Perception & Psychophysics 2000, 62 (5), 889-899 Orientation specificity in biological motion perception MARINA PAVLOVA and ALEXANDER SOKOLOV University ofTiibingen, Tiibingen, Germany and Institute ofPsychology, Russian Academy ofSciences, Moscow, Russia We addressed the issue of how display orientation affects the perception of biological motion. In Ex­ periment 1, spontaneous recognition of a point-light walker improved abruptly with image-plane dis­ play rotation from inverted to upright orientation. Withina range of orientations from 180°to 90°,it was dramatically impeded. Using ROC analysis, we showed (Experiments 2 and 3) that despite prior fa­ miliarization with a point-light figure at all orientations, its detectability within a mask decreased with a change in orientation from upright to a range of 90°-180°. In Experiment 4, a priming effect in bio­ logical motion was observed only if a prime corresponded to a range of deviations from upright orien­ tation within which the display was spontaneously recognizable. The findings indicate that display ori­ entation nonmonotonically affects the perception of biological motion. Moreover,top-down influence on the perception of biological motion is limited by display orientation. The ability to correctly identify an event is ofgreat eco­ mans and animals depicted in animated cartoons (Mitkin logical importance for an organism, because it supports & Pavlova, 1990), despite the highly reduced and unusual functional behavior in a continuously changing environ­ structural information available. By 3-5 months of age, ment. Since the classic work ofGestalt psychology, there infants can discriminate a computer-simulated point­ has been phenomenological evidence for veridical per­ light walker from a similar display in which local rigid­ ception of simple events depicted only by rigid motions ity between dots is perturbed (Bertenthal, Proffitt, & of several points. As will be shown later, the visual sys­ Kramer, 1987) or from displays ofidentical absolute mo­ tem is highly sensitive to information about the invariant tions with scrambled spatial relations between the moving structure in complex everyday events, such as biological points (Bertenthal, Proffitt, & Cutting, 1984). Even cats motion patterns that consist ofa set ofmoving dots on the (Blake, 1993) and bottlenose dolphins (Herman, Morel­ main joints ofan invisible walker. Samuels, & Pack, 1990) are able to perceive point-light Despite the potential perceptual ambiguity, humans displays. readily extract the invariant structure from biological Humans recognize biological motion despite the distor­ motion. In his initial study, Gunnar Johansson (1973, tions caused by reverse transformation (showing the film 1976) showed that adults need only 0.1-0.2 sec to iden­ backwards) or by changing the presentation rate to about tify displays with filmed biological motion. Mather and 0.5 times faster or slower than normal (Pavlova, 1992, West (1993) extended these findings to perception ofan­ 1995). Ahlstrom, Blake, and Ahlstrom (1997) reported imated point-light figures of quadrupeds. Preschoolers that observers can discriminate between canonical and 3-4 years of age were able to recognize point-light hu- phase-scrambled point-light figures, and that they tolerate variations in dot contrast and spatial frequency. Much research has established that with upright orien­ Parts ofthe study were presented at the 12th-14th Annual Meetings tation ofa target, observers can detect a point-light figure of the International Society for Psychophysics and at the I9th-2 Ist embedded in a simultaneous moving-dot mask (Berten­ European Conferences on Visual Perception. M.P. was supported by the thal & Pinto, 1994; Cutting, Moore, & Morrison, 1988; Deutsche Forschungsgemeinschaft (436RUS 17/14/97)and by the Max­ Planck-Gesellschaft. Parts of the paper were done at the Institut fur Mather, Radford, & West, 1992; Neri, Morrone, & Burr, Arbeitsphysiologie, Dortmund, the Institut fiir Psychologie, Christian­ 1998; Thornton, Pinto, & Shiffrar, 1998). Yet,when a dis­ Albrechts-Universitat zu Kiel, and the MPI for Biological Cybernetics, play was presented for less than 0.8 sec, the ability to de­ Tiibingen. We thank Niels Birbaumer, Isabelle and Heinrich Biilthoff, termine the apparent direction (facing) ofa camouflaged C. Richard Cavonius, Walter H. Ehrenstein, Dieter Heyer, Laurence walker fell to chance level (Cutting et aL, 1988). Maloney, Rainer Mausfeld, and Sverker Runeson for stimulating dis­ cussions and Bennett Bertenthal for providing us with the walker-gen­ On the other hand, it has recently been shown in a num­ erating program. We especially thank John C. Baird for valuable advice ber ofpsychophysical and perceptual studies that display on an earlier version ofthe manuscript and Arseny Sokolov for moral inversion dramatically impedes the perception ofbiolog­ support. We also thank two anonymous reviewers for helpful comments. ical motion. Upside-down presentation prevents infants Correspondence concerning this article should be addressed to M, Pav­ lova, Institute of Medical Psychology and Behavioral Neurobiologyl from discriminating a point-light walker from similar MEG-Center, University of Tiibingen, Gartenstr, 29, D-72074 Tiibin­ displays (Bertenthal et aL, 1984; Bertenthal et aL, 1987; gen, Germany (e-mail: [email protected]). Bertenthal, Proffitt, Spetner, & Thomas, 1985). Adults 889 Copyright 2000 Psychonomic Society, Inc. 890 PAVLOVAAND SOKOLOV and 3- to 6-year-olds failed in the recognition ofinverted (Proffitt et aI., 1984). Even infants can discriminate be­ animated cartoons (Mitkin & Pavlova, 1990; Pavlova, tween a computer-simulated canonical walker and a figure 1989). When presented through multiple apertures at without occlusion (Bertenthal et aI., 1985). On the other orientations of 0°, 90°, and 180°, only an upright-ori­ hand, the lack ofocclusion in filmed biological motion ented walker was reliably identified (Shiffrar, Lichtey, & is ineffective for its perception (Runeson, 1994), appar­ Heptulla Chatterjee, 1997). Under observation through a ently because ofredundant stimulus information in nat­ Dove prism, even performance in such a relatively easy ural displays. task as discrimination between texture-defined canonical Dynamic constraints that reflect sensitivity to a match and phase-scrambled walkers was degraded (Ahlstrom between event kinematics (configuration oftrajectories et aI., 1997). Sumi (1984) reported that inversion ofthe and velocities) and dynamics (mass- and force-related original Johansson films led to an impression of unnat­ information) may also limit nonveridical interpretations ural movements even when observers were aware of the of biological motion. Sensitivity to dynamic properties manipulation of a display or when responses were re­ revealed from kinematics was demonstrated when esti­ stricted to categories ofhuman/nonhuman motion. Previ­ mating the relative weight of moving balls in collisions ous experience with upside-down displays affected their (Flynn, 1994) or the weight ofan object lifted by a point­ recognition very little. Furthermore, inversion ofpoint­ light person (Runeson & Frykholm, 1983). Earlier work light displays disrupted the ability to judge basic emotions indirectly favors the role ofdynamic constraints in event represented in dance (Dittrich, Troscianko, Lea, & Mor­ perception. Observers were able to distinguish humans gan, 1996), types ofhuman actions (Dittrich, 1993), and in action from animated point-light puppets (Johansson, the gender of a walker (Barclay, Cutting, & Kozlowski, 1976). Gender discrimination in filmed displays signifi­ 1978). cantly decreased with a change in presentation rate (three Taken together, these findings raise the issue of the times slower than normal), mainly because ofdistortions nature oforientation specificity in the perception of'bio­ in the perception ofthe gravitational force (Barclay et aI., logical motion. The pattern of intrastimulus kinematics 1978; see also Mather et aI., 1992). Bingham, Schmidt, remains the same regardless ofdisplay orientation. Thus, and Rosenblum (1995) reported that among three view­ contrary to the assumptions ofthe computational structure­ ing conditions (upright display and observer, upright dis­ from-motion models (Hoffman & Flinchbaugh, 1982; play and inverted observer, and inverted display and up­ Sugie & Kato, 1987; Webb & Aggarwal, 1982), extracting right observer), the last one produced the most errors in structure from biological motion does not depend only on the identification of simple point-light events (such as the relative motions ofits elements defined by local rigid­ falling leaves or a pendulum). They suggested that event ity or by an elongated axis of symmetry. Obviously, the recognition depends on absolute display orientation in the perceptual system implements additional constraints for gravitational field, rather than on the relative orientation the unambiguous interpretation ofbiological motion. Two of the display and the observer. Much less is known, types of such constraints can be distinguished: (1) eco­ however, about how the perception of biological motion logical constraints that represent sensitivity ofthe visual depends on such constraints. More specifically, when the system to regularities in the environment (e.g., occlusion or natural orientation ofan event is changed, does the dis­ gravity) and (2) knowledge-based or internal constraints, crepancy between perceived