Permanent Magnetism. 311 



sion of the wires changes rapidly, and affects the results. The 

 condition of the instrument can always be ascertained by 

 making observations in pairs, so as to give zeros. If these 

 agree, confidence may be felt. With bad wires I have had 

 two consecutive zeros differ by several degrees. 



As to the use of the bifilar for absolute measure, I will only 

 observe, with reference to a recent paper of Kohlrausch*, that 

 I share his predilection for the method, though I do not 

 attempt to attain his degree of accuracy. But one part of his 

 work seems inadmissible for accurate purposes; and that is the 

 determination of the torsion by calculation, or by experiment 

 with a separate piece of the suspension wire. In my experi- 

 ence wires differ. I determine the torsion of the wires in use 

 by balancing it against some approximately known magnetic 

 moment, the torsion just exceeding 180°. It must be deter- 

 mined both ways (right and left). It is impossible to assume 

 that the zero of torsion coincides with the zero position. If 

 it did so once, it would not continue to do so after a weight 

 had remained deflected in the bifilar for a day or two, unless 

 with an exceptionally good wire. There is continual change 

 in consequence of what the Germans call " elastiseke Nacli- 

 ivirkung" which I may describe as "imperfect elasticity." I 

 suspect that a change of this kind had something to do with 

 the change noticed in the note at p. 215 of my previous paper. 



I may just mention that the magnetic meridian is most 

 conveniently determined by hanging an electromagnet in the 

 bifilar, by the method subsequently described. When on 

 reversing the current no change of position takes place, the 

 axis is truly in the magnetic meridian. The moment of the 

 electromagnet must be a little less than that of the bifilar. 



The difficulty in making any deduction from the observed 

 ratio of the moments of the compound permanent magnet, 

 when joined up and separated, consisted in our ignorance of 

 the amount of the air-resistances of the magnetic circuits. I 

 have succeeded in completely determining these by means of 

 soft-iron electromagnets, having nearly the same dimensions 

 as the steel, both joined up and separated. The pieces of soft 

 iron only differ slightly in diameter from the steel. The dif- 

 ference was an oversight. The diameter of the soft iron is 

 1*90 centim., that of the steel 1*97. Each length of 1*58 cen- 

 tim., whether separate or forming part of a long bar, is wound 

 with ten turns of wire. 



A mercury contact is arranged under the cradle, by which 

 currents can be transmitted to the coils with little friction. It 

 was found that semicircular channels, with dipping wires, 

 * Wiedemann's Annalen, xvii. p. 737. 



