Our study has shown that the tensile bar or dumbbell gives higher values for failure stress and strain than does either the ring or oval specimen. This should be expected since the tensile bar permits a primarily uniaxial test and one without the stress-concentrating pins used for the ring and oval. However, it is very difficult to obtain accurate low-strain data with the tensile bar. Also, the strain rate is not constant because of tab deformation. The ring, oval, and tensile bar show agreement for values of stress and strain above 100 per cent strain but below failure. For soft, low modulus stocks the ring and oval agree at low strains. For stock with tensile modulus values above 1000 psi, the bending stress and strain, required to straighten the ring to its in-test configuration, significantly affect the data. There is less ambiguity in reducing the data from tests run with the oval than from those run with the ring, whether reduction is done manually or through direct acquisition by computer. This is again because of the initial straightening of the ring. For both the ring and oval, the material in contact with the pin experiences non-uniaxial deformation and becomes the site of failure. Yielding, as defined in this paper, can be related to hysteresis effects and hence to the phenomenon known as cyclic softening. The stress at 300 per cent strain and the stress and strain at failure are not sufficient to characterize stress-strain behavior, particularly for thermoplastic elastomers.