Abstract
In this paper we will firstly consider the range of validity of Taylor's “stress-dilatancy” hypothesis. From this study we conclude that Taylor's rule is strictly valid only for coaxial deformations. For smooth non-coaxial deformations Taylor's rule must be modified according to the rule proposed by Gutierrez and Ishihara (2000), which includes a “non-coaxiality” correction factor. Strictly speaking the stress-dilatancy rule breaks down if in a process an abrupt rotation of principal axes is imposed, as this is the case at the onset of localization. During this phase the corresponding stress function decreases monotonously, whereas the dilatancy oscillates between large negative and large positive values. Finally it is shown that in globally undrained, displacement controlled tests this incipient post-failure contractancy to dilatancy oscillation should result in a dynamic instability.
