Abstract
It is vitally important for humans to quickly detect living creatures in the environment, and to analyze their behavior to facilitate action understanding and high-level social inference. The current study examined the ability of human observers to spontaneously perceive complex social behaviors using spatially-scrambled point-light displays of human social interaction. Specifically, we investigated the importance of global form information, intrinsic joint movements, extrinsic whole-body movements, and critically, the congruency between intrinsic/extrinsic motions. Motion congruency provides an important theoretical constraint on biological motion due to the causal relationship between limb movements and the direction of global body motion. In Experiment 1, we discovered using a free response paradigm, that even for spatially-scrambled point-light displays from which global human form has been removed, naïve observers (55%, n=37) spontaneously reported animate/social traits in various displays that originated from social interactions (e.g. dancing, tug-of-war, high-five). However when congruency between intrinsic/extrinsic motion was violated, observers were less likely to infer socially interactive traits (37%, n=27), and were more likely to attribute physical/mechanical traits. Experiment 2 showed that observers (n=33) could accurately discriminate meaningful interaction in spatially-scrambled displays of human salsa dance (2IFC task), as long as congruency was maintained between intrinsic and extrinsic motion. Violating this constraint resulted in chance performance. In Experiment 3, observers (n=10) rated the degree of interactivity (scale 1-5) between spatially-scrambled salsa dancers. Stimuli that violated the motion congruency constraint were consistently rated as less interactive than congruent displays, regardless of whether the dancers were truly engaged in meaningful interaction or not. These results suggest a hierarchical system in which basic filters first operate to detect animate creatures based on a set of fundamental constraints (e.g. gravity, motion congruency, body structure). Only if these constraints are satisfied, higher-level processes would be engaged to support action understanding and social inference.
Meeting abstract presented at VSS 2014