In ascertaining the relation between sunspots and earthquakes it would be well if we could employ the strict method of correlation coefficients. This, however, is impossible for the entire century, for the record is by no means homogeneous. The earlier decades are represented by only about one-fourth as many earthquakes as the later ones, a condition which is presumably due to lack of information. This makes no difference with the method employed in Table 7, since years with many and few sunspots are distributed almost equally throughout the entire nineteenth century, but it renders the method of correlation coefficients inapplicable. During the period from 1850 onward the record is much more nearly homogeneous, though not completely so. Even in these later decades, however, allowance must be made for the fact that there are more earthquakes in winter than in summer, the average number per month for the fifty years being as follows: Jan. 2.8 May 2.4 Sept. 2.5 Feb. 2.4 June 2.3 Oct. 2.6 Mar. 2.5 July 2.4 Nov. 2.7 Apr. 2.4 Aug. 2.4 Dec. 2.8 The correlation coefficient between the departures from these monthly averages and the corresponding departures from the monthly averages of the sunspots for the same period, 1850-1899, are as follows: Sunspots and earthquakes of same month: +0.042, or 1.5 times the probable error. Sunspots of a given month and earthquakes of that month and the next: +0.084, or 3.1 times the probable error. Sunspots of three consecutive months and earthquakes of three consecutive months allowing a lag of one month, i.e., sunspots of January, February, and March compared with earthquakes of February, March, and April; sunspots of February, March, and April with earthquakes of March, April, and May, etc.; +0.112, or 4.1 times the probable error.