Abstract The process of faulting within a crustal-scale rift basin, subjected to transpression, is simulated in a series of small-scale experiments using sand and silicone layers. The structural scenario involved three stages: (1) extension; (2) sedimentation; (3) coeval shortening and strike-slip motion. A sandpack, placed above a basal silicone layer, was submitted to extension and produced a system of horst and grabens. Following the extension phase, the resulting surface topography was covered with silicone material, thus introducing a potential décollement between the pre- and post-rift sediments. These latter strata were made of sand. After sedimentation, deformation by transpression was applied using the same amount of shortening and an increasing strike-slip displacement in the different experiments. The strain partitioning increased with the amount of horizontal shear, and strike-slip faults developed at the more advanced stages of transpression. Two phases of faulting were observed before the horst and graben faults could be reactivated. The first phase was characterized by conjugate reverse faults, striking parallel to the rift-bounding faults, compatible with a stress regime in compression. A second phase of faulting was recognized at the end of transpression, leading to the generation of Riedel R-type strike-slip faults. The stress regimes responsible for the kinematics of fault generation and reactivation were interpreted using the Coulomb failure criterion and assumed a friction angle of 30° for the undisturbed sand. The reactivation of the steep normal faults in the horst and graben occurred only after a large strike-slip displacement. The general sequence of fault generation and reactivation suggested a temporal change in the stress regime. This change was caused by the permutation of the minimum and the intermediate principal stress axes and also by a progressive rotation, in the horizontal plane, of the axis of the maximum compressive stress. The spatial variation of the stress regime was also strongly controlled by the geometry of the interbedded silicone layer. A regular and undeformed post-rift silicone layer introduced a more efficient mechanical decoupling between the post-rift cover and the stretched basin. In summary, when a pre-existing graben was present, there was a succession of two distinct ‘tectonic phases’, whereas without a rift, the resulting fault kinematics reflected a single stress state and one tectonic phase.