Most of the seismic moment release of the complex earthquake sequence beneath the South Sandwich Islands occurred on the central part of the SS megathrust. Significant aftershock activity indicates that the central and southern megathrust was subsequently activated, i.e., where young South America lithosphere is subducted. Seismic activity thus seems to have been restricted by the lateral termination in the south of the SS Trench.   Relatively little energy release occurred on the northern part of the megathrust. It was hypothesized by Govers and Wortel (2005) that here the South America slab breaks away from the surface part of the plate at the active STEP. Geochemical observations and earthquake P-axes orientations do not seem to agree with the hypothesis and we investigate the cause.   We show results of new physical analog lab models that aim to elucidate what controls the geometry of the lithospheric STEP Fault. We study lithospheric tearing in the process of STEP evolution, which is dynamically driven by the buoyancy of the subducting slab. In our experiments, the lithosphere as well as asthenosphere are viscoelastic media in a free subduction setup. A stress-dependent rheology plays a major role in localization of strain in tearing processes of lithosphere such as slab break-off. The results show that the highly curved northern plate boundary is a STEP Fault following from lithospheric tearing at a depth of ~100km. This is a modification of the original STEP model of Govers and Wortel (2005). This is consistent with available observations along the northern Sandwich plate boundary, and likely exists in other STEP regions. The region’s largest recorded event, the 1929 Mw 8.3 earthquake, may reflect horizontal extension perpendicular to the STEP fault, which is also expected based on our experiments.