NIPHER — MAGNETIC DETERMINATIONS IN MISSOURI, 79' 1^7 



in selecting brass for a new magnetometer, it was discovered that 

 the central (north) leveling screw of Declinometer No. 3 was 

 strongly magnetic, and that the side screws in the magnet box 

 were also magnetic, although in a less degree. It required 60 

 milligrammes of iron when equidistant to counteract the attrac- 

 tion of the lower end of the leveling screw, which evidently con- 

 tains a small mass of iron. The correction which this necessi- 

 tates in the intensity work of 1878 -will be discussed further on. 



Great difficulty was found in obtaining brass sufficiently free 

 from iron, and it was finally found necessary to obtain selected 

 specimens of copper and zinc and make a special cast. Even 

 this brass was not pure, and although the screws introduced in 

 Declinometer No. 3 (viz., the front leveling screw and two side 

 screws) were more nearly pure than the old ones, the corrections 

 have not been reduced, the effect having been to change the 

 direction of the resultant couple. I am also led to suspect that 

 brass may be contaminated by contact with steel tools, although 

 experiments on this point have not given concordant results, and 

 further attention must be given to it. 



The full extent of this defect was not discovered until the close 

 of the work of 1S79, as the pressure of other work prevented me 

 from giving it the attention which it should have received. 



In order to determine the uncertainty thus introduced into our 

 determination of declination, the deflections produced by the four 

 screws used during 1878 were carefully measured for all of the 

 24 possible arrangements. The errors are 



Greatest possible error, dz 2'. 



Probable error,* Jz o.'a- 



The two new screws used in 1879 were always kept on the 

 same side of the box, although they may have been transferred 

 from one side to the other during the summer. This reduces the 

 number of possible arrangements to 8, and the errors were found 

 to be for 1879 — 



Greatest possible error, zb 2.' 



Probable error, ito.'^. 



The uncertainty therefore lies inside the ordinary limits of accu- 

 racy in declination work, and the relative errors are even less, as 

 it was always the custom — to which there were, perhaps, two or 



* That is, half of the errors are numerically less than O. ' 2) the rest being between 

 zh O.'. and ± 2.'o. 



