6 Mr. E. H. Griffiths and Miss Marshall on the 



not essential, since, when taking the mean of a number of 

 observations, the effect of any error in Sq would be greatly 

 diminished. We believe that the above order of accuracy 

 w^as exceeded in the measurement of the respective terms. 

 In the case of the experiments with 3 cells at 20° an order of 

 accuracy of 1 in 50 would be required in the value of Q s £ s to 

 secure the same standard ; hence we do not regard our deter- 

 minations at that temperature as possessing an equal authority 

 with those at higher temperatures. 



4. The Method of maintaining the Space surrounding the 

 Calorimeter at a Constant Temperature 6 . 



A full description of the somewhat elaborate apparatus 

 designed for this purpose will be found on pp. 374-378 of 

 Paper J, and some improvements which were subsequently 

 added are described on pp. 274-276 of Paper W. This 

 portion of the apparatus did not, for some unknown reason, 

 appear to be working v. ith the same perfection this year as it 

 did in the summer of 1894. Oscillations in (the external 

 temperature) , amounting to yJo° C, were on several occasions 

 observed, the greatest change during the whole series being 

 •0145°. This, however, w r as exceptional and, as an inspection 

 of our final tables will show, the change in the course of an 

 experiment (*. e. in about half an hour) was usually only 

 about a few thousandths of a degree at temperatures varying 

 from 20° to 50° (J.; the constancy of O was therefore sufficient 

 for our purpose. 



Had it been possible, we should have preferred to determine 

 6 by means of a platinum- thermometer, as we could thus 

 have detected smaller changes ; but a third observer would, 

 in that case, have been required, and circumstances did not 

 permit of this addition*. 



The mercury -thermometers used for indicating the tem- 

 perature of the steel chamber in which the calorimeter was 

 suspended were graduated in millimetres, and had been care- 

 fully calibrated and compared with tw r o Tonnelot thermo- 

 meters standardized at the Bureau International f. The 

 steins of these thermometers (where they projected above the 

 tank-lid) were surrounded by glass tubes up which a stream 



* We believe that such irregularities as present themselves in our 

 final results are partly due to the employment of mercury-thermometers 

 for the determination of the charges in 6 , 



t See " The Measurement of Temperature/' Science Progress, Sept. 

 1894. 



