PHYSICAL PROPERTIES OF WATER, ETC. 35 



the same in all. It was, however, not much required, for the pressure could be adjusted 

 with considerable accuracy, and (especially when the large gun was used) the changes 

 of temperature were very slow. 



The next step was to enter, as shown in Plate II. fig. 3, all the results obtained 

 from the various piezometers at each definite temperature and pressure, with the view 

 of selecting the most probable value. The amount of discordance was in all cases very 

 much the same as that shown in the plate for the series of experiments at two tons' 

 pressure and the one temperature 5° C. It will be observed that the extreme limits of 

 divergence from the mean are not more than about two units in the third significant 

 place. For a pressure of one ton this corresponds to about half a millimetre in the 

 position of the indices, so that after what has been said about their peculiarities of 

 behaviour it may obviously be treated as unavoidable error. Thus the ordinary 

 process of taking means is applicable, unless the observations themselves show some 

 peculiarity which forbids the use of this method. 



All the results of observations made up to June 1887 (with the help of the Amagat 

 gauge) having been treated in this way, the following mean values of apparent average 

 compressibility (multiplied by 10 s ) were deduced from them : — 



[I think it extremely probable that the small irregularities among the last three 

 numbers in each pressure column may be due to want of uniformity of temperature 

 throughout the column of water in the pressure chamber. The day-temperature of 

 the cellar is, in summer, always a good deal above that at night, so that in the forenoon 

 (when the experiments were made) the gun and its contents were steadily growing 

 warmer. Thus the column of water was not at a uniform temperature. The 

 assumed temperature was the mean of the readings before the vessel containing the 

 piezometers was inserted, and after it was taken out. While it was in the chamber, 

 the contents could not be properly stirred except by raising and depressing the vessel 

 itself.] 



The points thus determined were laid down (marked with a *) as in Plate 1., 

 and smooth curves were drawn libcrd manu among them. From these curves the 



