180 Profs. Richardson and Cooke on the Heat developed 



with the method of taking galvanometer readings described 

 below. When the conditions had become satisfactory the 

 resistance R x was adjusted first of all, so that there was no 

 deflexion of the galvanometer when the Wheatstoue's bridge 

 E.M.F. was operative. Then the cell C 2 was reversed and the 

 balancing resistance R 4 adjusted until there was no deflexion 

 of the galvanometer. The battery C 2 was now reversed 

 so that Ci and C 2 were in series, and since the direct effect 

 of the thermionic current on the galvanometer has been com- 

 pensated for, any deflexion produced when the thermionic 

 current is turned on can only arise from the alteration its 

 heating effect produces in the resistance of the strip S. There 

 will, of course, be some heating effect due to the increase of 

 the current flowing through the strip on account of its Joule 

 effect. It is easy to show that with the current used in the 

 previous experiments this effect was small compared with the 

 effects measured, aud the correction for this will be considered 

 later. On turning on the thermionic current it was found 

 that the balance of the galvanometer was disturbed, the 

 deflexion being very rapid at first but gradually dying away 

 to a small drift which was difficult to distinguish from the 

 gradual drift of the galvanometer. Preliminary experiments 

 showed that much the greater part of the heating effect 

 occurred in the first half minute. It is difficult to be certain 

 exactly how much, but, with a thin strip, 90 per cent., and 

 probably more, of the total change of resistance developed took 

 place within this time. Presumably, in different experiments 

 under similar conditions the same fraction of the final rise of 

 temperature would be developed in equal times, so that the 

 method was adopted of always measuring the increase of 

 temperature, or the change of resistance, to which it is pro- 

 portional, developed in a given interval. This was usually 

 30 seconds, but in some of the earlier experiments only 

 15 seconds, while more recently intervals of one or two 

 minutes have been used. As a matter of experience, how- 

 ever, it seems that 30 seconds is the best interval to use as it 

 is long enough to obtain the bulk of the effect, and to enable 

 the periodic oscillations of the galvanometer to die down, 

 while avoiding to a very considerable extent the errors arising 

 from the drifting of the temperature to which the heatings 

 with longer intervals are more subject. We have made 

 several experiments to test the point, but have not been able 

 to convince ourselves that there is any difference in the values 

 obtained when different times are employed, except such as 

 might arise from casual fluctuations. In order to eliminate 

 the effect of these as far as possible, long series of readings 



