56 REPORTS ON THE STATE OF SCIENCE. — 1915. 



original form. The edges of the eight triangles form three great circles 

 at right angles, and it was expected that the epicentres would cluster 

 about these edges. But only one of these great circles, GKECA, fulfils 

 this expectation. Of a second, LKFCD, it was already remarked in 

 the last Report that it was not at present a conspicuous line of earthquake 

 activity, but that the geographical features characterising it (the Red 

 Sea, Italy, the Alps, the American Lakes, California) suggested such 

 activity in the past. 



The principal change required by the hypothesis as stated in the 

 last Report concerns the third great circle, which is formed by EF and 

 the boundaries of the map. For this we must clearly substitute a great 

 circle LFW, with ZD, which falls midway between the two now discarded. 

 This makes the general hypothesis really simpler than before, substituting 

 two circles, still at right angles, for the three formerly suggested, and 

 retaining many of the important features of symmetry. The two circles 

 retained have been indicated by a triplet of lines. (A thick line would 

 have obscured some of the epicentre dots.) 



The third circle cutting both these at right angles would be along 

 DF, then F to the South Pole, and South Pole to L. There is something 

 to be said for including this in the system, but it does not account for 

 much that is not already accounted for by the other two, so that for the 

 present we may omit it. 



We could, of course, suit the great majoritv of epicentres better by 

 drawing, instead of FL, some line nearer to E ; and, instead of FW, 

 some line nearer to C. This means leaving the great circle and sub- 

 stituting some small circle. ' Libbey's Circle,' which was drawn by 

 Milne on the earthquake maps in the Reports for 1903-1909, is approxi- 

 mately a small circle parallel to WFL, at about 20° from it, and would 

 suit the facts very well. But the elementary simplicity of the hypo- 

 thesis is then lost, and it seems preferable to retain the simple form above 

 indicated for comparison with future facts ; and perhaps even with a 

 revised version of the facts already used, for the determinations repre- 

 sented on the diagram are of different values — in some cases well estab- 

 lished, in others very uncertain. 



V. Improvements of the Milne Seismograph. 



Section IV. of the last Report is devoted to a discussion of the times 

 recorded by different seismographs for the beginning of P and S. It 

 is shown that while the probable error of the Milne instrument is dis- 

 tinctly greater than that of modern instruments, the favourable geo- 

 graphical distribution of these pioneer seismographs renders them still 

 capable of giving valuable information. At the same time it is clear 

 that the time has arrived when it is desirable to give the Milne seismo- 

 graphs 



(a) a higher magnification ; 



(b) some form of damping. 



The former consideration is put first because seismology is at present 

 very definitely concerned with the determination of the times of arrival 

 of P and S. Not only do the tables for these times of arrival require 

 corrections, but it seems probable that the phenomena themselves are 

 not always rightly identified. It was pointed out that at distances from 



