We have made over ten thousand measurements of PP-P and SS-S differential travel times from long-period Global Digital Seismograph Network recordings of all events with mb≥5.5 which occurred during the years 1976 to 1986. We have obtained reasonable global coverage, with very good sampling of several regions of the world. The residuals range from -5 to 5 s for PP-P and -12 to 12 s for SS-S and reveal a large-scale pattern of heterogeneity when plotted at the reflected phase's bounce point. Our experiments indicate that lower-mantle structure and source-received structure can each contribute approximately ¿0.5 s to our measured PP-P residuals so there is considerable signal to be explained. The pattern observed in the PP-P measurements is similar to the pattern observed in the SS-S measurements, with the SS-S residuals 2 to 4 times larger in magnitude. Comparisons of our measured residuals to those predicted by the upper-mantle models of Woodhouse and Dziewonski show that the overall patterns are quite similar but the amplitude of the model residuals is roughly a factor of 2 too small. Comparisons with the predictions of a whole-mantle model of Tanimoto again shows that the predicted pattern of residuals is reasonably consistent with the observations but now the predicted residuals are too large by about a factor of 2. This variation in predicted amplitudes is probably mainly due to differences in the near-surface structure of the models.

We have also binned our measurements according to the tectonic regionalization GTR1 of Jordan and find a qualitative correlation of average residual with tectonic region. In particular, Precambrian shields show a strong anomaly, and there is a correlation of residual size with the age of oceanic crust at the bounce point. For all tectonic regions the ratio of SS-S to PP-P residuals is approximately 2. This ratio is consistent with a thermal origin for the observed signal. Finally, our measurements show no compelling evidence for azimuthal anisotropy which might be related to fossil spreading direction or the direction of absolute plate motion. Such a signal is probably overwhelmed by the large signal from three-dimensional structure. ¿1991 American Geophysical Union