A double dissociation between action and perception in bimanual grasping: evidence from the Ponzo and the Wundt–Jastrow illusions

Research on visuomotor control suggests that visually guided actions toward objects rely on functionally distinct computations with respect to perception. For example, a double dissociation between grasping and between perceptual estimates was reported in previous experiments that pit real against illusory object size differences in the context of the Ponzo illusion. While most previous research on the relation between action and perception focused on one-handed grasping, everyday visuomotor interactions also entail the simultaneous use of both hands to grasp objects that are larger in size. Here, we examined whether this double dissociation extends to bimanual movement control. In Experiment 1, participants were presented with different-sized objects embedded in the Ponzo Illusion. In Experiment 2, we tested whether the dissociation between perception and action extends to a different illusion, the Wundt–Jastrow illusion, which has not been previously used in grasping experiments. In both experiments, bimanual grasping trajectories reflected the differences in physical size between the objects; At the same time, perceptual estimates reflected the differences in illusory size between the objects. These results suggest that the double dissociation between action and perception generalizes to bimanual movement control. Unlike conscious perception, bimanual grasping movements are tuned to real-world metrics, and can potentially resist irrelevant information on relative size and depth.

Scientific RepoRtS | (2020) 10:14665 | https://doi.org/10.1038/s41598-020-71734-z www.nature.com/scientificreports/ in which we measured the physical difference between the objects that could generate an illusory effect in the opposite direction. In the illusory background condition (grasping and manual estimations), two objects were placed on the upper and lower sides of a standard or inverted version of the Ponzo illusion (Fig. 1). The purpose of using the inverted configuration was to allow proper control for measuring the effect of the illusion for objects located at the same distance from the participant. Without the inverted configuration, the illusory effect would always be confounded with distance from the observer, and distance from the observer has been shown to have a significant effect on grip apertures 34 . Therefore, to focus on the effect of the illusion, it is important to compare grip apertures between objects located at the same distance from the participant. In the control grasping condition, the objects were placed against a white background. The objects were always placed at fixed locations. The distance between the initial location of the fingers and the nearer object was 24 cm, and between the two objects was 22.5 cm. The participants' viewing distance was approximately 40 cm for the nearer object. procedure. The procedure was based on the procedure used by Ganel et al. 18 and was adapted for bimanual grasping control. In Experiment 1a, participants grasped objects placed on the illusory background or the control, non-illusory background. The two conditions were performed in separate blocks, and block order was counterbalanced between participants. Before each trial, participants placed their index fingers around a small circular disk (2 cm in diameter), used as the starting point, while the goggles were set on their non-transparent state. In the illusory background condition, the participant was asked to lift the object she perceived as the shorter/ longer in the pair using both her hands. A verbal command signaled the required judgment ("short"/"long"), and an additional tone signaled movement initiation. In the control grasping condition, the participant was asked to lift the closer/farther object according to the corresponding verbal command ("close"/"far"). Each trial began with the verbal command which was followed by the openning of the goggles that stayed in their transparent state for 5 s, allowing on-line vision of the display and objects during the trial. The go signal was presented 1 s after the opening of the goggles.
The design of Experiment 1b was similar to the one used in Experiment 1a (illusory background condition), except that now the participants estimated the sizes of objects by manually adjusting the distance between their fingers following the opening of the goggles. Perceptual estimations were performed on the lower right side of the www.nature.com/scientificreports/ tabletop. To allow equal haptic feedback to the one provided in the grasping condition, following each estimation trial the participants were asked to grasp and lift the target object using both hands. Following perceptual estimates, a "go" signal to grasp the object was presented 5 s after the opening of the goggles. The goggles remained open for additional 3 s to provide online vision during grasp. In each illusory background condition (grasping; perceptual estimations), a total of sixty experimental trials were used. The standard configuration was presented for thirty consecutive trials, and the inverted configuration was presented in the remaining trials. Presentation order was counterbalanced between participants. Overall, there were 44 critical incongruent trials, in which the physically longer object (22 cm) was placed in the illusory background that induced an illusory effect in the opposite direction (22 trials in each illusory configuration). There were eight additional incongruent catch trials in which the physically shorter object was placed together with a noticeably larger object, 24.5 cm in length. In this configuration, the illusory background did not lead to reversal of the perceived size differences. In the eight remaining catch trials, the smaller and the larger objects were placed in illusory congruent locations. Trial order was pseudorandomized. In the control grasping condition, there were a total of thirty-two trials; in half of the trials, the 21 cm object was closer to the participant, and in the remaining trials the 22 cm object was closer (presentation order was counterbalanced). The 24.5 cm object was not presented in the control condition. Each condition started with a few practice trials. Data analysis. We recorded the 3D trajectories of the fingers in each trial. Movement onset was set as the point in time at which the aperture between two index fingers increased by more than 0.1 mm per frame for at least 75 ms. Movement offset was set as the point in time at which the aperture between the fingers changed less than 0.1 mm per frame for at least 125 ms (25 frames), but only after reaching the maximum grip aperture (MGA). The criteria used to detect movement onset and offset were based on the values used in a previous study from our lab in which we focused on unimanual grasps 35 . The values were slightly adjusted to allow effective identification of movement onset during bimanual grasping movements.

Data analysis.
For our primary analysis, we calculated the average MGA for each object size. The MGA provides reliable measure of the sensitivity of the grip aperture to object size 27,28 . To further analyze the movement trajectory, we divided the raw movement data to 11 equal intervals and calculated the average aperture in each interval (see Online Appendix 1). Detailed data for each of the participants in experiments 1 and 2 is provided in Online Appendix 2.
To avoid a potential confounding effect of obstacle avoidance when reaching out to grasp the distant object, we focused our main analysis only on near objects. Previous research suggests that physical stimuli in the target's environment can be treated as obstacles and affect kinematic aspects of the action toward the target [37][38][39][40] . For example, studies have shown that stimuli in the target surroundings could lead to deviations in movement trajectories from their original location 39,40 . As can be seen in Figs. 1 and 3, in the current study the near objects could be treated as obstacles, and could potentially affect the apertures between the fingers in flight. This would not have been a major issue if the obstacles were identical in size. However, they are not; when reaching out to grasp the physically far small object, the obstacle is always large (22 cm in Experiment 1) and when reaching out to grasp the physically large far object, the obstacle is always small (21 cm in Experiment 1). This potential confound makes it difficult to interpret the results of grasping objects in a far location. We therefore do not include the objects in this location in our main analysis and focus on the near location. Grasping the near objects is free of potential influences of obstacle avoidance and provides valid measure of grasping performance of objects located in a fixed distance from the participants' starting position.
Thus, we compared the MGA obtained in two separate configurations of the illusion (standard/ inverted). In Experiment 1, the critical comparison was between the 21 cm object presented on the inverted configuration of the illusion (and perceived in most trials as larger) and the 22 cm object on the standard configuration of the illusion (perceived in most trials as smaller). We note that previous studies argued that such a between-blocks analysis could potentially prevent unwarranted effects of an attentional mismatch 12 .
In addition, as in Ganel et al. 's 18 study, in order to disentangle the effects of apparent and real size on grip apertures, we focused our analysis on the subset of trials in which the participants erroneously judged the object as shorter/longer due to the illusion. In Experiment 1a (illusory background condition), the participants' mean error rate for the incongruent trials was 80.9% (SD = 11.3, 35.2 trials on average). The error rate for near objects was 72% (SD = 15, 15.8 trials). The overall error rate in the perceptual estimation task was slightly larger overall (79%, SD = 16, 34.7 trials), as well as for the near objects (75%, SD = 17, 16.5). However, the difference in the overall error rate between the estimation and grasping conditions was not significant [t (24) = 0.3, p = 0.7], as well as for the near objects [t (24) Results and discussion. The main results are presented in Fig. 2. As can be seen in the figure, a dissociable pattern of performance was obtained for grasping grip apertures and for perceptual estimates of size. For grasping (Experiment 1a), the scaling of apertures reflected the actual size differences between the objects rather than their apparent sizes [t (13) = 4.2, p = 0.001]. The MGA was larger for an object that was perceived as smaller (253, 260 mm, for the short and large objects, respectively). In sharp contrast to grasping, manual estimations (Experiment 1b) went in the same direction as the overt perceptual judgments and were biased by the illusion. Therefore, the distance between the fingers during perceptual estimations reflected the illusory size difference between the objects [215, 205 mm, for the short and large objects, respectively, t (12) = 2.6, p = 0.02]. A mixed model ANOVA was conducted on grip apertures with experiment as a between-subjects independent variable and size as a within-subject factor. There was a significant main effect for experiment [F ( . The pattern of results in the control grasping condition in Experiment 1a was similar to that found in the main grasping condition with the illusory background. The MGA was smaller for the 21 cm object compared to the 22 cm object [260, 267 mm, t (13) = 6.0-, p < 0.001]. To look for potential differences between the illusion and control grasping condition, a repeated-measures ANOVA test with condition and size as within-subject variables was conducted on the MGA data. The test showed a main effect for condition [F (1,13) = 13.5, p = 0.003, η p 2 = 0.50], with larger apertures in the control condition compared to the illusion condition (263, 256 mm, respectively). The main effect of size was also significant [F (1,13) = 51.7, p < 0.001, η p 2 = 0.80]. The difference in overall MGA between the control condition and the illusory condition was not predicted and we do not have a solid account for this effect. More importantly, however, there was no interaction between condition and size [F (1,13) = 0.13, p = 0.7], indicating similar sensitivity to physical size differences in the two conditions. An analysis of the grip apertures across the movement trajectory in Experiment 1a is presented in Online Appendix 1. The pattern of results was similar to that obtained for the MGA data. Again, grip apertures reflected the real size differences between the objects. Mean MGAs and grip apertures in the manual estimation condition across the different distances from the target are presented in Table 1.
The results of experiments 1a and 1b extend and replicate previous findings obtained in the domain of unimanual control 18 . While bimanual estimations reflected the illusory size difference between the objects, bimanual grasping apertures reflected their actual difference in size. In Experiment 2, we tested whether this pattern of results extends to a different illusion of size, the Wundt-Jastrow illusion.

Experiment 2
Fourteen participants (4 males, average age = 25.5, SD = 2.3), who did not participate in Experiment 1, participated in Experiment 2a (bimanual grasping condition), and ten other participants, who did not participate in Experiment 1, (1 male, average age = 23.6, SD = 1.7) participated in Experiment 2b (bimanual estimation condition). The results of two participants were excluded from the analysis. One participant (Experiment 2a), who did not follow the instructions, and another participant (Experiment 2b) due to a failure to capture her movements.
Stimuli and design. The overall design was similar to the one used in Experiment 1, except that now participants grasped different sized fan-shaped objects that constitute the Wundt-Jastrow Illusion. In the classic (2d) version of this illusion, the two fans are presented next to one other so that the inner circle of one object is vertical to the exterior circle of the other. This spatial organization leads to the perception that the inner fan is considerably bigger than the outer fan 41,42 (also see 43 on the similarities between the Ponzo and the Wundt-Jastrow illusions). In the modified 3D version used in Experiment 2, the two fans were physically different in size. For the critical incongruent trials, the physically shorter object was placed in the inner position and appeared (in most trials) as larger (Fig. 3a,b). Three different-sized objects were used; all objects were 4 cm in width and 2 cm in depth. The lengths between the top edges were 17.5, 18.5 and 21.5 cm for the small, medium, and big objects, respectively (the radiuses between the top edges to a reference point located 11 cm below its center point were 68, 70, 88 degrees, respectively). The largest object was used for the incongruent catch trials. The distance between the fingers' starting position to the nearest object was the same as in Experiment 1. The horizontal distance between the two target objects was 3 cm. As in Experiment 1 (illusory background conditions), a total of sixty experimental trials were presented, and each version of the illusion (standard/ inverted) was presented in 30 consecutive trials (mini-block order was counterbalanced between participants). The participants were asked to lift one of the objects by their external edges (Fig. 3a). Overall rates of susceptibility to the illusion in the grasping experiment were relatively high (90.1%, SD = 7, 40 trials on average). The error rate for near objects was 85% (SD = 11, 18.7 trials). The overall error rate for the perceptual estimations was 81% (SD = 14, 35.6 trials) and 76%, for the near objects (SD = 18, 16.7 trials). The differences between grasping and manual estimations in terms of overall error rate [t (20) = 1.9, p = 0.06] as well as for near objects [t (20) = 1.5, p = 0.14] did not reach statistical significance. In contrast, in the perceptual estimation task, finger apertures reflected the illusory size differences between the objects [186, 171 mm, for the short and large objects, respectively, t (8) = 3.9, p = 0.004]. These findings document a pattern of a double dissociation between grasping and perceptual estimates in terms of their sensitivity to real and to illusory size. A mixedmodel ANOVA with experiment as a between-subjects independent variable and size as a within-subject factor was conducted on the aperture data. There was a main effect for condition [F (1,20) = 41, p < 0.001, η p 2 = 0.67]; grip apertures were larger for grasping compared to manual estimations task (221, 179 mm, respectively). The main effect of size was also significant [F (1,20) = 7.7, p = 0.01, η p 2 = 0.28]. More importantly, there was a significant interaction between condition and size [F (1,20) = 25.6, p < 0.001, η p 2 = 0.56], which reflected the opposite effects of physical and illusory size differences on grasping and on perceptual estimates. An additional analysis of the movement trajectory in the bimanual grasping condition is presented in Online Appendix 1. As in Experiment 1, the grip apertures reflected the actual size differences between the objects. Mean grip apertures for the remaining conditions are presented in Table 1.

General discussion
The results show that the double dissociation between grasping and perceptual judgments, previously documented for unimanual grasping, extends to bimanual grasping control. A similar pattern of results was obtained in experiments 1 and 2. In particular, grip apertures during grasping reflected the real size differences between the objects. This pattern of results was found in the same trials in which the participants erroneously perceived the larger object in the pair as the smaller one. Therefore, in the context of both illusions, grip apertures during bimanual grasping were accurately scaled to the physical size differences between the objects. Furthermore, when participants adjusted the distance between their fingers to match the perceived size of the objects, their perceptual estimates reflected the illusory rather than the physical differences in size. These findings are in line with the results of previous studies conducted on unimanual motor control 5,9,18,44 and suggest that bimanual and unimanual grasping control may rely on shared visuomotor representations based on real rather than on relative world metrics.
Our findings are also consistent with a recent study from our lab in which we tested the adherence of bimanual grasping and bimanual estimations to Weber's law. In this study 31 , participants were asked to either grasp or estimate different-sized objects using both their hands. In line with findings from unimanual movement control, JNDs measured for estimation increased with size, adhering to Weber's law. However, JNDs during grasping did not increase with size, in violation of Weber's law. These results were interpreted based on the idea that the control of visually guided action relies on a more analytic processing style compared to perceptual processing 31,45 . The current study extends these findings by showing that bimanual grasping movements are refractory to relative, www.nature.com/scientificreports/ task-irrelevant information. These results suggest that the division of labor between the system that mediates the control of action and the system that mediates perceptual processing generalizes to bimanual motor control.
In the current set of experiments, a closed-loop design, in which vision is allowed throughout the entire trial was used both in the grasping and in the manual estimation conditions. This was done to ensure similar visual feedback in the two conditions. We note that there are pros and cons for using closed-loop and openloop designs in grasping experiments. On the one hand, open-loop designs prevent the participants from using feedback from their fingers during grasp 46 . But on the other hand, they make the task of grasping less like what happens in the real world 19,47 . Therefore, it is difficult to tell if any illusory effect during open-loop grasping is due to the prevention of visual feedback from the fingers and objects or due to the inefficiency of the task in hand 47 . In the current study we used a closed-loop design, a design that we typically use in other studies in our lab 35,48 . Therefore, the current findings reflect performance during grasping and perceptual estimations when full vision is allowed throughout the entire trial.
Another issue related to the current experimental design is the requirement from participants to grasp the objects using their both hands from its horizontal plane (see Figs. 1 and 3). We note that such a requirement may affect the ecological validity of the design because unless explicitly asked for, participants may have grasped the objects along their vertical plane using unimanual grasps rather than along the horizontal plane that requires picking up the objects with both hands. This issue could be attenuated if we would have used a set of wider objects, which would make it difficult to grasp the objects with one hand and would limit grasping choices to bimanual grasping. Yet, using wider objects would also limit their potential presentation in the context of illusions, which is why we chose the current set of objects for the two experiments. It could be argued, however, that the objects that we used can afford unimanual grasps, which may in turn lead to additional visual processing and perhaps to performing the movements in an unusual manner. Yet, we also note that previous research showed that unusual and awkward grasps are more likely to be affected by illusions 19,30 . Therefore, a potential effect of unusual grasp would be in a direction opposite to the direction of the current findings. Therefore, the current results support the idea that despite the potential effect of unusual grasp affordance or requirements, grasping movements were still performed efficiently, and could escape potential influences of illusory distortions.
There is a current debate among investigators as to the extent to which (some) visual illusions affect action control. For instance, some previous studies argued that the resistance of reach to grasp movements to the Ebbinghaus Illusion could be accounted for by methodological issues 15 . Unlike in the case of the Ebbinghaus illusion, previous studies that used the Ponzo illusion showed a more consistent pattern of a dissociation between action and perception in terms of their resistance to the illusory effect 9,18-20 . The current design allowed us to test the relative contribution of real and of illusory size differences to action and to perception 18 . Our results provide support for the idea that bimanual grasping can effectively resist illusory context.
To the best of our knowledge, the present study was the first to test the potential effect of the Wundt-Jastrow illusion on grasping movements (Experiment 2). One advantage of using this modified version of the illusion in grasping experiments (compared to other commonly used illusions) is the lack of pictorial background. According to previous studies, the presence of illusory background could be treated as an obstacle and affect the in-flight scaling of the fingers prior to grasp 7,49 (but see 50 ). It is reasonable to assume that in a similar fashion to bimanual grasping, unimanual grasping movements will resist the Wundt-Jastrow Illusion. This conclusion, however, awaits further experimentation.
Previous studies proposed that bimanual grasping control relies on a different, although partly overlapping neural circuitry, compared to unimanual grasping, and involves the encoding of different aspects of the visual scene 22,23 . The findings of the current study highlight the functional similarity between bimanual and unimanual grasping. Nevertheless, it is still possible that bimanual and unimanual motor control differ with respect to their vulnerability to perceptual illusions. Indeed, previous research has shown that type of interaction with the object could modulate vulnerability to perceptual effects 19,35,[51][52][53][54] . Future studies should explore potential differences between unimanual and bimanual control in terms of the computations that mediate these actions.
To summarize, the present findings extend the previously established dissociation between action and perception in the context of visual illusions to bimanual grasping control. The findings suggest that despite potential differences between unimanual and bimanual control, both types of actions share a common representation of the visual environment. This representation differs from the one that mediates the visual perception of objects and is more closely tuned to their actual properties.
Statement regarding experimental protocols. All experimental protocols in experiments 1 and 2 was approved by the Human Subjects Research Committee at Ben-Gurion University of the Negev (submission #1257). The study adhered to the ethical standards of the Declaration of Helsinki. Informed consent was obtained from all participants of experiment 1 and 2.