The load losses of an induction motor, defined as the difference between the actual losses and the conventional segregated losses under normal running conditions, have, in the past been difficult either to measure or to calculate accurately. An earlier paper gave methods of determining similar losses at larger slips and at reduced voltage, in terms of calculations and measurements of the torque/speed curve. The calculations allowed for saturation in the tooth tips, but not for saturation due to the full main flux which is present at normal voltage. The measurements were not made at full voltage because of heating considerations, and no useful measurements at normal full load were obtained because the load losses are a very small fraction of the input. In this paper, the method of calculation is modified, and improved testing methods are used to overcome these limitations, and thus to derive more reliable methods of measuring and calculating the load losses. Apparatus is described for the automatic recording of speed/torque and power-input curves at full voltage, from which the harmonic torques caused by the load losses can immediately be measured. The paper also describes a new back-to-back test for induction motors, which constitutes the most accurate and reliable method of measuring load losses under full-load conditions to date. Ten different rotors with carefully controlled parameters, used in conjunction with two identical stators, were tested in the back-to-back arrangement, and the results are compared with the theoretical computations and with the reverse-rotation test carried out on the same machines.