ON SEISMOLOGICAL INVESTIGATIONS. 57 



value have all the appearance of accidental errors. They are as follows 

 in units of '001 : — 



Table XXI. 

 Differences from the ratio 1"800 for ratio S/P in units of "OOl. 



Of the largest residuals that at A = 33° is due to the sudden rise of the 

 P residual to + 1* between two values of — 11^ ; a rise not confirmed by 

 S and probably spurious. The rise of P to +8^ at 48° also bears the 

 mark of accident. At 98° the correction of -57* to the S tables is 

 probably too large. Looking at the residuals in Table XX. we see that 

 they are probably made up of two groups, separated by an interval of at 

 least 65-\ One group, probably the true S, would have a mean cor- 

 rection of -57^ + 30*= -27« say, and the other of -57«-305= -879 

 say. This latter is probably the Y phenomenon beginning to declare 

 itself. With this interpretation the — 29 residual would become + 7. 



Hence it may be that we should do well to adopt a constant ratio 

 I'SOO, thus strengthening the determinations of both P and S by 

 the tie. 



Let us now examine very briefly the values of either P or S near the 

 epicentre. They are clearly affected much in the same way, and one 

 of them will suffice ; say P. We may, however, use the values of S, 

 reduced in the ratio 1"80, to strengthen the detemiination of P. Thus 

 we have : 



A =8° 13° 18° 23° 28° 23° 



Corrected P 118 205 257 308 357 417 

 From S 122 203 257 308 361 407 



Mean 120 204 257 308 359 411 



Mean 5? 150 IBS 106 10-2 102 10-4 



It seems difficult to avoid a sensible rise of the average 8P up 

 to A =10°. The 16"8 is only an average value, and the maximum 

 must be greater still. This rise in value cannot be explained by any 

 reasonable supposition as to the depth of the focus : for though this 

 provides an initial rise in value, the rise is very slight. We are driven 

 to suppose some important change in density just within the surface 

 of the Earth. We can avoid this supposition in two ways only: — 



(a) By discrediting the observations. On this head nothing more 

 need be said : the evidence is before us. 



(b) By adding a constant to the whole tables both for P and S. 



If we add (say) 20^°, then the mean SP for the first 8° would 

 be 140/8 = 17s-5, greater than the 16«-8 which follows. Even then the 



