The aim of this study was to determine the extent to which the resistance provided by variable-cam resistance training machines match joint torque capability. Eight variable-cam knee extension machines from six different manufacturers were assessed. Resistive torque for a constant weight/load was measured at five knee joint angles. Knee extensor muscle torque capability--the angle-torque relationship--of 10 healthy young men was determined isometrically and dynamically at the same five angles. After normalization, the two relationships were compared. The angle - torque relationship of the knee extensors displayed an inverted "U" shape as expected, with dynamic torque changing by +40% on the ascending limb (1.75 - 1.40 rad) and -60% on the descending limb (1.05-0.35 rad) of the angle - torque relationship. The angle-torque relationship of the different training machines was highly variable, but consistently less curvilinear and significantly different from knee extensor capabilities, with changes in torque of +2.5 to +22.2% (ascending limb) and +37.6 to -20.5% (descending limb). It is recommended that variable-cam resistance training machines be designed with the angle - torque relationship of the relevant joint and the inertial profile of the lifting exercise in mind.