66 MODERN SEISMOLOGY 
The three values for the epicentre do not differ by more than 
20 km. 
The advantages of this method are that it is quite inde- 
pendent of (1) the time at the two stations, and (2) the deter- 
mination of S, and thus free from any error that attaches to 
the empirical time curves. It should thus prove of great value 
in improving the empirical time curves, more especially for 
short distances where the influence of finite depth of focus is 
considerable. For this reason I consider that an instrument 
which would give the azimuth directly would be of great 
service even if the remaining part of the seismogram had to 
be sacrificed. 
We have now to consider how the primary time curves are 
to be obtained. 
We shall suppose that we have available the times of in- 
cidence of P and S at a number of stations. Before these can 
be arranged we require to know the position of the epicentre 
so that the distances 4 can be computed. In some cases (e.g. 
the great Messina earthquake, 1908) the epicentre is known 
with considerable accuracy from local knowledge. But,in many 
cases such information is not available or cannot be relied on, 
and then some other method must be used. 
We have seen that an extension of Galitzin’s method of 
azimuths may give the epicentre directly. So far it has not 
been used in the preparation of time curves, but there is little 
doubt that it is the most satisfactory method we can have. 
When observations of P have been obtained at several 
stations known to be not very far from the epicentre, we may 
however get a fairly good determination of the position of the 
epicentre by a method used by Zoéppritz (Gott. Nach., 
1907, l.c.). If for instance P occurs at precisely the same in- 
stant at three stations not too far from the epicentre, the 
epicentre would be the unique point which is equidistant from 
the three stations. If the times differ we may proceed as 
follows: Let A, B, and C be the stations and let X be the 
epicentre ; we then have the equations 
XA=1(7), XB=2(y+)), XC=0,(7+9) 
where # and g are the observed time intervals in seconds 
