Center of mass states render multi-joint torques throughout standing balance recovery

Successful reactive balance control requires coordinated modulation of hip, knee, and ankle torques. Stabilizing joint torques arise from feedforward neural signals that modulate the musculoskeletal systems intrinsic mechanical properties, namely muscle short-range stiffness, and neural feedback pathways that activate muscles in response to sensory input. Although feedforward and feedback pathways are known to modulate the torque at each joint, the role of each pathway to the balance-correcting response across joints is poorly understood. Since the feedforward and feedback torque responses act at different delays following perturbations to balance, we modified the sensorimotor response model (SRM), previously used to analyze the muscle activation response to perturbations, to consist of parallel feedback loops with different delays. Each loop within the model is driven by the same information, center of mass (CoM) kinematics, but each loop has an independent delay. We evaluated if a parallel loop SRM could decompose the reactive torques into the feedforward and feedback contributions during balance-correcting responses to backward support surface translations at four magnitudes. The SRM accurately reconstructed reactive joint torques at the hip, knee, and ankle, across all perturbation magnitudes (R2>0.84 & VAF>0.83). Moreover, the hip and knee exhibited feedforward and feedback components, while the ankle only exhibited feedback components. The lack of a feedforward component at the ankle may occur because the compliance of the Achilles tendon attenuates muscle short-range stiffness. Our model may provide a framework for evaluating changes in the feedforward and feedback contributions to balance that occur due to aging, injury, or disease. NEWS AND NOTEWORTHYReactive balance control requires coordination of neurally-mediated feedforward and feedback pathways to generate stabilizing joint torques at the hip, knee, and ankle. Using a sensorimotor response model, we decomposed reactive joint torques into feedforward and feedback contributions based on delays relative to center of mass kinematics. Responses across joints were driven by the same signals, but contributions from feedforward versus feedback pathways differed, likely due to differences in musculotendon properties between proximal and distal muscles.