814 Proceedings of the Royal Society 
It seems, however, that all the observations give considerably 
too small a change of temperature ; for the part due to the first 
power of the pressure is from 30 to 40 per cent, less than that 
assigned by Thomson’s formula and his numerical data. One 
obvious cause of this is the small quantity of water in the com- 
pression apparatus, compared with the large mass of metal in 
contact with it. This would tend to diminish all the results, 
whether heating or cooling ; and the more so the more deliberately 
the experiments were performed. Another cause is the heating (by 
compression) of the external mercury in the pressure gauge. Thus 
the pressures are always overestimated ; the more so the more 
rapidly the experiments are conducted. A third cause, which may 
also have some effect, is the time required by the thermo-electric 
junction to assume the exact temperature of the surrounding 
liquid. 
Be this, however, as it may, the following table shows the nature 
of the agreement between the results of my original experiments 
(ante, p. 218) and the data derived from the present investigations. 
The gauge and the compression apparatus were the same as in my 
experiments of last year ; the galvanometer, the thermo-electric junc- 
tions, and the observers were all different. The column MSO gives 
the whole heating or cooling effect at 15° *5 C., calculated for different 
pressures from the results of the investigation by Professor Marshall 
and his coadjutors. The column T contains the results of my direct 
experiments at that temperature. 
p (tons) 
MSO 
T 
Thomson. 
1 
0131 C. 
0139 C. 
0-177 C. 
2 
0-294 
0-311 
0-355 
3 
0-465 
0-516 
0-533 
4 
0-665 
0-750 
0-711 
It will be noticed that there is, again, a fair agreement ; though 
the results are, as a rule, lower than those calculated from Thomson’s 
formula. My own agree most nearly with Thomson’s formula, 
probably because they were very rapidly conducted. As they stand, 
they give about 3° C. for the effect of 1 ton on the maximum 
density point. It is to be observed that if we could get the 
requisite corrections for conduction, and for compression of 
