similar result emerged for judgementby the non-dominant hand. The susceptibility score of 0.46 cm was highly significant(t (17) = 3.71, p < .002).

3.3. Discussion

The grasping data closely replicate the data from Experiment 1. Again grip size scaling on thenon-dominant hand was insensitive to the stimulus differences (even more so than in Experiment1), while the dominant hand behaved much more disciplined and reflected stimulus differencesminutely. There was no indication that the illusion affected the dominant hand�s grip scaling.In stark contrast, when finger-thumb scaling was used as conscious estimates of stimulus lengths,estimates always went significantly in the direction of the illusion even in the ‘‘against illusion’’condition, where the actual lengths differed in the opposite direction.

Knowing from Experiment 1 that despite the need for bimanual grasping the non-dominanthand showed such imprecise grip size scaling, we expected a similar finding for the finger-thumb

1 There is a change in the degrees of freedom because data of one participant are incomplete.

E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284 283

scaling as indicators of perceived size. To our surprise, in this task the non-dominant hand pro-duced results comparable to the dominant hand, showing comparable sensitivity to stimulus dif-ferences, though the influence of the illusion was not as strong. This indicates that, despite theapparent parallelism in task demands between grasping and conscious judgment by thumb-fingerspan, there still remained some attentional asymmetry between the grasping and judgment task.

4. General discussion

The main objective in this study was to find illusory stimuli, which do not require visual‘‘obstacles’’ in the path of the grasping hand, to avoid extraneous reasons for grip size scalingapart from the length of the to be grasped object. Such extraneous reasons can have potentialeffects that can be exploited by opposing interpretations. As Jacob and Jeannerod (2003) pointout, they detract from the interpretation that grasping is less influenced by illusions thanconscious judgment, because the observed difference could be due to such extraneous reasonsrather than a difference in internal representation.

Such visual obstacles may, however, also lead to the opposite effect of changing the grip scalingin the direction of the illusion. The data from Haffenden and Goodale (1998, 2000) provide con-crete evidence for this danger in the Ebbinghaus/Titchener illusion. But also the Muller-Lyer illu-sion may be particularly prone to this, because the arrow wings at the end of the stick may beperceived by the motor system as belonging to the object to be grasped. Consequently, the gripsize will be larger for these objects but for a reason different from the illusory effect created bythese arrow wings.

In sum, because such extraneous reasons were avoided in our stimuli our results provide im-proved support for the dual representation theory. There is little indication that grip size scalingof the dominant hand is subject to the illusion (and the non-dominant hand showed some signs ofbeing influenced only in Experiment 1). In contrast, comparisons judgments in Experiment 1 andmanual estimates on both hands in Experiment 2 as measures of conscious perception showedvery strong effects of the illusion.

References

Aglioti, S., DeSouza, J. F., & Goodale, M. A. (1995). Size-contrast illusions deceive the eye but not the hand. Current

Biology, 5(6), 679–685.Bridgeman, B., Gemmer, A., Forsman, T., & Huemer, V. (2000). Processing spatial information in the sensorimotor

branch of the visual system. Vision Research, 40, 3539–3552.Bridgeman, B., Kirch, M., & Sperling, A. (1981). Segregation of cognitive and motor aspects of visual function using

induced motion. Perception & Psychophysics, 29, 336–342.Bridgeman, B., Peery, S., & Anand, S. (1997). Interaction of cognitive and sensorimotor maps of visual space.

Perception & Psychophysics, 59(3), 456–469.Daprati, E., & Gentilucci, M. (1997). Grasping an illusion. Neuropsychologica, 35, 1577–1582.Franz, V. H., Bulthoff, H. H., & Fahle, M. (2003). Grasp effects of the Ebbinhaus illusion: Obstacle avoidance is not the

explanation. Experimental brain research, 149, 470–477.Franz, V. H., Gegenfurtner, K. R., Bulthoff, H. H., & Fahle, M. (2000). Grasping visual illusions: No evidence for a

dissociation between perception and action. Psychological Science, 11, 20–25.

284 E. Stottinger, J. Perner / Consciousness and Cognition 15 (2006) 269–284

Franz, V. H. (2001). Action does not resist visual illusions. Trends in Cognitive Sciences, 5, 457–459.Franz, V. H., Fahle, M., Bulthoff, H. H., & Gegenfurtner, K. R. (2001). Effects of visual illusions on grasping. Journal

of Experimental Psychology: Human Perception and Performance, 27, 1124–1144.Goodale, M. A., Jakobson, L. S., & Keillor, J. M. (1994). Difference in the visual control of pantomimed and natural

grasping movements. Neuropsychologica, 32, 1159–1178.Haffenden, A. M., & Goodale, M. A. (1998). The effect of pictorial illusion on perception and prehension. Journal of

Cognitive Neuroscience, 10, 122–136.Haffenden, A. M., & Goodale, M. A. (2000). Independent effects of pictorial displays on perception and action. Vision

Research, 40, 1597–1607.Haffenden, A. M., Schiff, K. C., & Goodale, M. A. (2001). The dissociation between perception and action in the

Ebbinghaus illusion: Nonillusory effects of pictorial cues on grasp. Current Biology, 11, 177–181.Jackson, S. R., & Shaw, A. (2000). The Ponzo illusion affects grip-force but not grip-aperture scaling during prehension

movements. Journal of experimental psychology: Human perception and performance, 26, 418–423.Jacob, P., & Jeannerod, M. (2003). Ways of seeing. The scope and limits of visual cognition. Oxford: Oxford University

Press.Kwok, R. M., & Braddick, O. J. (2003). When does Tichener circles illusion exert an effect on grasping? Two- and three-

dimensional targets. Neuropsychologica, 41, 932–940.Millner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford: Oxford University Press.Marcel, A. J. (1993). Slippage in the unity of consciousness. In G. R. Bock & J. Marsh (Eds.), Experimental and

theoretical studies of consciousness (pp. 168–186). Chichester: Wiley.Otto-de Haart, E. G., Carey, D. P., & Milne, A. B. (1999). More thoughts on perceiving and grasping the Muller-Lyer

illusion. Neuropsychologica, 37, 1437–1444.Paulignan, Y., MacKenzie, C., Marteniuk, R., & Jeannerod, M. (1991). Selective perturbation of visual input during

prehension movements. Experimental Brain Research, 83, 502–512.Pavani, F., Boscagli, I., Benvenuti, F., Rabuffetti, M., & Farne, A. (1999). Are perception and action affected differently

by the Titchener circles illusion? Experimental Brain Research, 127, 95–101.Perenin, M. T., & Rossetti, Y. (1996). Grasping without form discrimination in an hemianopic field. NeuroReport, 7,

793–797.Poppel, E., Held, R., & Frost, D. (1973). Residual visual function after brain wounds involving the central visual

pathways in man. Nature, 243, 295–296.Steingruber, H.J., & Lienert, G.A. (1976). Hand-dominanz-test (HDT). Hogrefe, Gottingen, Toronto, Zurich, 1976.Smeets, J. B. J., & Brenner, E. (1999). A new view on grasping. Motor Control, 3, 237–271.Vishton, P. M., Rea, J. G., Cutting, J. E., & Nunez, L. N. (1999). Comparing effects of the horizontal–vertical illusion

on grip scaling and judgment: Relative versus absolute, not perception versus action. Journal of Experimental

Psychology: Human Perception and Performance, 25, 1659–1672.Weiskrantz, L., Sanders, M. D., & Marshall, J. (1974). Visual capacity in the hemianopic field following a restricted

occipital ablation. Brain, 97, 709–728.Weiskrantz, L. (1998). Blindsight: A case study and implications. Oxford: Oxford University Press.Westwood, D. A., & Dubrowski, A. (2000). The effect of illusory size on force production when grasping objects.

Experimental brain research, 135, 535–543.Wong, E., & Mack, A. (1981). Saccadic programming and perceived location. Acta Psychologica, 48, 123–131.