SECT. 1] SUB -OCEANIC STRUCTURAL EXPLORATION BY SEISMIC SURFACE WAVES 121 



Clearly, all factors affecting the results have not been taken into account in 

 the studies described above on oceanic mantle structure ; yet, despite the pre- 

 liminary nature of these results, it is clear that sub-oceanic and sub-continental 

 mantle structure are not the same. Additional data are necessary to ascribe 

 the observations either to regional differences in temperature distribution or in 

 composition in the mantle. Further analysis of long-period surface- wave data, 

 particularly of phase-velocity measurements for purely oceanic paths, are 

 certain to provide additional information on the sub-oceanic mantle. 



4. Evidence from Crustal Surface Waves 



Crustal surface waves are so named because their group-velocity curves 

 have a normally dispersed branch (i.e. a segment for which group velocity 

 increases with period) as a result of lower seismic velocities of the crustal layers, 

 and, in the case of oceanic Rayleigh waves, also because of the water layer. The 

 study of dispersed surface-wave trains, such as those of crustal surface waves, 

 was handicapped, for a long time by methods of reading which apply best to ray 

 theory analysis of body waves, i.e. only the arrival time of the first motion or 

 of the maximum amplitude was read. This arrival time was related to the period 

 of the first readable oscillation or the period, of the maximum amplitude. This 

 method necessitated reading large numbers of seismograms to obtain limited 

 dispersion data with great scatter. The lower group velocities of late waves of 

 long trains were never considered. Differences in velocities of surface waves 

 for different parts of the earth were known from the early data (see Macelwane, 

 1951, and Gutenberg, 1951, for summary). However, few precise interpreta- 

 tions were attempted partly because of the scatter of the data. Moreover, 

 in the case of Rayleigh waves, the correct theory was not applied until later. 

 An important improvement in technique was made by Bullen (1939) who 

 measured period versus arrival time for a number of successive cycles in a 

 Rayleigh-wave train, smoothed the values and converted arrival time to group 

 velocity. Wilson (1940) read the several swings on oceanic Love-wave seismo- 

 grams and related each period to its proper group velocity, thus making use of 

 data from the entire train of surface waves from a single shock. Following this 

 principle, Ewing and Press (1952) plotted the number of each wave crest 

 against its arrival time throughout an entire oceanic Rayleigh-wave train and, 

 from this, obtained a smooth relation between period and group velocity, 

 clearly showing the possibility of obtaining a large number of dispersion data 

 from a single seismogram, as in Fig. 9 (which appeared in their paper). This 

 method, which is now widely used, resulted in such a striking improvement in 

 the quality of observations that most experimental surface-wave dispersion 

 data of present-day significance have been derived by it. Almost all of the 

 data reported here have been analyzed by this technique. An earlier summary 

 of results was given by Press and Ewing (1955) and a detailed review of theory 

 and results was given by Ewing, Jardetzky and Press (1957, chap. 4) and by 

 Ewing and Press (1956a). 



