38 ART. 4. K. nOXDA AXD T. TERADA. 



SO as to make them as equal as possible, in order that in the 

 comparison of periods for different fields, the effects of amplitudes 

 misht be neglected. One of the observers sio;nalled at each com- 

 plete oscillations, while the other recorded tlie time by a mean 

 time chronometer. According to the usual method of time-obser- 

 vation, 60 consecutive oscillations were availed of for the deter- 

 mination of its period. The first and the last 10 were signalled 

 and recorded ; we thus obtained 10 sets of time records for 50 

 complete oscillations. The mean of these observations gave the 

 period of oscillation, which usually ranged from 10 to 20 seconds. 

 Its value may be considered accurate to —Tqaq^ of a second. These 

 processes were repeated for a series of successively increasing fields, 

 the demagnetization being of course effected before each experi- 

 ment. After a series of observations was taken, the experiment 

 at no field was again repeated, and we usually found the result 

 fairly unchanged. 



Since the magnetizing coil was waterjacketed, the heating- 

 effect was inappreciable. The magnetizing current was measured 

 both before and after each experiment, and the mean was taken. 

 The current remained fairly constant during each experiment, 

 except in a few cases. 



Though the period of oscillation was long and tlie oscillating 

 weights were moderately distant from the lower end of the coil, 

 it was necessary to determine experimentally the retarding effect 

 of the Foucault current, which was liable to be produced in the 

 system viljrating in the magnetic field. For this purpose, a copper 

 wire of the same size îis tlie specimen was oscillated in a number 

 of fields and tlie corresponding periods were determined. We 

 found that for the periods used in our experiments, the effect was 

 negligeably small. Logarithmic decrements for different fields 



