The 1900-km-long, trench-parallel Sumatran fault accommodates a significant amount of the right-lateral component of oblique convergence between the Eurasian and Indian/Australian plates from 10 ¿N to 7 ¿S. Our detailed map of the fault, compiled from topographic maps and stereographic aerial photographs, shows that unlike many other great strike-slip faults, the Sumatran fault is highly segmented. Cross-strike width of step overs between the 19 major subaerial segments is commonly many kilometers. The influence of these step overs on historical seismic source dimensions suggests that the dimensions of future events will also be influenced by fault geometry. Geomorphic offsets along the fault range as high as ~20 km and may represent the total offset across the fault. If this is so, other structures must have accommodated much of the dextral component of oblique convergence during the past few million years. Our analysis of stretching of the forearc region, near the southern tip of Sumatra, constrains the combined dextral slip on the Sumatran and Mentawai faults to be no more than 100 km in the past few million years. The shape and location of the Sumatran fault and the active volcanic arc are highly correlated with the shape and character of the underlying subducting oceanic lithosphere. Nonetheless, active volcanic centers of the Sumatran volcanic arc have not influenced noticeably the geometry of the active Sumatran fault. On the basis of its geologic history and pattern of deformation, we divide the Sumatran plate margin into northern, central and southern domains. We support previous proposals that the geometry and character of the subducting Investigator fracture zone are affecting the shape and evolution of the Sumatran fault system within the central domain. The southern domain is the most regular. The Sumatran fault there comprises six right-stepping segments. This pattern indicates that the overall trend of the fault deviates 4¿ clockwise from the slip vector between the two blocks it separates. The regularity of this section and its association with the portion of the subduction zone that generated the giant (MW9) earthquake of 1833 suggest that a geometrically simple subducting slab results in both simple strike-slip faulting and unusually large subduction earthquakes. ¿ 2000 American Geophysical Union