ON SEISMO LOGICAL INVESTIGATIONS. 37 



104/7 mouths) deducted from the Catalogue of Destructive Earthquakes 

 compiled under Milne's superintendence. It was natural to examine 

 the independent Catalogue 0/ Chinese Earthquakes compiled by Milne's 

 Japanese assistant, Shinobu Hirota, and published in the 1908 Report 

 of this Committee (Section XI.), with additions by Professor E. H. 

 Parker in the 1909 Eeport (Section XII.). The result was to confirm 

 the periodicity and to define it more exactly as of period 451"805 days = 

 14'8438 months = l'2370 years. The investigation is given in the 

 Monthly Notices R.A.S., Ixxix., p. 461, and it is pointed out that 

 the periodicity seems to be affected by one of long period (about 78 

 years). This led to the examination of the same Chinese series for 

 long periods (see Man. Not. R.A.S. Ixxix., p. 531), of which several 

 seem to be worth further investigation. The most notable is not the 

 one above mentioned (78 years), but one of about 240 years (which 

 may therefore be 3x78 years), which is conspicuous in the Chinese 

 earthquakes and was also found in the recoi-ds of Nile flood. It is, 

 however, only faintly traceable in Milne's Catalogue of Destructive 

 Earthquakes, and the question arises how far the heterogeneous nature 

 of the latter can be held responsible for the loss of this periodicity, 

 and how far, on the other hand, the Chinese records can be regarded 

 as possessing the necessary homogeneity. There is no doubt of the 

 defective nature of the material in the Destructive Earthquakes in the 

 early centuries. The increase in volume of the records is so consider- 

 able as quite possibly to overwhelm any signs of periodicity. For 

 example, let us limit attention to European earthquakes and to those 

 marked III. by Milne (i.e., as ' having destroyed towns and devastated 

 districts '). It might be supposed that these would be recorded with 

 some approach to completeness, yet the numbers for successive periods 

 of 180 years are as shown: — 



AD 631 — 811 — 991 — 1171 — 1351 — 1531 — 1711 — 1891 

 11 9 14 22 19 39 117 



Unless there has been an improbable increase in the number of 

 such quakes, the figures for 1711-1891 show that oaily about one in 

 ten was recorded in earlier centuries. If this happens for European 

 records, others will scarcely be in better case, and when we compound 

 the different sources it is perhaps not surprising if the accidental errors 

 are large enough to mask periodic phenomena. 



The Chinese records are also probably far from complete, but they 

 have an appearance of much greater steadiness of some kind, which 

 may quite possibly be real. A critical discussion by Chinese scholars 

 would be of great interest. 



Meanwhile we turn to some numerical relations among the periodi- 

 cities indicated, which seem to strengthen the evidence for their reality. 



Firstly, remark that the period of 451'805 days (or 1'2370 years) 

 is sensibly different from that found for the movements of the earth's 

 axis from astronomical observations. The most recent discussion of 

 this latter period by Sir F. W. Dyson (Mon. Not. R.A.S., Ixxviii., 

 p. 452) gives it as about 432 days, or accurately 7'10/6 years = 

 1'1833 years. Neither determination can be so much in error as 

 1919. H 



