Multimodal Bayesian combination of visual information about object size with observation of an actor: cue integration by the mirror neuron system?
Monica Gori, Allessandra Sciutti, Giulio Sandini, David Burr
Poster
Time: 2009-07-01 09:00 AM – 10:30 AM
Last modified: 2009-06-04
Abstract
Purpose of the study: Recent results suggest that humans can integrate optimally information from multimodal systems. In this study we investigated whether a) optimal visuo-haptic integration can extend outside the haptic workspace; and b) if integration can occur between our visual signals of object size with those resulting from observing an actor grasping the object.
Methods: We measured size discrimination of physical spheres of different diameter (varying around 50 mm), positioned at varied distances from the subjects (42-120 cm). In the first experiment subjects reported (in 2AFC) which of two spheres appeared bigger, using only visual information. The standard was always presented at 60 cm distance, while the comparison could be positioned at the same or different distance (in random order). In separate sessions, we asked subjects to match with their index and thumb the dimension of a sphere, and measured by Optotrak aperture between the two fingers. The size discrimination judgement and the matching task were afterwards repeated in three other conditions: subjects observe the spheres, and also an actor reaching to and grasping the object; an actor reaching to and touching the object; or a static hand holding the object. We also measured performance with conflicting visuo-haptic information, to evaluate the integration.
Results: Our results show that subjects overestimate object size for distances larger than their haptic workspace. The bias was also observed when subjects responded by matching their grip size. Interestingly, the bias decreased and precision improved when subjects observed an actor grasping: the control conditions (pointing and static hand) had little or no effect.
Conclusion: Our results suggest that object outside the haptic workspace are perceived larger than they should be, increasing as a function of distance. However, this internal bias can be reduced by observing of an actor grasping the object. The Bayesian ideal-integration model predicts well both the abolition of overestimation, and the improvement in thresholds. We suggest that the “Mirror system� could mediate this process.
Methods: We measured size discrimination of physical spheres of different diameter (varying around 50 mm), positioned at varied distances from the subjects (42-120 cm). In the first experiment subjects reported (in 2AFC) which of two spheres appeared bigger, using only visual information. The standard was always presented at 60 cm distance, while the comparison could be positioned at the same or different distance (in random order). In separate sessions, we asked subjects to match with their index and thumb the dimension of a sphere, and measured by Optotrak aperture between the two fingers. The size discrimination judgement and the matching task were afterwards repeated in three other conditions: subjects observe the spheres, and also an actor reaching to and grasping the object; an actor reaching to and touching the object; or a static hand holding the object. We also measured performance with conflicting visuo-haptic information, to evaluate the integration.
Results: Our results show that subjects overestimate object size for distances larger than their haptic workspace. The bias was also observed when subjects responded by matching their grip size. Interestingly, the bias decreased and precision improved when subjects observed an actor grasping: the control conditions (pointing and static hand) had little or no effect.
Conclusion: Our results suggest that object outside the haptic workspace are perceived larger than they should be, increasing as a function of distance. However, this internal bias can be reduced by observing of an actor grasping the object. The Bayesian ideal-integration model predicts well both the abolition of overestimation, and the improvement in thresholds. We suggest that the “Mirror system� could mediate this process.