259 
BETWEEN THE FREEZING AND BOILING-POINTS. 
materials must all have involved the same error, always in the same direction. At 
the same time my cadmium cells must also have been in error to exactly the same 
amount and in the same direction, in order to give a ratio to my Clark cells identical 
with that obtained for the cells at the Reichsanstalt, which have been compared 
directly with the Cavendish standards used by Griffiths. If the error is attributed 
to the value of my resistance, then we must reject the signed certificates of 
II standard ohms from the Electrical Standards Committee of the British Asso¬ 
ciation, as well as a true ohm from the German Reichsanstalt, as being in error. It 
is far more likely that the values of my constants agreed to 1 in 10,000 with those 
used by Griffiths and by Schuster and Gannon respectively, and that the 
difference in our results is to be attributed to some constant source of error as yet 
undiscovered in our methods of calorimetry. However, the values obtained by these 
observers using the same method differ by nearly 1 part in 1000 from each other, 
which is not so good an agreement as exists between the measurements of Reynolds 
and Moorby, Rowland, and myself, using widely different methods. At the same 
time the method used by Rowland is essentially the same as that used by 
Griffiths, and is subject to similar calorimetric errors. Owing to the great care 
and trouble taken by Griffiths to carry out his experiments, it is difficult to see 
where the difference between our two results can be. Moreover, the temperature 
coefficient obtained by Griffiths, although a linear one over the range of his 
experiments, is almost exactly a mean to the curve in my experiments over the same 
range. 
The individual observations by the present method agree very well amongst 
themselves,- but although it may be correctly said that the mere repetition of 
observations does not necessarily eliminate errors of experiment, yet it is possible to 
vary the conditions so thoroughly by the continuous flow method of calorimetry as to 
leave little room for any systematic error. In addition to varying rise of tem¬ 
perature, water flow and electric current, the present measurements have been made 
to the same order of accuracy by varying the shape and resistance of the electric 
heating conductor, by using flow T -tubes of different sizes, and by employing calori¬ 
meters with different values of heat-loss, this last being identical to the cooling 
correction in the older methods of calorimetry. 
It may be questioned whether the separate determination of the cooling effect by 
special experiment and its subsequent application as a correction to calorimetric 
experiments, can be relied on to an accuracy greater than 1 part in 1000. The 
variations in the radiation loss measured from time to time in the present experiments 
are so large that unless it had been separately determined and eliminated from the 
final result for each experiment, large errors would have been introduced. Indeed, it 
appears that the cooling correction is a far more uncertain factor in methods of 
calorimetry than has been hitherto sufficiently realized. All questions, however, 
relating to the absolute values of the standards used in the present results in no way 
2 L 2 
