BASE-LINE VALUES. 



79 



taken by a practised observer like Mr. BKRNACCHI. On the other hand, it is difficult to imagine what 

 natural cause could lead to alterations of 50' or even of 30' per month in the base-line value of a Declination 

 magnetograph, the alteration continuing for several months in the same direction. The base line, it 

 need hardly be mentioned, is due to light reflected from a mirror, which is supposed to be fixed, on to 

 photographic paper on a drum which is also fixed. In the Kew magnetograph it is doubtful whether the 

 total alteration in the base line between 1890 and 1900 attained to as much as 1'. 



11. Table III gives the values deduced for the base line of the Horizontal-Force curves from individual 

 observations, the values thence calculated for individual months, and the values actually used. The last 

 column gives the excess of the base-line value for each month over that for the previous month. As 

 already explained in the case of the Declination, the differences between the base-line values accepted for 

 successive months appear as discontinuities modified in April, 1903, by a change of scale value between 

 the values assigned respectively to the first midnight of a month and the last midnight of the previous 

 month. As with the Declination, it is difficult to account for the large apparent variations in the base-line 

 value. Owing to the limited number of absolute observations, observational errors doubtless come in ; 

 but they can hardly account for any large fraction of the larger discontinuities. The collimator magnets 

 used for the absolute observations were old, and there was but little change in their magnetic moments, 

 especially in that of 25A, the magnet chiefly employed. If the absolute observations had been faulty 

 a circumstance improbable in view of Mr. BERNACCHI'S experience this would have shown itself through 

 irregularity in the values given by the individual observations for the moment of 25A. The accordance, 

 however, in the values for the moment is satisfactory and suggests that the probable error in individual 

 absolute observations of Horizontal Force was at most from 10-y to 20y. Moreover, errors in the absolute 

 determinations would naturally vary irregularly in sign, while the apparent monthly changes in the values 

 of the base line in Table III appear on the whole systematic. 



TABLE III. Values of the Horizontal-Force Base Line. 



Applied also during March. May base-line value used after Noon on April : 

 t After August 20. September base-line value applied. 



