The difficulties involved in the accurate measurement of the flow stress of materials at high rates of deformation are reviewed, and methods of overcoming these difficulties are discussed. It is concluded that the best experimental method is that in which a split Hopkinson bar is adapted for torsion testing, and the loading pulse is approximately square in shape and has a relatively short duration. A description is given of apparatus which was developed to achieve this type of loading, the input wave being generated explosively. This wave was found to contain large amplitude fluctuations, and to eliminate these a mechanical “pulse smoother” was used. The operation of this device is described, and the test procedure, method of calibration, and possible sources of inaccuracy are discussed. Results are presented for 1100-0 aluminum alloy deformed at strain rates of the order of 800 sec−1 and are compared with those obtained at a nominal strain rate of 10−4 sec−1; the results are also compared with those obtained in compressive tests by other workers.