Proceedings Knirsch Verlag Kirchentellinsfurt

ion of Physical Properties from Complex Object Interactions—The Case of Elasticity Manfred Nusseck, Roland Fleming and Heinrich H. Bülthoff MPI for Biological Cybernetics, Tübingen [email protected] For humans it is useful to be able to visually infer an object’s physical properties (e.g. weight, hardness or elasticity). One potentially important source of information is the way that an object moves and interacts with other objects in the environment. For example, the way that a ball bounces could inform us about its elasticity. There have been several explorations of what are the necessary and typical visual cues in a bouncing event. However, in most previous work the stimuli consisted of a ball bouncing repeatedly on a simple horizontal plane. Warren, Kim and Husney (1987) showed that under these circumstances, there are at least three heuristic cues to elasticity: relative height, relative period and relative velocity of the bounces. We wanted to test whether the visual system can interpret more complex bouncing events in which these simple cues are not present in the display. Can subjects abstract something more sophisticated from the trajectories of bouncing objects, or must they rely on these simple heuristics? To test this, we used the Virtools Physics Pack to simulate a ball falling through an array of horizontal cylindrical pegs housed in a vertical box. The ball fell from a random location above the box, bounced a number of times and finally fell out of a hole at the bottom. This stimulus allows us to completely randomize the duration that the ball needs to get through this box, the number of collisions, the velocity of the ball and the height of the rebounds, because the angle of the collisions with the pegs is always different. Subjects performed an elasticity matching task. Subjects were presented with two pegboxes simultaneously. The left-hand box was the Test ball, whose elasticity was chosen at random by the computer. The right-hand box contained the Match ball, whose elasticity could be adjusted by the subject. The subject?s task was to adjust this elasticity of the Match ball until it appeared to have the same behavior as the Test ball. The results show that subjects generally performed poorly in this task, despite dramatic variations in the elasticity of the ball. However, we found large individual differences, in which some subjects were able to perform the task above chance levels. Our results suggest that subjects normally rely on simple heuristics to estimate elasticity (e.g. bounce height), which, by design, were eliminated from our stimuli. Further research is needed to investigate which additional, complex cues were used in some cases to abstract the behaviour of the ball. 136 Visual Cognition: Objects and Faces Motion Matters: Facial Motion Improves Delayed Visual Search Performance Karin Pilz, Ian M. Thornton and Heinrich H. Bülthoff MPI for Biological Cybernetics, Tübingen [email protected] In the last few years there has been growing interest in the role that motion might play in the perception and representation of facial identity. In separate experiments, we explored how learning is affected by two different types of movement: the non-rigid motion of the face, typically associated with expression and communication, and the equally familiar rigid motion that occurs whenever a person approaches you in depth. Traditional old/new recognition tasks have previously yielded mixed results in the context of facial motion. Therefore, we decided to use a delayed visual search paradigm. Observers were familiarised with two target individuals, one seen in motion, the other via static snapshots. All images were non-degraded and consisted of video sequences in Experiment 1 and 3D heads on a walking avatar body in Experiment 2. After a delay of several minutes, observers were asked to search for their targets in static search arrays. Crucially, during this test phase, the static search arrays were identical regardless of how the face was first seen during learning. Nevertheless, in both experiments, faces that were learned in motion were found more quickly and more accurately than faces that had been learned from snapshots. These findings provide further evidence that facial motion can affect identity decisions, both in the presence of intact form cues and across extended periods of time. Visual Cognition: Objects and Faces 137 Perceived Dynamics of Static Images Enables Emotional Attribution Arseny A. Sokolov, Alexander Sokolov and Marina Pavlova University Hospital of Tübingen, University Hospital of Ulm [email protected] Perception of intentions and dispositions of others is an essential ingredient of adaptive dailylife social behaviour. Dynamics of moving images allows for veridical perception of social attributes (e.g., [1,2]). Anecdotal observations in art, science, and popular culture indicate that dynamic imbalance can be revealed in static images [3]. Here we ask whether perceived dynamics of abstract figures is related to emotional attribution. Participants first estimated instability of geometric figures rotated in 15-deg steps in the image plane, and then ranked the intensity of basic emotions that can be ascribed to the figures. No substantial link occurred between the deviation of the figures from the vertical orientation and perceived instability. Irrespective of shape, the strong positive correlation was found between negative emotions and perceived instability. By contrast, the positive emotions were inversely linked with the figure deviation from vertical orientation. The work demonstrates for the first time that dynamics conveyed by static images enables specific emotional attribution, and agree well with the assumption that neural networks for production of movements and understanding the dispositions of others are intimately tied [4]. The findings are also of importance for exploring the ability to reveal social properties through dynamics in abnormal development, for example, in patients with early brain injury [5] or autistic spectrum disorders [6]. [1] Heider, F. and Simmel, M. (1944): Am. J. Psychol. 57, 243–259. [2] Runeson, S., and Frykholm, G. (1983): J. Exp. Psychol. General 112, 585–615. [3] Cutting, J. (2002): Perception 31, 1165–1193. [4] Gallese, V.et al. (2004): Trends Cogn. Sci. 8, 396–403. [5] Pavlova, M. et al. (2005): Cereb. Cortex 15, in press. [6] Blake, R. et al. (2003): Psychol. Sci. 14, 151–157. 138 Visual Cognition: Objects and Faces Characterization of Face Processing in Patients with Congenital Prosopagnosia by Simultaneous MEG/EEG Recordings Joachim E. Weber, Claus-Christian Carbon, Martina Grueter, Tilmann H. Sander, Thomas Grueter, Gabriel Curio, Lutz Trahms and Andreas Lueschow Charité—University Medicine Berlin, Institute of Psychology, Freie Universität Berlin, University of Münster, Physikalisch-Technische Bundesanstalt, Berlin [email protected] Congenital prosopagnosia (cPA) is assumed to be a severe and selective impairment in face processing from early childhood. 13 individuals with cPA were compared to a group of matched controls by means of simultaneous measurements of MEG and EEG. Evoked responses were recorded in a viewing task with a sequence of faces and houses. The EEG/MEG evoked components N170 and its analogon the M170 are considered to be linked to structural encoding of faces rather than to later processes responsible for individual identification. It is well known for normal subjects that the M/N170 elicited by faces is larger than for other visual objects indicating intercategorical discrimination at this stage of processing. Our measurements clearly show that the M/N170 elicited by faces and houses is decreased for the prosopagnosics compared to the controls but in contrast to previous single case studies a significant amplitude difference between the two categories is preserved. This preserved intercategorical amplitude difference for the M/N170 in subjects with cPA can explain their remaining ability to discriminate faces from houses and this finding adds to existing evidence that M/N170 label a stage of intercategorical discrimination. Another interesting result concerns the latency of the M170 for faces, which peaks about 20 ms later in subjects with cPA compared to normals. I would especially like to thank the Sonnenfeld foundation for the financial support given to this study. Visual Cognition: Objects and Faces 139 Mechanisms of Recovering Shape Properties from Perfectly Mirrored Objects Ulrich Weidenbacher, Pierre Bayerl, Roland Fleming and Heiko Neumann Department of Neural Information Processing, University of Ulm, MPI for Biological Cybernetics, Tübingen [email protected] When we look at a perfectly mirrored object, such as a polished kettle, we generally have a remarkably strong impression of its 3D shape. This leads to the question of whether there is a mechanism to completely recover the shape of a mirrored object from a single static image (e.g. a photograph). Without explicit knowledge of the surrounding scene, this is theoretically impossible because many possible combinations of illumination from the surrounding scene and surface properties can generate the same image (i.e. it is an ill-posed problem). Therefore, the only way to extract information about object shape is to constrain the possible combinations of object shape and illumination. If we assume that the reflected scene contains isotropic contrast information, then there is a close relation between the surface curvature of an object (specifically the second derivatives of the surface function) and the distortions of the reflected scenery [1]. In this contribution we present two different computational methods for analysing images of mirrored objects to recover certain properties of 3D shape. Our first method is a statistical approach, based on principal components of the image gradient computed in a local neighborhood, known as the structure tensor. In this context, the eigenvectors of the tensor tell us the orientation of curvature and the eigenvalues of the tensor give us information about the anisotropy of curvature (ratio of maximal and minimal curvature). Our second method is a biologically motivated approach, based on Gabor filters and grouping. We apply an iterative refinement in a simple model of cortical feedforward/feedback processing [2]. Context information is collected by cells with long-range lateral connections. This information is fed back to enhance regions where local information matches the top-down reentry pattern provided by the larger context. Our approach shows that under the assumption mentioned above, it is possible to recover two characteristic curvature properties of mirrored objects: (i) the direction of maximal and minimal curvature and (ii) the anisotropy of curvature. Our simulations demonstrate that both methods (the statistical and biological motivated approach) lead to comparable results and that the models perform well even if the assumption of isotropic contrasts in the scenery is violated to a certain degree. [1] Fleming, Torralba and Adelson. Journal of Vision, 4, 2004 [2] Neumann and Sepp, Biol. Cybern., 81, 1999 Visual Cognition: Space Perception and Navigation 142 Visual Cognition: Space Perception and Navigation The Influence of Body Tilt on Mental Rotation and Subjective Visual Vertical Sarah D. Glaus and Thomas Jarchow Universität Zürich, Man Vehicle Laboratory, MIT, Cambridge MA, USA [email protected] According to earlier spatial-orientation studies, an A-effect occurs when an illuminated bar is adjusted to a subjectively vertical position (SVV) in the dark. The A-effect increases quasi linear up to135◦ body tilt and collapses in body tilts greater than 135◦. In mental rotation studies the question remains whether subjects respond faster to objects that are in line with physical or with retinal frame of references, e.g. when subjects are tilted and the retinal and the gravitational reference frame are misaligned. The present study combines SVV and mental rotation with the aim to (1) replicate the classical SVV pattern, and to (2) investigate whether a similar pattern can be found in reaction times of mental rotation task. In this experiment, subjects were placed in the positions of 0◦(=upright position), 45◦, 90◦, 135◦, and 180◦(=upside down) on a tilt board in a dark room. In each position, subjects performed two separate tasks: One consisted of adjustments of an illuminated bar to the subjective vertical position (SVV). In the other, subjects were asked to decide whether the writing of meaningless consonant triplets was normal or mirrored. The triplets were rotated in the plane of the writing and reaction time of subjects was recorded. Results on the SVV task are in accordance with the findings of earlier studies. Further results show that in the mental rotation task subjects responded fastest to triplets that corresponded to the respective SVV, i.e. an “A-effect” in mental rotation is found. This effect occurs up to body tilts of 90◦. At 135◦, the fastest reaction time is found for triplets oriented between SVV and retinal orientation, which could be interpreted as transition from a SVV-coordinate system to a retinal one. At 180◦, fasted reaction time is found for retinal orientation, in contrast to SVV-results. The authors conclude that two separate systems are involved in mental rotation, one in line with SVV and one with retinal coordinates, the former applying to smaller, the latter to larger tilt domains. Visual Cognition: Space Perception and Navigation 143 Can A-Priori Instructions Induce a Viewpoint Independent Mental Representation of a Dynamic Scene? Markus Huff, Bärbel Garsoffky and Stephan Schwan Knowledge Media Research Center (IWM-KMRC), Tübingen, Germany