during the Absorption of Electrons \>y Platinum. 183 



The change of resistance is proportional to the watts 

 supplied within the order of accuracy of the measurements. 

 The error in these measurements is considerably greater than 

 that of the method adopted in the investigation, on account 

 of the difficulty of determining the correction for the sensi- 

 tiveness of the Wheatstone's bridge system, which is, of 

 course, different for each current used. 



§ 4. Reduction of the Galvanometer Deflexions. 



It has been pointed out that it was not possible to get the 

 temperature of the grid absolutely steady. As a rule there 

 would be a regular drift corresponding to a gradual rise of 

 temperature ; in other, less common, instances the drift 

 would be in the opposite direction. Sometimes, of course, 

 there would be an irregular variation of the zero, but unless 

 this was small the readings then obtained were discarded. 



In order to eliminate the effect of the drift arising from 

 changes in the steady temperature of the grid, the method 

 was adopted of taking a large series of readings in succession, 

 for equal periods of time, with the thermionic current alter- 

 nately on and off. Thus the thermionic current would be 

 allowed to flow into the grid for, say, 30 seconds. This gave 

 rise to a deflexion of the galvanometer indicating a rise of 

 temperature. The thermionic current would then be turned 

 off' for an equal period, and the galvanometer spot would be 

 deflected back in the opposite direction. It would not, as a 

 rule, reach the initial zero in this period of time, owing to 

 the occurrence of the gradual drift. Again, on turning on 

 the thermionic current for the same length of time, a de- 

 flexion would be obtained in the original direction ; but at 

 the end of this period the reading would not be the same as 

 in the first case. The effect of the drift will be eliminated 

 if we subtract the mean of the first and third readings from 

 the second ; and if we subtract the mean of the second and 

 fourth readings from the third, we shall get independent 

 values of the effect, from which the drift has been eliminated. 

 Treating the successive readings in this way, it will be seen 

 that we can obtain any desired number of individual deter- 

 minations of the effect. Moreover, if we take a sufficiently 

 large number, not only will the steady drift be eliminated, 

 but the effects of any chance fluctuation in the temperature 

 of the grid will also be obliterated, As a matter of fact, one 

 could see at a glance whether the chance fluctuations were 

 considerable or not, and only those readings in which they 

 were insignificant were retained. 



It will be seen that this method of reducing the readings 



