])1{. M. W. TUAVEKS, MIL (E SEXTEi;, AND Di;. A. -lAQrEKOl) 
of the observed values is based, as the tables show, on eight measurements 
of the pressure on the gas, the temperature of the dead-space, &c., when the bulb 
rtf the thei-mometer was immersed in ice, and twelve similar measurements where 
the bulb was surrounded with steam. The agreement between the observatirais is 
satisfoctory. 
As has already been pointed out at the commencement of this paper, the best 
determinations of the pressure coefficient for hydrogen are those of CHAPrris and 
()XNES. Chappujs’ work is to be found in the ‘ Travaux et Memoires du Bureau 
International des Folds et Mesures,’ vol. 6 (1888), and in the subsequent volumes; 
an account of his work is also to be found in the ‘Bapportsdu Congres International de 
Physique ’ (Paris, 1901). The results obtained in 1887 with his large thermometer, 
which has a platinum-iridium ladb of nearly 1 litre capacity filled at an initial 
pressure of 1000 millims. of mercury, are as follows :—■ 
Value of the coefficient. 
j\lean values. 
0-0036G271 
1 
0-00;36621S 
0-00366225 
1 
0-00366231 
0-00366254 
0-00366256 
, 
0-00366270 
1 
1 
0-00366269 
J 
The mean value of four determinations carried out in 1809 autli the same 
thermometer was 0'003(5G296, and of five determinations made in 1895 with a bulb 
of “ verre dur ” was 0'OOOGG217. The mean value of the seven determinations 
made in 1887 is, liowever, retained as the probable value of the coefficient, which 
foiins the basis tor the definition of tlie so-called normal scale of temperature. 
The tliree determinations of the coefficient by OxxES (‘ T'ornmnnicatiojis from the 
Phvsical Laboratorv of the University of Leiden,’ No. 09) are as follows :— 
0'003GG-28, 
0 •0030021, 
0-0030028. 
As in the case of Ciiappuls’ ex})eriments, the initial pressure in the thermometer was 
about 1000 millims. of mercurv. 
Though the highest initial pressure at which our measurements were made was 
700 millims. of mercuiy, there is, as we shall ])resently show, no reason for assuming 
that the coefficient varies with the pressure, and the agreement between Chappuis’ 
results and our own may be considered as confirmatorv of the latter. The combined 
results may prol)ably be considered correct to the fiftli significant figure. 
