Contributing Factors of Person Recognition in Natural Environments
33.566, Sunday, 18-May, 8:30 am - 12:30 pm, Pavilion
Carina A. Hahn1, P. Jonathon Phillips2, Alice J. O'Toole1; 1The University of Texas at Dallas, 2National Institute of Standards and Technology
Little is known about recognizing moving people in naturalistic environments. We examined recognition as it occurs as a person approaches from a distance. Participants were visually familiarized with 30 identities. Recognition was tested subsequently with 8s videos of 60 people (half familiar) walking toward a camera (see supplementary material). The quality of identity information in the face and body varied naturally across the video sequence as the viewing distance changed. With this paradigm, we examined: 1) how the accumulation of diverse identity-related information affects recognition accuracy, and 2) the relative contribution of the face and body on recognition over the video’s temporal sequence. In the whole-video recognition test, participants viewed entire videos, making familiarity judgments at three equally spaced time-points. Performance (d’) was above chance and improved as people advanced toward the camera. In the segment test, participants viewed only one part of the video (beginning, middle, or end). Combined, the results from these two conditions indicated that the time of the response (beginning, middle, or end) predicted accuracy, rather than the amount of video seen. This suggests that the video segment immediately preceding a response determined recognition accuracy, with no evidence that people accumulated identity information across segments. Next, we examined the contribution of faces and bodies to recognition by blurring faces or bodies at test. Accuracy with faces-only was substantially better than with bodies-only and equaled performance in conditions where the whole person was seen. Notably, accuracy with faces-only improved as people approached the camera, whereas accuracy with bodies-only remained stable but above chance (see supplementary material). In summary, when viewing someone approaching in natural environments, people do not incorporate identity information over time. Instead, they rely primarily on close-up views of the face and the information viewed most recently, even with useful identity information across viewing distances.