18 



DISCUSSION OF THE HORIZONTAL COMPONENT 



The first three observations were not made at the Girard College grounds ; and 

 it appears from Prof. Loomis' observation when compared with Dr. Locke's, that a 

 correction of 0.023 in the value of X should be added to these; to the twelfth observa- 

 tion I have assigned only half weight ; it was probably made during a disturbance. 

 From the general discussion an annual diminution in the horizontal force of 0.0011 

 parts was deduced for a number of stations on the Atlantic coast. At Toronto (vol. 

 Ill of General Sabine's Discussion) the annual decrease was found 0.0010 in parts 

 of the horizontal force. Being somewhat guided by these results, after several 

 trials, the following combination of the results in the table has been adopted, as 

 perhaps best representing the values for the time during which the Girard College 

 observations were made, these latter being merely of a differential character: 



Combination. 

 1, 2, 3 

 5, 1, S, 9 

 10, 11, 12, 13 



Mean epoch. 

 1837.1 

 1842.6 

 1852.3 



Mean horiz'l int. X. 

 4.191 

 4.174 

 4.145 



The annual diminution of X is 0.0030, or, when expressed in parts of the hori- 

 zontal force, = 0.0007; its equivalent in scale divisions is 19.2. The total annual 

 change was found to be 220 scale divisions ; hence, 200.8 scale divisions of annual 

 change is due to loss of magnetism of the bar. 



The mean epoch is 1844.0, and the corresponding mean X = 4.170; the mean 

 epoch of the observation taken at the Girard College, is January, 1843, for which, 

 therefore, the mean value of X = 4.173. This value has been adopted whenever 

 it was desirable to introduce the horizontal force in absolute measure. 



Separation of the Larger Disturbances. The observations having been referred 

 to a uniform temperature, and corrected for progressive change, Peirce's criterion 

 was applied separately to each month. For this purpose, a systematic application 

 was made extending over the whole series of observations, commencing with the 

 hour and the month of July, next with the hour 2 and August, followed by hour 

 4 and September, and so on in regular progression. This process eliminates from 

 the result the diurnal variation and the annual variation of the disturbances them- 

 selves. The value for O b in July, 1840, was omitted as affected by two very large 

 disturbances. The following table shows the limiting value of difference from the 



1 Added while this paper is passing through the press. 



