PHYSICAL PROPERTIES OF WATER, ETC. 31 



tive trial determined to employ exclusively the lesser of the two small pistons. Some 

 time was spent upon a comparison of the indications of this instrument with those of 

 the external gauge, with the result that single indications of the latter could not be 

 trusted within about 1 per cent., though the mean of a number of observations was 

 occasionally very close to the truth. I therefore put aside all the compression observa- 

 tions already made, and commenced afresh with the same piezometers as before, and 

 with the Amagat gauge exclusively. 



In the summer of 1886 I obtained a long series of determinations at about 11°'8 C, 

 and others at 14°"2 and 15° C. In December of the same year I worked for a long 

 time between 3° and 3° - 5 C. All of these were with the large Fraser gun. 



In June 1887, with the new compression apparatus, I secured numerous deter- 

 minations at 0°"4 C. 



In July the piezometers were filled with solutions of salt of various strengths, and 

 examined at temperatures near 19° C. and 1° C. In November these were again 

 examined, this time in the large gun at about 9° C. ; and the piezometers were again 

 rilled, some with fresh water and some with sea-water. 



During the winter complete series of observations in the large gun were obtained 

 at about 7°, 5°, 3°% 2°-3, 1°-1 ; and, finally (on March 16, 1888), at 0°-5 C. 



The piezometers were, once more, filled with the salt solutions, as I considered that 

 I had obtained sufficient data for fresh water and for sea-water; except in the one 

 important particular of the exact values of the ratio of their compressibilities at one or 

 two definite temperatures and pressures. 



These were finally obtained in May and June 1888, with piezometers considerably 

 larger and more delicate than the former set. 



VII. Final Kesults and Empirical Formulae for Fresh Water. 



Although my readings and calculations were throughout carried to four significant 

 figures, I soon found that (for reasons already sufficiently given in Section I.) only three 

 of these could be trusted even in the average of a number of successive experiments, 

 and that the third might occasionally (especially with sea-water) err by an entire unit 

 or two ; at most | per cent, of the whole quantity measured. Of course, now and then 

 there occurred results so inconsistent with the rest as to indicate, without any doubt, a 

 displacement of the index by upward or (more frequently) downward currents. 



This was made obvious by comparison of the indications of any one piezometer m 

 successive experiments at the same temperature and pressure ; but it was even more 

 easily seen in the relative behaviour of a number of piezometers which were simul- 

 taneously exposed to exactly the same temperature and pressure several times in 



