ON SEISMOLOGICAL INVESTIGATIONS. 271 



These results had just been tabulated when a letter was received from Dr. Harold 

 Jeffreys calling attention, in enthusiastic terms, to Prof. Gutenberg's paper Erdbeben- 

 wellen Vila, in Gott. Nach. 1914, and it was at once seen that the readings tabulated as 

 S refer to Gutenberg's ray ScPcS; that is, a ray which travels as S until it reaches the 

 liquid core of the earth, is then transformed into P, and finally emerges as S. Since 

 the middle part of its path is described with the velocity of P, which is greater than that 

 of S, it naturally arrives before S. The figures given by Gutenberg compare with the 

 adopted tables lor S as below : — 



A 54 65 70 77 79-5 87°0 94°5 102 



It will be seen that throughout the range A = 80° to A «= 1 10°, from which thef ormula 

 (80° — a)x4-6s. was deduced, the difierence between it and the value of ScPcS— S 

 assigned by Gutenberg is constant at about +7s. It changes a little for values of A 

 back to 54°, but this only means that the formula for the difference from S is only 

 approximately linear ; and it is rather remarkable that the approximation should be 

 80 close. In this region ScPcS follows S, and is not very likely to be recorded. 



But the large negative residuals from S were noticed in 1917 in discussing the 

 observations of 1913 {The Large Earthquakes of 1913, B.A. Seism. Ctee., 1917). 



The last reference need not, however, be followed here. The main point is that for 

 this particular earthquake nearly all the thirty -five observatories between a =80° 

 and A = 110° which attempt to record S, record ScPcS instead. There are only two 

 possible exceptions, Bidston (+98s) and Honolulu { + 33s), and of these the latter has 

 a positive residual not much greater than the — 29s for Barcelona. Since for these 

 values of A ScPcS precedes S, and S is generally counted as the earhest indication of a 

 change from P, it is easy to see how the earUer ScPcS is preferred to the real S, provided 

 it is large enough. [For values of A less than 80° ScPcS follows S according either to 

 Gutenberg's theory or the above suggested empirical formula ; and that S should then 

 be preferred is equally natural.] 



But it is not always that ScPcS is large enough to be mistaken for S, as can be seen 

 from the counts of residuals for the principal earthquakes in the five years 1918-1922 

 now under discussion. It will suffice here to give the counts for every 10 sec. in S and 

 every 5° of a , and to restrict ourselves to the groups near the maximum. 



Nos. of S residuals for the years 1918-1922 :— 



s. s. s. s. s. s. s. S. S. 8. 



Thus for values of A between 75° and 85° the maximum is definitely near zero ; 

 there are two large groups enclosing zero, and the fall on either side is rapid. Between 

 85° and 90° there are five comparable groups near maximum ; from 90° to 95° there 

 are seven. This is the effect of the double max. for S and ScPcS. But then the S max. 



