A multivariate differential-phase model based on bio-kinematic asymmetry accounts for estimation biases in point-light walker directions.

Observers exhibit systematic biases when estimating the walking direction of point-light walkers (PLWs). Previous research suggests that biomechanical information plays a key role in this process, yet the underlying computational mechanisms remain poorly understood. To address this question, we developed a multivariate differential-phase model that formalizes how biokinetic asymmetries between the left and right sides of the body contribute to the bias in the estimation of biological motion directions. We first computed asymmetry indices (ASI) for key joint angles, ranges of motion, and gait parameters across different walking directions. Hierarchical clustering identified four functional modules from these ASIs. We then constructed and compared both single-factor and multi-factor differential-phase models in their ability to fit behavioral data. Results revealed that a four-factor model integrating knee and elbow joint angles, their ranges of motion, and step length, provided the best account of direction estimation biases (highest R2, lowest AIC/BIC). These findings provide a novel computational framework for understanding how bio-kinematic asymmetries contribute to directional estimation biases in biological motion, advancing our understanding of the mechanisms underlying biological motion processing.