Abstract
Human observers are remarkably proficient at identifying and discriminating other humans in their perceptual environment. What representations underlie this success? We tested whether humans represent the relations between body parts using a behavioral adaptation paradigm. In Experiment 1, observers viewed digital renderings of arms, and were asked to judge the relative length of the forearm compared to the upper arm ("too long" vs "too short"). Forearms were presented at 0 (veridical), ±5, ±10, and ±15 percent scaled lengths. After completing baseline judgements, observers adapted to an arm containing a relatively shortened forearm (by 25%) for either 2 minutes or 10 minutes, after which they completed post-test judgements. Psychometric functions relating observer response to arm scaling were computed, and PSEs were estimated as the 50% point of the function. Following adaptation to a short forearm, observers in the 10 minute group perceived forearms as significantly longer (by 6%, p < 0.05), an effect which decayed over time. A smaller trend was observed in the 2 minute group (~3%, p = 0.16), consistent with past work showing adaptation, as opposed to short-term bias, becomes stronger as exposure duration increases. In Experiment 2, using similar procedures, observers were tested with arm images at the same orientation as the adaptor and with images rotated by 90 degrees. Post-test trials were preceded by 4 seconds of top-up adaptation. Following adaptation to the short forearm, observers perceived forearms as significantly longer at both the orientation of the adapter (10%, p < .01) and at the 90 degree rotation (3%, p < .05), though the effect was reduced in the 90 degree condition. Taken together, the results of Experiments 1 and 2 suggest that humans perceive bodies using representations of relative limb length, and that these representations are at least partially invariant with respect to image rotation.
Meeting abstract presented at VSS 2017