It is unclear whether humans can voluntarily control dynamic and static properties in knee abduction-adduction, which may be important in performing functional tasks and preventing injuries, whether the main load is about the abduction axis or not. A joint-driving device was used to perturb the knee in abduction-adduction at full knee extension under both passive (muscle relaxed) and active (muscle contracted in abduction or adduction) conditions. Dynamic control properties in knee abduction-adduction were characterized by joint stiffness, viscosity, and limb inertia, and quasi-static knee torque-angle relationship was characterized by knee abduction-adduction laxity and quasi-static stiffness (at a 20Nm moment). It was found that the subjects were capable of generating net abduction and adduction moment through differential co-contraction of muscles crossing the medial and lateral sides of the knee, which helped to reduce the abduction-adduction joint laxity (p< or =0.01) and increase stiffness (p<0.027) and viscous damping. Knee abduction laxity was significantly lower than adduction laxity (p=0.043) and the quasi-static abduction stiffness was significantly higher than adduction stiffness (p<0.001). The knee joint showed significantly higher stiffness and viscosity in abduction-adduction than their counterparts in knee flexion-extension at comparable levels of joint torque (p<0.05). Similar to dynamic flexion-extension properties, the system damping ratio remained constant over different levels of contraction, indicating simplified control tasks for the central nervous system; while the natural undamped frequency increased considerably with abduction-adduction muscle contraction, presumably making the knee a quicker system during strenuous tasks involving strong muscle contraction.