ON SEISMOLOGICAL INVESTIGATIONS. 49 



validity of the corrections to tables at distances 75°-95° is supported 

 by the third group. After 95° the positive residuals in — C3 indicate 

 that the suggested cori-ections to P tables are perhaps excessive ; but 

 we cannot be guided by a single earthquake alone. Moreover, these 

 corrections are still under consideration and have not been adopted. 

 One necessary preliminary was the settlement of the anomalous records 

 here discussed ; and if these can be now regarded as due to anomalous 

 i-eflections the direct P records can be re-examined with greater con- 

 fidence. There is one further point which may account for part of 

 the discordance between 0°-95° and stations beyond 95° in the above 

 table. Several stations give two readings for P ; one marked e and the 

 other marked i. Thus: — 



The first record has been taken in all cases. It seems possible 

 that e might be recorded more frequently at nearer stations, but be too 

 faint at more distant stations. But this is little more than a con- 

 jecture. 



The hypothesis of an oscillation in the graph of ^P shown in fig. 4 

 means that there is an oscillation of similar kind in the increase of 

 density of the earth as we travel downwards. The inteipretation 

 suggested is that just below the ' crust ' there is a layer of unexpectedly 

 high density, in which P travels unusually quickly, followed by a 

 return to a density which is either actually less than that of the dense 

 layer above it, or perhaps ceases to increase at the same rate. No 

 theoretical examination of such a possible change of density has yet 

 been made ; but it is perhaps worth noting as a speculation * that this 

 notable oscillation might be followed by one or more smaller ones, the 

 effects of which on the times of P (and S) might be so small as to 

 have been hitherto completely masked by accidental errors. 



Hitherto attention has been confined to P for simplicity. But the 

 earthquake just discussed now enables us to test the behaviour of S 

 with facility ; for the epicentre is apparently well determined, and we 

 have found a satisfactory coiTection to the time at epicentre. Obser- 

 vations of S will thus give us at once the proper corrections to the 

 S tables. Before examining the observations, however, let us see what 

 we can infer about S from P. The ratio of the times for S and P is 

 very nearly constant (1-80) for all distances from the epicentre. With 

 the adopted (Shide) tables it is 



thus showing a slight rise in value. But corrections to these tables 

 have been proposed, and they tend to reduce the higher values. From 

 what has already been said of the possible changes in the tables required 



* These words were written before the evidence of a second oscillation given 

 below had been detected ; in fact, before the S records had been examined at all. 



1916 K 



