This paper measures and analyzes the saturation effects on torque-and current-slip curves of three squirrel-cage induction motors. Motor saturation is illustrated by three sets of measurements: I) torque and current for each slip measured at three voltage levels; II) short-circuit impedance measured at different current levels; and III) no-load impedance measured at different voltage levels. In test I, torque and current measured at reduced voltage are prorated to full voltage for comparison purposes. A double-cage model is used to model the motors, and the nonlinearity of their reactances is examined. In order to evaluate the individual weight of every nonlinear reactance on the observed saturated behavior, we try to fit the measurements considering that only one reactance of the double-cage model is nonlinear. Good agreement is obtained only when the stator leakage reactance is considered nonlinear. On the contrary, when magnetizing or remaining leakage reactances are considered nonlinear, they predict machine behaviors inconsistent with measurements. In other words, the saturation of the stator leakage reactance is the main contributor to torque and current behavior, while the saturation of the remaining reactances has a negligible influence. Consequently, when all voltage levels are considered, the squirrel-cage induction motors can be accurately modeled (with reasonable accuracy) with a double-cage model where only the stator leakage reactance is considered nonlinear. When only a constant voltage level is considered, the paper also proves that a linear double-cage model (all reactances are linear) accurately predicts machine behavior at such voltage level.