We interpret the three-dimensional seismic wave-speed structure of the Australian upper mantle by comparing its azimuthal anisotropy to estimates of past and present lithospheric deformation. We infer the fossil strain field from the orientation of gravity anomalies relative to topography, bypassing the need to extrapolate crustal measures, and derive the current direction of mantle deformation from present-day plate motion. Our observations provide the depth resolution necessary to distinguish fossil from contemporaneous deformation. The distribution of azimuthal seismic anisotropy is determined from multi-mode surface-wave propagation. Mechanical anisotropy, or the directional variation of isostatic compensation, is a proxy for the fossil strain field and is derived from a spectral coherence analysis of digital gravity and topography data in two wavelength bands. The joint interpretation of seismic and tectonic data resolves a rheological transition in the Australian upper mantle. At depths shallower than V150^200 km strong seismic anisotropy forms complex patterns. In this regime the seismic fast axes are at large angles to the directions of principal shortening, defining a mechanically coupled crust^mantle lid deformed by orogenic processes dominated by transpression. Here, seismic anisotropy may be considered 'frozen', which suggests that past deformation has left a coherent imprint on much of the lithospheric depth profile. The azimuthal seismic anisotropy below V200 km is weaker and preferentially aligned with the direction of the rapid motion of the Indo-Australian plate. The alignment of the fast axes with the direction of present-day absolute plate motion is indicative of deformation by simple shear of a dry olivine mantle. Motion expressed in the hot-spot reference frame matches the seismic observations better than the no-net-rotation reference frame. Thus, seismic anisotropy supports the notion that the hot-spot reference frame is the most physically reasonable. Independently from plate motion models, seismic anisotropy can be used to derive a best-fitting direction of overall mantle shear. ß