New high precision PIMMS Hf and Pb isotope data for 14--28 Ma basalts recovered during ODP Leg 187 are compared with zero-age dredge samples from the Australian-Antarctic Discordance (AAD). These new data show that combined Nd-Hf isotope systematics can be used as an effective discriminant between Indian and Pacific MORB source mantle domains. In particular, Indian mantle is displaced to lower ϵNd and higher ϵHf ratios compared to Pacific mantle. As with Pb isotope plots, there is almost no overlap between the two mantle types in Nd-Hf isotope space. On the basis of our new Nd-Hf isotope data, we demonstrate that Pacific MORB-source mantle was present near the eastern margin of the AAD from as early as 28 Ma, its boundary with Indian MORB-source mantle coinciding with the eastern edge of a basin-wide arcuate depth anomaly that is centered on the AAD. This observation rules out models requiring rapid migration of Pacific MORB mantle into the Indian Ocean basin since separation of Australia from Antarctica. Although temporal variations in isotopic composition can be discerned relative to the fracture zone boundary of the modern AAD at 127¿E, the distribution of different compositional groups appears to have remained much the same relative to the position of the residual depth anomaly for the past 30 m.y. Thus significant lateral flow of mantle along the ridge axis toward the interface appears unlikely. Instead, the dynamics that maintain both the residual depth anomaly and the isotopic boundary between Indian and Pacific mantle are due to eastward migration of the Australian and Antarctic plates over a stagnated, but slowly upwelling, slab oriented roughly orthogonal to the ridge axis. Temporal and spatial variations in the compositions of Indian MORB basalts within the AAD can be explained by progressive displacement of shallower Indian MORB-source mantle by deeper mantle having a higher ϵHf composition ascending ahead of the upwelling slab. Models for the origin of the distinctive composition of the Indian MORB-source based on recycling of a heterogeneous enriched component that consist of ancient altered ocean crust plus<10% pelagic sediment are inconsistent with Nd-Hf isotope systematics. Instead, the data can be explained by a model in which Indian mantle includes a significant proportion of material that was processed in the mantle wedge above a subduction zone and was subsequently mixed back into unprocessed upper mantle.