PHYSICAL PROPERTIES OF WATER, ETC. 39 



The former cause could have produced no measurable effect; but if the indices win 

 all somewhat stiffer to move in the second series than in the first, the discrepance 

 might be fully accounted for. 



Fifth. Between the two series all the piezometers had, for several months, been 

 filled with strong salt-solutions. Imperfect washing out of these solutions may have 

 had the effect of rendering the second series a set of experiments on water very 

 slightly salt. 



Sixth. To make my observations applicable to natural phenomena, I purposely 

 did not employ distilled water. The ordinary water supply of Edinburgh is of very 

 fair quality, and I took care that it should not be boiled longer than was absolutely 

 necessary to prevent air-bubbles from forming in the piezometers. But it comes from 

 different sources, and is supplied as a mixture containing these in proportions which 

 vary from time to time. From this cause also the substance operated upon may have 

 been slightly different in the two series of experiments. 



As will be seen in next section, I have obtained direct proof that the first series of 

 observations is to be preferred to the second, — though I have not been able to ascertain 

 definitely which of the above causes may have been most efficient in producing the 

 discrepancy. 



It will be observed that this discussion has nothing to do with the important 

 question, Does the compressibility of water diminish from the very first as the pressure 

 increases, as was asserted by Perkins ? The first and rudest of my experiments sufficed 

 to answer this definitely in the affirmative; though the contrary opinion has been 

 confidently advanced, and is very generally held to this day. 



The discussion deals with a much more refined and difficult question, viz. Is the 

 diminution of average compressibility simply proportional to the pressure for the 

 first few hundred atmospheres, or does the compressibility fall off more slowly than 

 that proportion would indicate, as the pressure is raised ? 



VIII. Reductions, Results, and Formulae fob Sea-Water. 



As already stated, three of the six piezometers employed were filled with fresh 

 water and three with sea-water, so that simultaneous observations were made on the 

 two substances. The accordance among the various observations made with sea-water, 

 at any one temperature and pressure, was not so good as it was with fresh water ; 

 especially when the smaller compression apparatus was used. There is some curious 

 action of salt upon the hairs attached to the indices, which has the effect of rendering 

 them too loose, however stiffly they may originally have fitted the tube. Treating the 

 observations of the first series exactly as described in the preceding section, I obtained 



