on i Ln ee 
ON SEISMOLOGICAL INVESTIGATIONS. 9 
WIII. Tables for P and S. 
‘The Large Earthquakes of 1913’ were collated and printed in a 
special pamphlet of xii+74 pages. In the Xli pages of introduction a 
provisional analysis of the residuals for P and § is given ; but, the deduc- 
tions there made not proving workable, a new analysis of all the 1913 
and 1914 material has been undertaken, in which the residuals for different 
types of machine were kept separate. The help of Dr. G. W. Walker, 
F.R.S., in pointing out some errors and unworkable deductions is grate- 
fully acknowledged. 
The correction of greatest importance refers to the identification of 
S at distances exceeding 90°. The provisional analysis of the 1913 results 
shows that there are several phenomena which have been confused as 8, 
put are really separate. They were denoted First Set, Second Set, &c. 
The First Set lay near the adopted tables, but there were few of them ; 
the Second Set, arriving about a minute earlier than the First at A=100°, 
is favoured by the great majority of records from 4=90° to 105° (115 
records against 29), and for this reason was assumed to be the true 8. 
This assumption led to the inference that the times of transmission for 
S and P became nearly constant beyond 4100", and an explanation was 
suggested why they became faint or even disappeared (p. vii). But 
Dr. G. W. Walker pointed out some grave objections to this identification, 
which must clearly be given up, with its consequent inferences. The 
true § is the ‘ First Set,’ not far from the existing tables; the “Second 
Set’ is probably the Y phenomenon to which attention was called in the 
20th Report, and it follows that most observatories have recorded Y in 
mistake for S; only at Pulkovo and in cases where special care is taken 
has the true § been identified. The examples given in the 20th Report 
will serve to show how readily the mistake may be made. Y comes before 
S; and if the first big movement is taken, it is natural to record Y. 
The discussion in the introduction to the 1913 earthquakes was made 
in terms of S—P, under the assumption that the ratio of S to P was nearly 
constant. But this assumption was one of the faulty consequences of the 
wrong identification of S, and falls with it. Hence a new discussion of 
S and P separately was undertaken, using the whole of the 1913 and 1914 
material together. The data were written on cards which could be 
arranged in various ways, and it was determined to separate the different 
types of machine. It was found that there was enough material to find 
the errors of the tables from Galitzin machines alone, and even more for 
Wiechert machines alone. Accordingly these two determinations were 
made and are given side by side in Table I. 
The first column shows the mean A for the group. On the cards A 
was entered to 1°, and a long list of residuals was made for every 1°; but 
examination showed that little was gained by grouping in less than 5° sets. 
Under G are shown the mean residuals for Galitzin machines, followed by 
the number of records, and under W the corresponding numbers for 
Wiechert machines. In forming these means obvious mistakes were 
excluded ; there is no practical difficulty in doing this except near A=90° 
for 8, to which we must devote special attention. 
_ It will be seen that there is a systematic difference W—G of + 3°3s. for P 
and +5:5s. for S. lf these mean values be subtracted from the columns 
for W-G, the numerical mean of the residual differences is +2°3s. for P 
and +3'5s. for S. 
