Tsunamis generated by the great Indonesian earthquakes of 26 December 2004 (Mw 9.3) and 28 March 2005 (Mw 8.6) produced high-frequency (>5 mHz) dispersed signals recorded by hydrophone stations offshore from Diego Garcia and Cape Leeuwin, Australia, and by many seismic stations in the Indian Ocean and on the coast of Antarctica. For the first and greater earthquake, the dispersed energy is commonly observed to 30 mHz and, in one case, to 60 mHz. The high-frequency signals are consistent with being generated at pointlike sources. The earliest arriving tsunami signals originate from a point near 4.3° N, 93.8° E, determined using event- to-station distances estimated by matching predicted dispersion to observations. The location is further constrained with azimuth estimates from an array of hydrophones. Fine structure in the tsunami signal indicates a second high-frequency source just south of Great Nicobar Island near 6.5° N, 93.6° E. The point sources are located close to the maximum slip area determined in several other seismic and tsunami studies. The dispersion of much later-arriving energy is consistent with travel over longer paths, and matches predictions for reflections of the tsunami from bathymetric features in the Indian Ocean basin. For the 28 March 2005 tsunami, the high- frequency dispersion is observed at the Diego Garcia hydrophone station and the AIS seismic station, and tsunami signals without apparent dispersion are seen at four other seismic stations. Phase velocities estimated at hydroacoustic stations agree with linear dispersion theory at frequencies above 12 mHz.