Many functional tasks require regulating appropriate forces or torques even under unpredictable disturbances. However, how this regulation can be achieved remains poorly understood. Limb impedance describes the relationship between externally imposed displacements to the limb and the changes in force or torque generated in response. Low limb impedance is preferred during torque regulation tasks. However, low-frequency impedance increases with muscle activation, which is counterproductive to torque regulation. The purpose of this study was to quantify the ability to voluntarily reduce limb impedance during torque regulation tasks, and to assess if the observed performance is near optimal given the challenges posed by activation-dependent muscle properties and time delays in the neuromuscular system. By examining elbow impedance measured in experiments and predicted by a biomechanical model with an optimal controller, our results demonstrated that individuals can reduce the low-frequency components (below 1Hz) of elbow impedance during forceful contractions, and that this performance is similar to those predicted by an optimal feedback controller. These findings suggest that neural feedback can compensate for intrinsic muscle properties in a near-optimal manner, thereby allowing torque to be regulated at frequencies below ∼ 1 Hz.