Ring faults in volcanoes have been recognized for a long time, but their mechanics of formation is still poorly understood. While the subsidence on a ring fault during a large eruption from the associated chamber is easily understood, the initiation of the fault itself has been difficult to explain. For a ring fault to form, the tensile and shear stresses at the surface of the volcano must peak at a certain radial distance from the surface point above the center of the chamber. Empirical evidence, however, shows that during most periods of unrest in active volcanoes the stress field is not favorable for the initiation of ring faults. We made several tens of boundary-element models of magma chambers of various geometries and subject to different loading conditions. The results indicate that overpressure or underpressure in the chamber as the only loading is unlikely to initiate ring faults. For a spherical chamber subject to horizontal extension or doming, and a sill-like chamber subject to horizontal extension, the tensile and shear stresses at the surface of the volcano peak at a certain radial distance from the surface point above the center of the chamber. Then, however, the maximum stresses occur at the boundary of the chamber itself, which would normally lead to sheet injection rather than to ring-fault formation. By contrast, a sill-like magma chamber subject to doming gives rise to a stress field suitable for the initiation of a ring fault.¿ 1997 American Geophysical Union