100 MR. K. D. OLDHAM OX A SEASONAL VARIATION [vol. lxxiv, 



December and January, as representing midwinter conditions, 

 showed a larger excess of night shocks than the general average of 

 the winter half-year. 



The actual figures and ratios referred to above are set forth in 

 detail in tabular statement No. 1 (p. 101), and in this each group 

 of half-yearly ratios of frequency may be taken as a single event. 

 In any division of the year into two halves the ratio of day to night 

 shocks would very improbably be exactly the same as for the whole : 

 the actual divergence is in no case greater than might reasonably 

 be expected if the variation were purely fortuitous, and the odds 

 are even that the variation would be in either direction. No 

 conclusion can, therefore, be drawn from any one group taken 

 separately ; but the probability that the variation would be in 

 the same direction in each of the three groups is only one-eighth. 

 Similarly, the two extremes in the Italian record may be taken as 

 independent events ; in each case, if merely accidental, the variation 

 might, with equal probability, be such as to make the ratio either 

 greater or less than the average of the whole of each half-year, but 

 the probability of their being in both cases greater is only one- 

 quarter, reducing the probability of the whole series falling in the 

 order actually found to one thirty-second. In other words, the 

 odds are 31 to 1 that the variations are not accidental but due to 

 some common cause, and, if the lunar tabulation of the Shillong 

 seismograph can be accepted as another event, the odds 

 become 63 to 1. 



From these considerations it may safely be laid down as a 

 general principle that earthquakes are rather more frequent during 

 the day in summer and less frequent in winter than the general 

 averages, with an opposite variation in frequency during the night ; 

 and that this relation may be expected wherever the record is 

 sufficiently complete, covers a sufficiently long period, and contains 

 not less than about 1000 separate shocks. This last proviso is 

 necessary, for, with so irregular a phenomenon as an earthquake, a 

 good average can only be obtained where a sufficiently large 

 number of separate records is dealt with. Experience has shown 

 that 400 is about the lowest limit which can be safely used, a 

 larger number is better; but the irregularities become so great, 

 when the number in each separate group dealt with falls below 400, 

 that small variations cease to be traceable. 



The extent of the Italian record makes it possible to carry the 

 investigation farther, by considering, separately, each of the two- 

 hour periods in which the record is tabulated. This has been done 

 in tabular statement No. 2 (p. 102), from which it will be seen that 

 the relation between day and night shocks which was found in the 

 case of the half-days is also met with in each of the two-hour periods 

 of each half. A similar tabulation of the Japanese catalogue gave 

 results in substantial agreement with the Italian, with irregularities 

 for the periods between 2 and 6 a.m. and 4 to 6 p.m. ; the Japan 

 record, however, is much less homogeneous than the Italian, for, of 

 the total number of shocks recorded, more than three-fifths are 



