ON SEISMOLOGICAL INVESTIGATIONS. 249 



Cj fits the observations best, especially for D«= 1 and D= 2. But it is a question 

 whether the fit ought to be good at these depths, for there may be many cases of small 

 focal depth which have escaped detection owing to insufficiency of observations ; 

 and when D=0 the number of cases is certainly very large, though it is not easy to 

 give a precise value to it. 



The single case D=9 occurs on 1926, February 9d. Oh. with epicentre 27°-0 S. 

 59°*5 W. It is the deepest focus yet determined ; but the evidence for this exceptional 

 depth seems good. The epicentre is practically fixed by observations at La Plata, 

 Sucre and La Paz, aU within 15° of it ; but nine observatories in Europe (azimuth 43°) 

 then require a correction to A of more than 10° ; and two others in N. America 

 (azimuth 346°) a correction of 10°. Four Asian stations at distances from 130° to 170° 

 receive [P] about 80 sec. early, and fourteen European stations with A from 94° to 117° 

 receive [S] about 2 min. too early. AU the evidence hangs together. Moreover, though 

 this is actually the greatest depth hitherto required, there are two cases requiring 

 0-080, not much less, on 1921 Dec. 18d. 15h. 29m. 24s. at 2°-5 S. 71°0 W., and on 

 1922 Sept. 4d. 17h. 4m. 8s. at 9°-0 S. 66°-0 W. It will be seen that all three extreme 

 cases occur in S. America. The six cases of 0-070 are as foUows : — - 



Lat. Long. 



Of these six only one is in S. America, so that the other three cases seem to have 

 a quite exceptional character. 



Periodicities. 



Attention has been chiefly directed to the period of approximately 21 minutes. 

 Two papers have been published in the Oeophysical Supplement to the R.A.S. Monthly 

 Notices ; one (April 1928) presenting the discussion of eleven separate series of repeti- 

 tions from the same epicentre, apparently shows that the precise value of the period 

 varies with the latitude, being about 21-2 min . near latitude 45° and less than 

 21-0 min. in seven other cases ; the second paper (October 1928) deals with the earth- 

 quakes recorded in the PhiUppines in the nine years 1918-1926, which consistently 

 show a periodicity near 20-993026 min. (given in the paper by an unfortunate oversight 

 as 20.993342 min.). The character of the variation is a sharp rise in frequency 

 (possibly per saltum) with a subsequent steady fall. 



The total counts for twelve equal subdivisions of the period adopted are as follows, 

 starting with the maximum : — 



=150 128 141 123 130 127 132 114 116 106 112 104 



Assuming a uniform slope of 3-4 per term we get 



Ci=142 138 135 

 0-Ci=+8 -10 +6 

 0-C.3= + 3 -5 +1 



There is a curious feature in the residuals 0— Ci, viz., the alternation of -f- and — 

 signs, which makes the mean odd residual +5 and the mean even residual —5. If 

 this difference is allowed for as in the line 0— Co, the sum of the squares of the residuals 

 is reduced from 461 to 159. But the main feature of the O totals is clearly the drop 

 of 3-4 per term. These are totals for nine years, so that for a single year the drop or 

 slope would be 3-4/9= -38. In the paper these slopes are calculated for the separate 

 years 1918-1926 and found to be 



+ -21, +-20, +-60, +-44, +-44, +-30, +-21, +-42, +-55, 



sho-wlng a mean value +-37 -with a mean departure of +-13. The evidence of the 

 nine years is thus satisfactorily consistent. 



