The authors design a global adaptive output feedback control for a fifth-order model of induction motors, which guarantees asymptotic tracking of smooth speed references on the basis of speed and slater current measurements, for any initial condition and for any unknown constant value of torque load and rotor resistance. The proposed seventh-order nonlinear compensator generates estimates both for the unknown parameters (torque load and rotor resistance) and for the unmeasured state variables (rotor fluxes); they converge to the corresponding true values under persistency of excitation which actually holds in typical operating conditions, In these cases the rotor flux modulus asymptotically tracks desired smooth reference signals which allows the motor to operate within saturation limits (so that modeling assumptions are met). As in field-oriented control, the control algorithm generates references for the magnetizing flux component and for the torque component of stator current which lead to significant simplifications for cut-rent-fed motors. Simulations show that the proposed controller iis suitable for high dynamic performance applications. This is confirmed by experiments since the control is robust with respect to modeling errors, sensors and actuators noise, control discretization, and simplification. Experimental comparisons with a classical indirect field-oriented control exhibit a significant improvement of power efficiency when rotor resistance differs from its nominal value.