186 Mr. N. Campbell on the 



position of the ionization-vessel may change the absolute 

 values of the absorption coefficient of solids as well as liquids 

 by as much as 10 per cent., but they did not seem to have 

 any effect upon their relative values or upon the shape of 

 the absorption-curve for the liquids. Other liquids were 

 tried — paraffin, aniline, alcohol, as well as several solutions, 

 but the curves with and without filter-paper all showed the 

 same peculiarities. In all cases the absorption coefficient 

 (deduced from the straight part of the curve) was less when 

 filter-paper was used than when it was not used. Different 

 thicknesses of filter-paper produced little, if any, effect upon 

 the observations. Since the curvature of the curves with and 

 without filter-paper are in opposite directions in their earlier 

 parts, it might be possible to find paper so thin that the curve 

 should be straight ; but the thinnest paper obtainable gave 

 results differing very little from the thickest. 



The shape of the early part of the curve without filter- 

 paper can be explained easily on the assumption that the 

 thin layers do not spread themselves evenly over the tray ; 

 for a non-uniform layer would absorb less rays than a 

 uniform layer of the same mass. But I cannot offer any 

 explanation of the other deviations from a straight line, 

 or (the most important point of all) of the difference between 

 the absorption coefficients calculated with and without the 

 filter-paper. 



It should be noted that the calculated probable error of 

 X/p for a liquid is always greater than that for a solid, and 

 greater than that estimated from the known errors of the 

 constituent observations. This fact would seem to indicate 

 that that part of the curve for solutions which appears to be 

 a straight line is not so in reality — as might be expected 

 from the form of the remainder of the curve. 



6. It appears doubtful, then, whether liquids possess a true 

 absorption coefficient as defined in the first paragraph ; at 

 any rate it is obvious that that quantity differs according to 

 the method of measurement. Nevertheless, it was thought 

 worth while to continue the comparison of the absorption of 

 solutions of different concentrations by adopting throughout 

 the same method of calculating the absorption coefficient. 

 Since rather more consistent results were obtained with the 

 use of filter-paper, that method was adopted. The absorption 

 is approximately exponential for masses of the liquid between 

 2 and 7 grammes ; some ten measurements were made on 

 each solution between these limits, and the value of \/p 

 obtained from them by a least square calculation. 



Diagrams 5-8 give the results obtained. Four solutions 



