44 ART. 4. K. HONDA AND T. TERADA. 



twisted and then magnetized. In our experimentg, instead of the 

 magnetic wire, a copper one was used for the compensation of the 

 twist at no field, in order to avoid any ambiguity caused by the 

 magnetization of the compensating wire. The specimens were cut 

 from the jorevious samples into about one-third of their k^ugths. 

 The copper wire was so chosen that its total twist for a given 

 couple was nearly equal to that of the magnetic specimen to be 

 tested. The two wires were connected in the same line by a cop- 

 per rod of moderate diameter with a small mirror attached to it, 

 and hung vertically from a torsion circle. The lower end of the 

 specimen was also rigidly connected to another copper rod. To 

 this rod, a rigid pin was perpendicularly fixed, the ends of which 

 slid in the vertical grooves cut on the inside of a flat hollow 

 cylinder at the center of the lower torsion circle. In this way, 

 the tension applied to the pan hanging on the lower end of the 

 lower co2:>per rod was transmitted independently of the twist. The 

 upper and lower ends of this connected system could therefore be 

 twisted to any desired amount by means of the torsion circles, 

 while the tension was constantly acting on the wire. A magnetiz- 

 ing coil, whose length was 30 cm. and 47r?i= 379.7, was placed 

 co-axially with the lower wire, which was, in our case, the speci- 

 men to be tested. The coil had a small resistance of 0.6i?, so 

 that the effect of heating was negligibly small, though the w^ater- 

 jacketed arrangement was dispensed with. A long scale distance 

 of 6.797 m. was used to increase the sensitiveness of the arrange- 

 ment. The sensitiveness was such that a deflection of 1 scale 

 division corresponded to a change of total twist of lö.Ü". 



To stop the vibration of the mirror, a short brass wire was 

 fixed horizontally to the vertical copper rod a little below the 



