Creating stimuli for the study of biological-motion perception

MATHIAS DEKEYSER 0 1 2 KARL VERFAILLIE ) 0 1 2 JAN VANRIE 0 1 2 0 The writing of this article was supported by Concerted Research Ef- fort Convention No. GOA 98/01, the Belgium Programme on Inter- university Poles of Attraction Contract No. P4/19, and the Fund for Sci- entific Research of Flanders. We thank James Cutting for comments on a previous version of this paper. J.V. is a research assistant of the Fund garding this article to K. Verfaillie, Department of Psychology , Uni- versity of Leuven, Tiensestraat 102, B-3000 Leuven, Belgium ( 1 University of Leuven , Leuven, Belgium 2 DEKEYSER, VERFAILLIE, AND VANRIE In the perception of biological motion, the stimulus information is confined to a small number of lights attached to the major joints of a moving person. Despite this drastic degradation of the stimulus information, the human visual apparatus organizes the swarm of moving dots into a vivid percept of a moving biological creature. Several techniques have been proposed to create point-light stimuli: placing dots at strategic locations on photographs or films, video recording a person with markers attached to the body, computer animation based on artificialsynthesis, and computer animation based on motion-capture data. A description is given of the technique we are currently using in our laboratory to produce animated point-light figures. The technique is based on a combination of motion capture and three-dimensional animation software (Character Studio, Autodesk, Inc., 1998). Some of the advantages of our approach are that the same actions can be shown from any viewpoint, that point-light versions, as well as versions with a full-fleshed character, can be created of the same actions, and that point lights can indicate the center of a joint (thereby eliminating several disadvantages associated with other techniques). - Johansson (1973) discovered that a few bright moving spots under the appropriate conditions can elicit vivid and apparently effortless perceptions of a moving person involved in readily identified activities. This phenomenon is known as the perception of biological motion. Johansson initially created these stimuli by attaching small light bulbs as markers to the main joints of an actor and then video recording the moving actor in a darkened room (an educational film, made at Johanssons lab, shows examples of these stimuli; Maas, Johansson, Jansson, & Runeson, 1971). Stills from the resulting movies do not elicit impressions of a human figure. However, when the movie is played, the impression of a moving person is compelling. It has been shown that untrained viewers can recover information about the actors personal identity, gender, emotions, or personality characteristics from a point-light stimulus. Verfaillie (2000) provides a brief overview of the history of the study of biological-motion perception. Several techniques have been proposed to create pointlight stimuli. Each time a new technique was introduced, new possibilities were offered, but new problems were created as well. An overview of these features will allow OVERVIEW OF POINT-LIGHT TECHNIQUES Starting From Photographs or Films One possible strategy to create biological-motion stimuli is to start from a series of photographs or films and to place dots at strategic locations on the body on a frame-byframe basis. For instance, Mather and West (1993) used this technique with the well-known sequences of highspeed photographs of moving humans and animals, which Eadweard Muybridge made in the second half of the 19th century (Muybridge, 1955). One disadvantage of this technique is that it gives only a two-dimensional (2-D) version of a point-light action. Moreover, and this is the techniques main disadvantage, stimulus construction is extremely labor intensive. The next technique suffers less from this problem. Video Recording For his first recordings, Johansson (1973) used a technique that can be traced back at least to Marey (1894), who used it for chronophotography. Johansson used light bulbs to mark ankle, wrist, knee, elbow, hip, and shoulder joints of an actor dressed in black and filmed him or her with a video camera in a darkened room. When the movie was shown to participants in an experiment, the contrast was turned to near maximum and the brightness to near minimum, resulting in a display consisting of a dark background and bright dots. The main advantage of this technique is that it produces very natural point-light animations; the richness and complexity of real-life animate kinematics are preserved. This method shares the disadvantage associated with the previous technique, in that it produces only 2-D pointlight animations. Moreover, it has another drawback. When the actors depth orientation changes with respect to the camera, the light bulbs are sometimes occluded by the joints that they mark.1 In this case, the light bulb is no longer visible, even though the joint itself is not occluded by any other body part. For instance, suppose an actor has a light bulb attached to the outside of the right knee. When the actor is in a sagittal orientation facing to the right with respect to the camera, this knee will be marked by a visible point light. However, when the actor is facing to the left with respect to the camera, no point light on the right knee will be visible, because the knee will hide the markers light. To make the point light on the right knee visible in a leftward orientation, a point light has to be attached to the knees inside. Consequently, the orientation of the actor with respect to the camera cannot change within one take. For different actor orientations, different marker configurations have to be used. The classic solution for this problem (Johansson, 1973) is the use of retroreflective material in the form of a ribbon or tape. This material is wound around the joint or, in the case of hips or shoulders, is attached as a large patch to cover the side of the body part as well as the front and back. With this marker configuration, orientation can vary during one scene. Although this method has been quite popular (e.g., Cutting & Kozlowski, 1977), it has one important inconvenience: The projected size of a marker continuously changes, depending on the cameras angle and the actors orientation. The video-based technique (and to some extent the technique based on photographsas well) involves direct recordings of natural events. Original films or photographs are subjected to subtractive manipulations:Everything except for lights attached to the main joints of the actor are made invisible. As an experimenter, one has direct control only over the distal event (the actor and the actions in the world) and less over the proximal event (the point lights on the computer screen). Following Runeson (1994), these techniques can therefore be described as the distal approach. In a proximal approach, in contrast, biological-motiondisplays are synthesized from elemen (...truncated)