This study investigates the stability of PP-P travel time measurements using a waveform crosscorrelation method on both broadband and long-period data. This study finds correspondence between 50% of the PP-P travel times read from broadband and long-period data, but also finds 50% of the measurements differ more than 1.0 second. The inconsistent measurements are demonstrated to be due to two causes. (1) Diffraction near the core-mantle boundary causes dispersion of the P phase which yields an underestimate of the PP-P travel time determined from long-period data. This effect is identified at epicentral distances as small as 88¿ and may amount to several seconds. (2) Interference of the PP phase with secondary arrivals causes distortion of the PP waveform and produces a non-systematic measurement error. This effect are most clearly seen on broadband seismograms and may even hamper an unambiguous broadband PP-P travel time measurement. Although strong high-frequency effects associated with interference are filtered out on the long-period data, the long-period PP-waveform may still be contaminated. We conclude that the accuracy of PP-P travel time measurements is on the order of 1 s for both broadband and long-period seismograms. ¿ American Geophysical Union 1996