If a rod is twisted while subjected to longitudinal compression it will support without fracture angles of twist many‐fold greater and maximum torques somewhat greater than is possible in the absence of load. Under compressional load the curve of shearing stress against shearing strain rises to a maximum and then sinks with a long drawn out tail to an approximate asymptote. Fracture is never complete, but some coherence always remains, probably due to cold welding. The maximum torque is not marked by any visible beginning of fracturing or other discontinuity. The strain hardening curve in torsion, therefore, under proper circumstances passes through a maximum. The whole mechanism of strain hardening appears to be different in torsion and in tension, and reasons are given for anticipating such a difference because of the difference of the atomic kinetics in torsion and tension. It is shown in particular that the method of correlating tension and torsion through the ``octahedral'' coordinates which is applicable for small strains is not applicable to the large strains which are the subject of the present discussion. It is shown that the equations of conventional plasticity theory correctly reproduce certain qualitative aspects of the secondary longitudinal and radial flow which accompany twisting, but it is possible to establish large failures of isotropy not covered by the elementary theory. With regard to fracture, it is necessary to distinguish sharply between fracture in tension and in shear. The latter is not clean cut and it is probably possible to realize a continuous gradation of atomic disorganizations, culminating under proper conditions in complete shearing fracture.