OF TIIK MAGNETIC DECLINATION. 



the remaining 27 months in the years 1S43, '44, and '45, when discussed in the 



same manner, give a rather different result. 



Some improvements, however, can be made in the preceding investigation \<\ 

 omitting the December mean of 1844, which is obviously about 12 scale divisions 

 too small; the observed value is 535 d- .2, and the interpolated value 547''.0. An 

 examination of the first series shows a defect in the monthly means of 1841, be- 

 tween May and June, requiring a constant correction of + 8.0 scale divisions for 

 the remaining months after May, as may be seen by the following table : — 



Computed value for June, 57.9-2 



Mean 



—0.7 



The following values then remain for the discussion, and they should be con- 

 sidered as forming the basis from which the legitimate results are to be deduced. 

 The numbers marked with an asterisk have been increased by 8 d \0. Interpolated 

 values are between brackets, and were obtained by comparing the means of the 

 remaining months of the year with the corresponding means of every other year ; 

 by this process several values are obtained for each interpolated number ; the re- 

 sulting mean is given in the table. The high value of 1841, and the low value of 

 1844, for the month of May, in some measure compensate. 



This last result accords in general with that before deduced, but is much to be 

 preferred. 



From June to October the north end of the magnet is accordingly to the eastward 



of the mean annual position (after the elimination of the secular change), and in 



the remaining months of the year it is to the westward of this position. From 



the vernal equinox till after the summer solstice the motion is to the eastward or 



4 



