In recent years general merchandise and hardware stores have been marketing devices to be inserted in the power supply of refrigerators and other motor driven appliances that are advertised as a technology to reduce electric energy use and cost. The purpose of this research was to procure and test one of these devices. The testing included examining how the device modifies the standard 60 cps and sinusoidal voltage power supply and then to determine the corresponding change in electric power consumption by an induction motor as a function of mechanical load. The load is the product of motor speed which is nominally constant and the torque which is variable. The torque depends upon a combination of compressor characteristics and the low and high side pressures of the freon which is a function of the evaporator and condenser temperatures. Thus, to simulate the range of induction motor operating conditions for a refrigerator it is necessary to be able continuously vary and measure the torque over the range of possible operating conditions. This was achieved by connecting the motor to a shunt wound DC generator and with the stator of the generator also mounted on bearings to allow for the measurement of torque by means of a spring on a lever arm. Instrumentation included an oscilloscope to observe the voltage and current wave shapes, a strobe to measure shaft speed and also slip, Fast Fourier Transform (FFT) software to quantify the fractions of the total power at the fundamental and at higher harmonics and power measurements by both a conventional kilowatt hour meter and by sampling the current and voltage signals which are then multiplied to obtain instantaneous power and then averaged. The induction motor was then operated over a range of loads without and with the power conditioner. The power conditioner appears to operate on the basis of some sophisticated electric power engineering analysis. For the first few seconds of operation the power conditioner samples the current and voltage, but does not modify the power supply. Then, on the basis of the initial sampling it uses a proprietary neural network algorithm that is programmed on a reduced instruction set computing (RISC) microprocessor to calculate how the power supply should be modified for the particular load and motor characteristics. Our test results on our motor showed that the device saved electric energy at low to moderate loads, but resulted in higher electric energy consumption at the maximum rated load.