Woodward and Masters [1991a] measured the arrival times of the SS-wave, a downward radiated shear wave reflected once at the earth's surface, roughly at the mid-point of its path between the source and receiver, and subtracted from it the travel time of the direct shear wave (S). This difference is called a ``differential travel time'' and is relatively insensitive to the structure near the source and the receiver as well as the origin time of the earthquake and its geographic location. The differential travel times SS - S plotted at the mid-path reflection points show predominantly negative residuals under the shields and stable continental platforms, strongly positive residuals near the East Pacific Rise, and relatively faster times for the old ocean floor. Such a map, particularly if lightly smoothed by the cap averaging procedure, bears a strong resemblance to the shear velocity models in the 100--300 km depth range, and closely corresponds to what one would expect from our understanding of the thermal evolution of the continental and oceanic lithosphere.

Woodward and Masters [1991b] created a similarly informative picture by the plotting the ScS - S differential travel time anomalies at the core reflection points; ScS is the shear wave reflected from the core-mantle boundary. The maps shows a distinct ring of faster then normal differential travel times under the circum-Pacific ring and two regions of strongly (up to +10 s) positive residuals under the central Pacific and Africa. Woodward [1989] showed that ScS - S computed for a scaled model [ Dziewonski, 1984] predict his measurements rather well.

Woodward and Masters [1991a, b] measured the differential travel times from the long-period waveforms recorded by the Global Digital Seismographic Network. The long period (20 s) of the waveforms assures broad averaging along the ray path and lesser sensitivity to the local structure This is responsible for the high signal to noise ratio of this data set: over 80% of the variance of the cap averaged SS - S and ScS - S residuals can be explained in global inversions.

It is also possible to measure `absolute' travel times by cross-correlating the observed pulse with the synthetic computed for a reference earth model. Substantial sets of such data have been gathered by Masters and Bolton [1991], Su and Dziewonski [1991], Su [1993] and Liu and Dziewonski [1994].

Efforts were made to extend surface wave tomography to periods shorter than those obtained in the analyses of mantle surface waves (about 150 s). Zhang [1991] made measurements of dispersion of the fundamental mode Rayleigh (P-SV) and Love (SH) surface waves on the minor arc path in a period range from 75 to 250 seconds for many thousands of paths. It seems, however, that his results differ significantly from those obtained by others [ Wong, 1989; Masters, 1992; Su et al., 1992a, 1994].

Tanimoto [1993] and Ekström et al. [1993] extended the global analysis of surface wave dispersion to periods as short as 35 seconds. This involves the development of new data processing approaches.