AS all the continental discussions of the observations of near earthquakes have been carried out by graphical methods, and as I could not satisfy myself as to the precision obtainable by these methods, I have recently carried out a rediscussion of the principal series of data by the method of least squares. These refer to the Kulpa valley earthquake of 1909, the Wurtemberg one of 1911, the Tauern earthquake of 1923, and the Oppau explosion. The results indicate very definitely that there is an upper layer that transmits compressional waves with a velocity of 5.6 km./sec. (though a velocity of 5.4 km./sec. would fit the Oppau explosion slightly better) and a lower one where the velocity is 7.8 km./sec. In addition, the Tauern earthquake gave rise to a wave with a velocity of 6.2 km./sec., which must have travelled in an intermediate layer. The probable error of all these velocities does not exceed 0.1 km./sec. The result for the upper layer corresponds to that found for granite by E. D. Williamson and L. H. Adams. The recent work of L. H. Adams and R. E. Gibson gives a velocity of 6.4 km./sec. in basaltic glass, and of 8.4 km./sec. in dunite, at ordinary temperatures and at pressures corresponding to depths of some tens of kilometres. If we allow for the higher temperatures within the crust, the basaltic layer below the granite may be in a glassy state, as Daly has suggested, and the lower layer may well be dunite. The evidence indicates that there is no further sudden change to a depth of about 1200 km.