155 



o / 



Mean of January and the following December .... 75 18*90 N. 

 Mean of February and the following November . . 75 18*98 



Mean of March and the following October 75 18*63 



Mean of April and the following September 75 18*71 ,> 



Mean of May and the following August 75 1 7*70 ,, 



Mean of June and the following July 75 17*25 



Hence on the 1st of July the mean dip at Toronto would be derived 



as follows, viz. : 



o / 



From the four winter months, November to February . . 75 18*97 



From the four summer months, May to August 75 17*47 



Showing an excess of l'*5 in the winter months above the summer 

 months. 



The annual variation at Kew, as it may be derived from the 282 

 determinations in Tables L, II. and III., does not differ materially 

 from this conclusion. There are in these Tables 87 results obtained 

 in the four winter months of the different years, and 93 results 

 obtained in the four summer months of those years. If we collect 

 into separate means the results in the winter months of 1857-58, 

 1858-59, and 1859-60, numbering them (1), (3), and (5), and into 

 separate means the results in the summer months of 1858, 1859 and 

 1860, numbering them (2), (4), and (6), and if we compare (1) 

 and (3) with (2), (3) and (5) with (4), (2) and (4) with (3), and 

 (4) and (6) with (5), (by which comparisons the effects of secular 

 change are eliminated), we find an excess of 1'7 in the mean dip of 

 the winter months over that obtained from the summer months. The 

 mean of the two corrections, thus separately obtained at Toronto and 

 Kew, is l'*6; of which the half, or 0'*8, has been applied in the 

 Tables with the sign to the results in the winter months, and with 

 the -f sign to the results in the summer months. 



Probable error of a single determination of the Dip. It may be 

 desirable to state the probable error of a single determination as it may 

 be derived from the observations in the Tables, before and after the 

 application of the correction for annual variation. It will be seen 

 that the probable error is diminished by the application that has 

 been made of a correction on this account, 



