226 



consists in the removal of the acting magnet. The change 

 produced is doubled, and therefore the effect of errors of ob- 

 servation halved ; and the two parts of the observation may be 

 made at a very short interval, — an essential condition of accu- 

 racy in the elimination of the irregular changes. If the ob- 

 servations be repeated six or seven times in rapid succession, 

 and at a time in which the irregular changes are small, the 

 final result may be depended on to a few hundredths of a^ 

 minute. 



The following observation may serve as a favourable ex- 

 ample of the accuracy attainable by this process. It was made 

 to determine the effect of the action of the magnet of the 

 balance magnetometer upon that of the declinometer, the two 

 magnets being in the plane of the magnetic meridian, and the 

 distance of their centres nineteen feet. The third column 

 contains the differences of the corresponding readings in the 

 second column, and the means of the preceding and subsequent 

 readings. 



North end. 



Reading. 



Differences. 



(N-S) 



North 



51-37 





South 



5178 



-0-46 



North 



51-28 



-0-49 



South 



51-76 



- 0-47 



North 



51-30 



-0-47 



South 



51-78 



-0-49 



North 



51-28 







Mean 



= - 0-48 



Hence the error = - 0*24 scale divisions = - 0-17. 



This process cannot be employed when the disturbing- 

 action is that of a mass of soft iron, and we must, in that case, 

 have recourse to the less accurate and less easy method of 

 removal. The following will serve as an example of the mode 

 of dealing with such cases. 



