T-phase spectrograms and spectra derived from deep-ocean hydrophones for tsunamigenic and non-tsunamigenic earthquakes in the Pacific reveal the following: (1) substantial amounts of T-phase energy at high frequencies (i.e., 1 Hz to at least 35 Hz); (2) strong correlations of spectral strengths in the 10 Hz to 35 Hz frequency range to seismic moments; and (3) large spectral strengths at high frequencies for tsunamigenic earthquakes. These findings are consistent with the low attenuation of the SO und Fixing and Ranging (SOFAR) channel which permits efficient propagation of high frequencies to great distances and with the perception that large earthquakes may generate high frequency signals at the water sediment interface which can best be observed by hydroacoustic sensors in the low noise environment of the deep oceans. In addition, these studies confirm that: (1) source bathymetry can strongly influence the character of T-phases; (2) islands or seamounts near the source area can act as secondary radiators of energy which add to the complexity and duration of the coda; (3) there appears to be no significant dependence of T-phase strength on earthquake source mechanism; and (4) the strength of the T-phase is dependent on the purity of its oceanic travel path. The strong correlation of T-phase spectral strengths to seismic moments suggests that substantial improvements in the reliability of tsunami warnings may be possible by using T-phases as an additional, independent, single station method for estimating seismic moments. Finally, the high frequencies observed in the T-phase may be associated with important alternate mechanisms for tsunamigenesis other than faulting (i.e., slumpage, compaction, or liquefaction).