182 Mr. N. Campbell on the 



ten different values of x. (The latter quantity was always 

 determined from the mass of the substance placed in the 

 tray and the known area (a) of the tray. Since the value 

 of XI p and not of \ is required, a knowledge of the density 

 is not necessary.) It was found that the most satisfactorily 

 uniform layer of such solids as were used could be obtained 

 by reducing the solid to fine powder and sprinkling it over 

 the surface of the tray out of a vessel of the nature of a 

 pepper-pot. The usual graphical method of determining \/p 

 from such observations w T as not sufficiently delicate to take 

 full advantage of the accuracy of the observations : this 

 quantity was found by solving by the least square method 

 the equations of form : 



log I - log I = \/ / o,-.m 



resulting directly from the observations. I and \/p are the 

 unknowns. The calculated probable error was never more 

 than 1 per cent. 



However, since only comparative measurements on the 

 solutions of different concentrations were required, it was 

 thought in the first instance sufficient to make observations 

 on the value of I for only one value of x, the same for all 

 the solutions. Accordingly in each case the contents of 

 a pipette (6*691 c.c.) were poured upon the filter-paper. 

 Ten measurements of this nature were made on each 

 solution : the agreement between the calculated values of 

 \/p was such that the probable error of the mean was always 

 less than 0'8 per cent. 



The composition of the solutions was estimated by pre- 

 paring them by dilution of a saturated solution, the content 

 of which was known by analysis. The values thus obtained 

 were confirmed from density measurements with the aid of 

 Landolt and Bornstein's tables. 



4. Three solutions w^ere investigated in this manner, 

 namely, aqueous solutions of potassium carbonate, potassium 

 acetate, and calcium chloride. The following diagrams 

 (1, 2, 3) indicate the results obtained. 



The straight line on the diagram joins the points repre- 

 senting the pure solvent and the pure solute. If the law 

 enounced by Crowther for solid compounds w^ere true for 

 liquids, all the points would lie on this line. The points do 

 not lie on the line, and it will be noted that in the case of 

 each of the three solutions, a curve drawn through the 

 observed points seems to possess a common feature, namely* 

 a point of inflexion nearly midway between the solvent and 



