of Water and other TAquids. 411 



With ethyl iodide we get a perfectly continuous straight 

 line throughout. 



Carbon disulphide gives a fairly straight line from —20° 

 up to 45° ; at this last point some irregularity begins to show 

 itself, and we get two points at A and B which are noncon- 

 formable to an extraordinary extent. The difference at B 

 would correspond to an error of some 3°, and though it is 

 difficult to imagine how such an error could occur (as it would 

 have to exist in all the determinations above 51°), it is still 

 more difficult to imagine that the results are true. 



Ethyl alcohol gives three straight lines with changes at 53° 

 and 19°, noticeable as being nearly coincident with two of 

 those shown by water. 



Fig. 12 gives Pierre's results with water, the portion below 

 15° being copied from fig. 3, Plate VII. The character of 

 this curve and the lowness of its position is remarkable by 

 the side of the other curves. 



The boiling-points of the various liquids are inserted in the 

 Plate (b.-p.), but they do not seem to correspond to any sudden 

 changes when the substance still remains liquid. 



Although we cannot attribute very much weight to the 

 results obtained with these volatile liquids, the general cha- 

 racter of all the curves is in favour of the view that the first 

 differential is a series of straight or nearly straight lines, and 

 that consequently the density, as a function of temperature, 

 is represented by a series of parabolic curves of the second 

 order. Methyl bromide and acetate are the only doubtful 

 exceptions amongst the non-aqueous liquids examined ; and 

 in the diagrams for water the portion between 18° and 55° is 

 the only one which appears to be uniformly curved. 



MendeleefFs work on the expansion of alcoholic solutions, 

 Tschernay's on that of salt-solutions, and my own determina- 

 tions with sulphuric-acid solutions given above (Trans. Chem. 

 Soc. 1890, p. 119), may also be cited as further proof of the 

 rectilineal character of the first differential in such cases. 



That the arguments brought forward in this communication 

 in favour of the existence of more or less sudden changes in 

 the rate of alteration of the density with temperature can be 

 regarded as absolute proof cannot for a moment be maintained; 

 nor must it be supposed that the evidence here is in any 

 respect on the same footing with that on which the conclu- 

 sions in my previous communications depend. In the present 

 case the various changes depend on differences as small as 

 any in the most doubtful cases in former investigations, and 

 the conclusions drawn from them must always be doubtful. 

 The evidence derived from the data given by the different 



