Max Planck Institut für biologische Kybernetik. Spemannstraße Tübingen Germany

Transcription

1 MP Max Planck Institut für biologische Kybernetik Spemannstraße Tübingen Germany Technical Report No. 067 visual illusions: No evidence for a dissociation between perception and action. Volker H. Franz 1, Karl R. Gegenfurtner 1, Heinrich H. Bultho 1 & Manfred Fahle 2 February 1999 We wish to thank M. S. Banks and M. Jeannerod for helpful comments on earlier versions of the manuscript. Also we wish to thank M. A. Goodale and A. M. Haenden for long and fruitful discussions about our opposite views. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG) and by the Max{Planck Society. 1 AGBultho, E{mail: volker.franz@tuebingen.mpg.de 2 Department of Optometry and Visual Science, Goswell Road, London EC1V 7DD, UK This document is available as /pub/mpi-memos/tr-067.ps via anonymous ftp from ftp.kyb.tuebingen.mpg.de or from the World Wide Web,

2 visual illusions: No evidence for a dissociation between perception and action. Volker H. Franz, Karl R. Gegenfurtner, Heinrich H. Bultho & Manfred Fahle Abstract. Neuropsychological studies motivated the theory that the primate visual system might be organized into two parallel pathways, one for conscious perception and one to guide action (Milner & Goodale, 1995). Supporting evidence in healthy humans seemed to come from a dissociation in visual illusions: Aglioti, DeSouza, and Goodale (1995) reported that the Ebbinghaus (or Titchener) Illusion deceived perceptual judgments of size, but only marginally inuenced the size estimates used in grasping. Here we show that identical eects of the illusion are found if the perceptual and grasping tasks are appropriately matched. We show that the dierences found by Aglioti et al. (1995) can be accounted for by a hitherto unknown, super additive eect in the illusion. We conclude that the illusion does not provide evidence for the existence of two distinct pathways for perception and action in the visual system. Several theories state that visual information is processed in two dierent processing streams in the primate brain. The origin of these theories are two physiologically dierent projections from the primary visual areas to parietal and temporal areas of the cortex. Usually these projections are called the \dorsal" and the \ventral" stream. Based on lesion studies on monkeys, Ungerleider and Mishkin (1982) proposed that the main function of the dorsal stream is the analysis of the spatial relations between objects (\where" pathway), while the function of the ventral stream is the recognition of objects (\what" pathway). Based on neuropsychological studies (Goodale, Milner, Jakobson, & Carey, 1991; Goodale et al., 1994) showing a double dissociation between grasping an object and perceiving its shape, Goodale and Milner (1992; Milner & Goodale, 1995) re{interpreted this theory. They suggested that the function of the dorsal stream is not the analysis of the location of objects, but the computations that are needed to guide the manipulation of objects, while the function of the ventral stream is to perform computations that are necessary for object recognition and conscious perception. They argued that these computations have to full{ll totally dierent requirements. Computations for the guidance of actions have to be fast, have to precisely code the position of the object relative to the eector and only need a very short memory because the position of the object can change quickly. In contrast, computations for the purposes of object recognition do not need to be as precise and fast. Here it is much more important to evaluate an object in its context and to enable a long lasting representation. Therefore, Milner and Goodale (1995) suggested that there exist two dierent visual systems, one that guides motor actions and one that leads to object recognition and conscious perception. Of course, if such a division of labor exists in the primate brain, it should be reected not only in decits of neuropsychological patients, but also in the healthy visual system. Studies on grasping in humans seemed to provide this missing piece of evidence. In order to grasp, humans have to move their hand close to the target object. During the reach, the index nger and thumb open to a maximum aperture that is linearly related to object size (Fig. 1a; Jeannerod, 1981, 1984). This maximum preshape aperture is formed before the hand had any contact with the object. Therefore the maximum preshape aperture reects the size estimate being transferred to the motor system from the visual system (if no other, non{visual cues about object size are available). Because conscious perception is receptive to a number of size illusions (Coren & Girgus, 1978), the question arises whether the maximum preshape aperture will be aected as well. In principle, there are two possibilities. The rst is that the motor system uses the same internal representation of object size as perception does. This common representation would be inuenced by the illusion. Due to the linear relationship between object size and maximum pre- 1

3 a Aperture Starting reach Time Max. preshape aperture Touching object b Max. preshape aperture Grasped larger bec. of illusion Perceived larger bec. of illusion Physical size of object c Aluminum disc d Comparison circle Board with context circles Figure 1: Maximum preshape aperture and the apparatus of Experiment 1. a. Typical time course of the aperture between index nger and thumb during the transport component of a prehension movement. After about 75% of the reaching time, the ngers are opened maximally (maximum preshape aperture). b. The maximum preshape aperture is linearly related to the physical size of the object. If perception and grasping share a common representation of object size, then an illusory change in perceived size should lead to a predictable change in maximum preshape aperture. c. In Experiment 1 participants viewed a board with either large or small context circles drawn on it. In the center of the context circles an aluminum disc was positioned. In the grasping task, the participants grasped the disc. In the perceptual task, they adjusted a comparison circle displayed on the monitor to match the size of the aluminum disc. d. A participant grasping the aluminum disc. Note the infrared light{emitting diodes attached to thumb and index nger. They allow to record the trajectory of the nger tips. shape aperture, it would be possible to predict the inuence of the illusion on maximum preshape aperture (Fig. 1b). The second possibility is that the motor system uses a dierent representation of object size than perception. In this case the maximum preshape aperture could be unaected by the illusion. This is the prediction of the perception vs. action hypothesis of Milner and Goodale (1995). The computations for perception should focus on the relationship between an object and its surrounding objects, while the computations for action should focus on the relationship between an object and the eector to be used (in this case the hand). Because visual size illusions are often generated by special arrangements of objects, the theory would predict that these illusions should have only little (or even no) inuence on the motor system. The rst and most inuential study investigating this topic was performed by Aglioti et al. (1995) and was replicated by Haenden and Goodale (1998). These original studies used the Ebbinghaus (or Titchener) Illusion: a central circle surrounded by large circles appears smaller than when surrounded by small circles. To determine the inuence on grasping, the central circle was replaced with a disc which was grasped by the participants. Results showed a larger eect of the illusion on perception than on the maximum preshape aperture. This was interpreted as strong evidence for the theory of Milner and Goodale (1995). the Illusion However, there are several problems in these original studies which prompted us to try a replication using an improved and simplied design. First, even though the eect on perception was larger, the original studies also found inuences on grasping as did other investigators (Brenner & Smeets, 1996; Daprati & Gentilucci, 1997). The picture, however, is blurred because some studies reported a statistically signicant motor illusion (Aglioti et al., 1995; Daprati & Gentilucci, 1997) while in other studies the eect of the illusion on grasping failed to reach signicance (Haenden & Goodale, 1998; Brenner & Smeets, 1996). This is not very surprising because grasping shows a rela- 2

4 a Experiment 1: Single context versions b Original studies: Composite version Perception Direct comparison c Experiments 2 and 3: Separate comparisons Figure 2: Stimuli used in the Experiments 1 and in the original studies. a. In Experiment 1 the participants operated on only one Ebbinghaus gure at a time (single{context versions). b. In the original studies of Aglioti et al. (1995) and Haenden & Goodale (1998) asymmetric measures were used: To perform the grasping task, participants had to calculate only the size of one of the central discs. In the perceptual task, however, participants had to compare the two central discs directly, both being subjected to the illusion at the same time. c. A perceptual task that would be more similar to the grasping task of the original studies. In our Experiments 2 and 3, the three dierent methods to determine the perceptual eect of the illusion were compared. tively high variability. We tried to settle this topic by using more repetitions per participant and a larger sample size. Second, the original studies compared the in- uence of the illusion on the perceptual measures directly with the inuence on maximum preshape aperture. For example, Aglioti et al. (1995) found an inuence of 2.5 mm on perception and of 1.6 mm on grasping (Fig. 4b). Due to the statistically signicant dierence between these values, they concluded that the inuence of the illusion on the motor system was dissociated from the in- uence on perception. However, this conclusion is only valid if the function relating perceived size to physical size and the function relating maximum preshape aperture to physical size are both linear and have the same slopes. To see this, assume that maximum preshape aperture depended on physical size with a slope of 1=2, while perceived size had a slope of 1. In this case even the common{ representation model would predict only a motor illusion of half the size of the perceptual illusion. To obtain a good estimate of these slopes, we used a wider range of disc sizes than the original studies. Third, in order to make the perceptual and motor tasks as similar as possible, we presented only one Ebbinghaus gure at a time: A central disc was surrounded by either large or small context circles (single{context versions, Fig. 2a). In the 3

5 1 Perception Maximum grip aperture [mm] Effect of illusion Effect of illusion Adjusted size [mm] Size of 0.4central disc 0.6[mm] Figure 3: Eects of the Ebbinghaus Illusion on size perception and on maximum preshape aperture for each central disc size in Experiment 1. Solid lines represent data for small context circles, dashed lines for large context circles. The illusion inuenced perception (F (1; 19) = 112, p<:0) as well as grasping (F (1; 19) = 10.3, p = :005). The slopes for perception (s =1:1) and for grasping (s =1:2) were similar (t(19) = 1.51, p = :15). Mean eects of the illusion pooled over all disc sizes are shown in Fig. 4a. Error bars depict 1 standard error of the means (data are normalized to account for absolute dierences in hand sizes and aperture sizes between the participants). perceptual task, participants indicated the size of the central disc by adjusting the radius of an isolated circle displayed on a monitor (Coren & Girgus, 1972; Pressey, 1977). In the grasping task, participants grasped the central disc. This is different from the original studies in the following ways: The original studies used the composite version of the illusion (Fig. 2b) two Ebbinghaus gures with dierent context circles were presented simultaneously. In the perceptual task, the participants directly compared the two central discs. In the grasping task, however, the participants grasped only one of the discs. The overall eect of the illusion on grasping was then determined by adding the eects each context had on grasping. Note the asymmetry in this procedure. In grasping, the participants operated on only one Ebbinghaus gure at a time, while in the perceptual task they operated on both gures simultaneously. The implicit assumption in this procedure is that the perceptual eects of the two Ebbinghaus gures simply add up to yield the eect obtained by a direct comparison. Our Experiments 2 and 3 will show that this is not the case. A direct comparison between two Ebbinghaus gures yields a larger eect than if the perceptual eects are determined for each gure separately and then added (Fig. 2c). Results and Discussion In Experiment 1, we found highly signicant effects of the illusion on perception and on grasping (Fig. 3). Furthermore, maximum preshape aperture and perceived size were linearly related to physical size with similar slopes. This nding allowed a comparison of the illusion eects. The magnitude of the eects were equal in perception and in grasping and were in a range typically found for the Ebbinghaus Illusion (Fig. 4a; Coren & Girgus, 1972; Coren & Miller, 1974). These results clearly contradict the notion that the eects of the Ebbinghaus Illusion are dissociated between action and perception. Why, then, did the original studies nd a dierence between action and perception? A comparison shows that the grasping eects were similar between our Experiment 1 and the original studies (Fig. 4) but the perceptual eects were larger in the original studies. We hypothesized that this larger eects were due to the direct comparison between the two Ebbinghaus gures in the original studies. To test this hypothesis, we conducted 4

6 1 a Experiment 1 b Aglioti et al. c Haffenden & Goodale Mean effect of illusion [mm] Perception Perception Perception Figure 4: Overall eects of the Ebbinghaus Illusion on grasping and perception in Experiment 1 and in the original studies. Conditions in which a direct comparison between two Ebbinghaus gures was required are lled (for a description of the stimulus conditions see Fig. 2). a. In Experiment 1, the illusion aected grasping just as much as perception (t(19) = 0.13, p = :90, N = 20). b. In the Aglioti et al. (1995) study, the illusion aected grasping signicantly less than perception (p < :02, N = 14). c. Haenden & Goodale (1998) replicated the ndings of Aglioti et al. (1995) (N = 18). Figural elements were similarly sized in all studies, allowing a comparison of the illusion eects. Error bars depict 1 standard error of the means. two more experiments. In Experiment 2 we measured the perceptual eects for both the single{context versions (as in Experiment 1, Fig. 2a) and a direct comparison in the composite version (as in the original studies, Fig. 2b). Results replicated the dierences between the original studies and Experiment 1: The direct comparison yielded a much larger perceptual eect than the sum of the eects in the single{context versions (Fig. 5a) 1. In a third condition, we measured the perceptual eect of each Ebbinghaus gure of the composite version separately. The participants viewed the composite version, and alternately compared one of the central circles to an isolated circle (Fig. 2c). Again, the direct comparison yielded a larger eect than the sum of the eects in the two separate comparisons. These results reveal a super additive eect in the Ebbinghaus Illusion. If people directly compare two Ebbinghaus gures, they experience a larger 1 Because the Experiments 2 and 3 were purely perceptual, all stimuli were presented on a computer screen. This increased the similarity between central and context circles, and therefore yielded slightly larger eects of the illusion than Experiment 1 and the original studies (Coren & Miller, 1974). size illusion than is predicted by the sum of the size illusions experienced in each gure separately. Interestingly, most quantitative research on the Ebbinghaus Illusion has been based on the single{ context versions, while qualitative demonstrations of the illusion usually employ a direct comparison in the composite version and therefore exhibit an eect that is about 50% larger (Coren & Girgus, 1972, 1978). The super additivity contradicts the implicit assumption of the original studies that the perceptual eects of the two Ebbinghaus gures simply add up to yield the eect obtained by the direct comparison. Therefore, it is inappropriate to compare the motor and the perceptual illusions in those studies. However, one possible objection to this argument could be that in the original studies the participants had to directly compare the two central discs immediately before grasping (this was done as a control for the perceptual eect). Could this direct comparison, in which the participants were forced to attend to both central discs, eliminate the super additivity? In Experiment 3, we mimicked this procedure perceptually. The participants rst compared the two central circles 5

7 Mean effect of illusion [mm] a Experiment 2 b Experiment 3 c Experiment 3 First...: Direct comparison...then: Separate comparison Sum of Sum of Direct Sum of Direct 0single 0.2sep sep Figure 5: Eects of the Ebbinghaus Illusion in the Experiments 2 and 3 (for a description of the stimulus conditions see Fig. 2). a. Experiment 2: The eect of the direct comparison was larger than the sum of the eects in the two separate comparisons (t(17) = 2.27, p = :04) and than the sum of the eects in the single{context versions (t(17) = 3.68, p = :002). Single{context versions and separate comparisons showed similar eects (t(17) = 0.99, p = :34). b. Experiment 3: Even if participants performed a direct comparison immediately before each separate comparison, the direct comparison showed a larger eect than the sum of the two separate comparisons (t(11) = 3.45, p = :006). c. Typical trial of Experiment 3: Stimuli were presented for 1 sec. Within this time interval the participants rst performed a direct comparison between the two central discs as they did in the perceptual task of the original studies. Then, they compared one of the central circles to the isolated circle. This should mimick the grasping task of the original studies. Error bars depict 1 standard error of the means. directly and then, immediately afterwards, compared one of the central circles to the isolated circle (Fig. 5c). Both comparisons were performed within 1 sec, which is similar to the mean onset time for grasping in the Aglioti et al. (1995) study. Results show, that the super additivity was not eliminated by this immediate succession of direct comparison and separate comparison. The eect of the illusion in the direct comparison was again much larger than the sum of the eects in the two separate comparisons (Fig. 5b). Conclusions If perceptual and motor tasks are carefully matched, there are strikingly similar eects of the Ebbinghaus Illusion on perceived size and on maximum preshape aperture. The dierences found in the original studies by Aglioti et al. (1995) and Haenden and Goodale (1998) are due to a super additivity that selectively enhances the perceptual eects of the illusion. Therefore, the Ebbinghaus Illusion does not provide evidence for dierent processing mechanisms for perception and action. In contrast, the results strongly suggest that the same internal representation is used for perception and for grasping in this illusion. This outcome contradicts the predictions of the perception vs. action hypothesis of Milner and Goodale (1995) and removes one critical piece of evidence that is usually counted in favor of this theory (e.g. Jackson & Husain, 1997). Methods PARTICIPANTS. Students of the University of Tubingen, Germany, participated in the study and received a payment of 13 DM per hour. They had normal or corrected to normal vision. Twenty students took part in Experiment 1. They were right{handed according to the Edinburgh Inventory (Oldeld, 1971). Eighteen students took part in Experiment 2 and twelve students in Experiment 3. STIMULI. In all experiments, the large (small) 6

8 context elements were 5 (12) circles, 58 mm (10 mm) in diameter, the centers of the circles being 118 mm (60 mm) apart. In the composite version, the centers of the central discs were 140 mm apart. The isolated circle had a distance of 155 mm from the central disc in Experiment 1 and of 140 mm in the other experiments. The central discs in Experiment 1 were 28, 31, 34 or 37 mm in diameter (corresponding to 2.5, 2.7, 3.0, 3.3 deg visual angle) and 5 mm in height. In Experiment 2, the central circles had diameters of 28, 31 and 34 mm (corresponding to 2.5, 2.7, 3.0 deg visual angle) and in Experiment 3, they were 31, 32, 33 and 34 mm (corresponding to 2.7, 2.8, 2.9 and 3.0 deg visual angle). EXPERIMENT 1. The apparatus of Experiment 1 is shown in Fig. 1c,d. Special care was taken to minimize the perceptual dierence (Coren & Miller, 1974) between central disc and context by minimizing shadows and having the participants view the disc from above. The participants sat on a stool at a viewing distance of approximately 65 cm. In the grasping task, the participants grasped the central disc with their right (dominant) hand. They wore liquid-crystal shutter glasses and could neither see their hand nor the stimulus during grasping (open loop condition; Haenden & Goodale, 1998; Jeannerod, 1981; Post & Welch, 1996). The grasp trajectory was recorded using an Optotrak TM system: Three infrared light{emitting diodes were attached to thumb and index nger, and the maximum preshape aperture between the nger tips was calculated for each grasp. In the perceptual task, the participants adjusted the comparison circle displayed on the monitor to match the size of the central disc (Pressey, 1977). Each participant performed 72 grasps and 24 adjustments. EXPERIMENTS 2 and 3. In the perceptual experiments the participants sat in front of a computer screen at a viewing distance of approximately 65 cm. In Experiment 2 all three possibilities to assess the perceptual eect of the illusion were employed as shown in the lower part of Fig. 2. In the single context condition the participants adjusted the size of an isolated circle to match the size of the central circle in one Ebbinghaus gure. The eects of the large context circles and of the small context circles were added to obtain an estimate of the illusion strength. In the direct comparison condition, the participants adjusted the central circles of the two Ebbinghaus gures simultaneously. The dierence between the two central circles that was needed for the perception of equal size was used as measure for the illusion strength. In the separate comparison condition, participants viewed both Ebbinghaus gures, however, adjusted the isolated circle to match the size of only one of the central circles. The illusion strength was calculated the same way as in the single context condition. Each participant performed a total of 75 adjustments. A typical trial of Experiment 3 is shown in Fig. 5c. The participants performed the direct comparison and the separate comparison in direct succession, within 1 sec. Because this short time interval did not allow an adjustment procedure, the method of constant stimuli with two two{alternative forced choices was used. Each participant compared 462 congurations to complete the psychometric functions. References Aglioti, S., DeSouza, J. F. X., & Goodale, M. A. (1995). Size { contrast illusions deceive the eye but not the hand. Current Biology, 5 (6), 679 { 685. Brenner, E., & Smeets, J. B. J. (1996). Size illusion inuences how we lift but not how we grasp an object. Experimental Brain Research, 111, 473 { 476. Coren, S., & Girgus, J. S. (1972). A comparison of ve methods of illusion measurement. Behavioral Research Methods & Instruments, 4 (5), 240 { 244. Coren, S., & Girgus, J. S. (1978). Seeing is deceiving: The psychology of visual illusions. NJ: Lawrence Erlbaum. Coren, S., & Miller, J. (1974). Size contrast as a function of gural similarity. Perception & Psychophysics, 16 (2), 355 { 357. Daprati, E., & Gentilucci, M. (1997). an illusion. Neuropsychologia, 35 (12), 1577 { Goodale, M. A., Meenan, J. P., Bultho, H. H., Nicolle, D. A., Murphy, K. J., & Carolynn, I. R. (1994). Separate neural pathways for the visual analysis of object shape in perception and prehension. Current Biology, 4 (7), 604 { 610. Goodale, M. A., & Milner, A. D. (1992). Separate visual pathways for perception and action. Trends in Neurosciences, 15, 97 {

9 Goodale, M. A., Milner, A. D., Jakobson, L. S., & Carey, D.P. (1991). A neurological dissociation between perceiving objects and grasping them. Nature, 349, 154 { 156. Haenden, A. M., & Goodale, M. A. (1998). The eect of pictorial illusion on prehension and perception. Journal of Cognitive Neuroscience, 10 (1), 122 { 136. Jackson, S. R., & Husain, M. (1997). Visual control of hand action. Trends in Cognitive Sciences, 8 (1), 310 { 317. Jeannerod, M. (1981). Intersegmental coordination during reaching at natural visual objects. In J. Long & A. Baddley (Eds.), Attention and performance (Vol. 9, pp. 153 { 168). Erlbaum, Hillsdale. Jeannerod, M. (1984). The timing of natural prehension movements. Journal of Motor Behavior, 16 (3), 235 { 254. Milner, A. D., & Goodale, M. A. (1995). The visual brain in action. Oxford: University Press. Oldeld, R. C. (1971). The assessment and analysis of handedness: The Edinburgh Inventory. Neuropsychologia, 9, 97 { 113. Post, R. B., & Welch, R. B. (1996). Is there dissociation of perceptual and motor responses to gural illusions? Perception, 25 (5), 569 { 581. Pressey, A. W. (1977). Measuring the Titchener circles and Delboeuf illusions with the method of adjustment. Bulletin of the Psychonomic Society, 10 (2), 118 { 120. Ungerleider, L. G., & Mishkin, M. (1982). Two cortical visual systems. In D. J. Ingle, M. A. Goodale, & R. J. W. Manseld (Eds.), Analysis of visual behavior (pp. 549 { 586). MIT Press, Cambridge. 8