are unquestionably th e m ost complete ever constructed for an earth quake-disturbed country.
The discussion is really a working out of certain lines suggested in a paper on “ E arthquake Frequency,” communicated by me in May, 1885, to th e Seismological Society of Japan, and published in vol. 9 of the ‘ Transactions ’ of th a t Society. In th at paper I pointed out the im portance of subjecting earthquake statistics to some strict form of m athem atical analysis, and gave a simple arithm etical process for separating the annual and semi-annual periods in earthquake frequency. The results then obtained have been fully corroborated by D r. C. D avison in his paper “ On the Annual and Sem iannual Seismic Periods ” ( ‘ Phil. Trans.,’ vol. 184, 1893) ; and my suggestion th a t the annual period is connected with barometric pressure is also strongly supported by Dr. Ferd. Seidl in his pam phlet ‘ Die Beziehungen zwischen Erdbeben und Atmospharischen B ew egungen’ (Laibach, 1895). T he sem i-annual period, which was first clearly bro u g h t into evidence in my earlier paper, does not adm it of a very ready explanation.
In my paper of 1885 I also considered in some detail the various tidal actions w hich m ight reasonably be supposed to have a determ ining influence on earthquake frequency. From lack of m aterial it was not possible at th a t tim e to make a satisfactory search for lunar periodicities; but the rem arkable fulness of inform ation contained in Professor M ilne’s latest catalogues tem pted me to undertake the labour involved in (first) tabulating the statistics in term s of lunar periods, and (second) analysing harm onically the tables so prepared.
2. The Lunar Daily and H alf-daily ..— In one of the catalogues the earthquakes are classed according to district. D istricts 1 to 6 lie on the N .E. and E. coasts of Japan, reckoning from the n o rth ; districts 6 to 11 on th e S. coast; and 12 to 15 on the W. coast. D istricts 6 and 7 are the m ost im portant, the former being the region including Tokyo and Yokohama, and the latter the region including Nagoya, w hich was the scene of th e destructive earthquake of October 28, 1891. The investigation into a possible lunar daily period is conveniently based upon this classification into districts. H ad th a t not been done by Professor Milne the labour involved in taking into account differences in local tim e would have been enormous ; for, to compare the tim e of occurrence of a recorded earthquake w ith the im mediately preceding m eridian passage of the moon, it was necessary to apply corrections for longitude and local time.
The statistics for each district were, in the first instance, separated out and tabulated according to time of occurrence, estimated in hours after the im m ediately preceding passage of the moon. The method
