TIME OF EARTHQUAKE : RATE OF PROPAGATION. 293 



immediately over the point of measurement. It was subsequently 

 found to be better to place the zero line to one or other side and 

 measure the departure. 



(2) Errors in measuring quantities smaller than 10 minutes, this being the 



smallest division of the glass scale. 



(3) Error in determining the corrections to glass scale. 



(4) Error in measuring the parallax of the curve, being the small quantity 



by which the time scale overlaps the curve or vice versd. 



(5) Error in estimating the number of seconds before or after the hour at 



which the cut off falls, the driving clock being unprovided with a 

 second hand. 



Considering all these sources of error, which in the most unfavourable 

 case are cumulative, it would not be excessive to regard the probable error of a 

 single measurement as ±2 minutes, though in many cases fortuitous cancellings 

 would operate to produce results far closer to the truth. 



The accordance however of many of the results in the following pages, when 

 independent measures were taken by two or more observers, is such that there 

 should be no hesitation in accepting some groups as correct at least to 1 minute 

 time. 



There is moreover a check on the times found for each observatory by 

 comparing the intervals derived from the time of shock at each observatory with 

 the intervals obtained by measuring the interval between the shocks and the well 

 marked apices of disturbances on the same date. This method assumes that the 

 times of disturbance are simultaneous but, in view of the measurements of the 

 disturbances and the fact that such disturbances are proved to be simultaneous over 

 large areas, the assumption is not unwarranted. Two apices of disturbance have 

 been utilized and the measurement has been made in two ways. 



An additional advantage applies to these methods in that error of cut off 

 parallax and error of clock are not involved. 



In the first the absolute time of the apex of disturbance is found and the 

 time interval derived by applying the absolute time of the shock already deter- 

 mined. 



In the second the interval is measured direct and the only sources of error 

 are in the measurement of the small quantity by which the time is in excess of or 

 defect of a 10-minute division of the glass scale and in the correction to the length 

 of the glass scale itself. 



This method probably gives the closest approximation to the interval of time 

 betvy-een the registration of the shock at different observatories and hence the 

 velocity of transmission. 



If then 2 and .'J give results fairly in accordance, it will probably be best to 

 accept the time intervals given by 3 as being correct and then compare these with 

 the measurements of the time intervals obtained by the direct measurements of 

 the times of shocks. 



