of Substances in Solution. 131 



tion, and also that into which it diffused, was a very weak 

 solution of sulphuric acid of which the freezing-point was 

 0°'0I41. This prevented all but a trace of oxidation of the 

 pyrogallol. The tannin used was probably not pure; but the 

 difference in the calculated molecular weights deduced from 

 the determinations in the two series is due to the solutions 

 used being of different strengths, and to the fact that the 

 molecular depression of the freezing-point diminishes rapidly 

 as the strength of the solution increases. The depressions, I 

 may mention, appear to show some marked irregularity ; for 

 when the four determinations (the two in Table I. at the 

 initial strength, and the first two in Table II.) are plotted out, 

 they do not all lie on any simple curve, either the determi- 

 nation at 12 parts to 100 giving much too small a value, or 

 that at 16 parts too large a value. Another specimen of 

 tannin was examined as to the freezing-points, and the results, 

 which are the last four given in Table II., show the same 

 peculiarity. This sample, however, gave throughout smaller 

 values than that used in the diffusion-experiments, and the 

 three stronger solutions of it deposited some tannin as well as 

 ice in the freezing-point determination. The first sample did 

 not do so. 



In order to make the results of the first, series more easily 

 comparable with those of the second, the values for mv 2 

 obtained in the former have been multiplied by 2'577( = j;), 

 in accordance with the results obtained with glycerine and 

 cane-sugar, as mentioned above. The values for ?nv 2 are, as 

 will be seen, very far from being constant. They vary from 

 16"09 with pyrogallol (omitting the doubtful result with gallic 

 acid) to Ib'"81 with the weakest solution of cane-sugar; a 

 variation of some 200 per cent., which cannot be in any way 

 explained by the purely experimental error, since this, as has 

 been mentioned, does not exceed 2 to 3 per cent. When we 

 take the molecular weight deduced from the osmotic pressure 

 (freezing-point) itself, m', instead of the theoretical molecular 

 weight, the constancy is in most cases scarcely improved; 

 raid, indeed, in those three eases where these calculated mo- 

 lecular weights do not agree with the simple theoretical 

 weight, the values for m f v 2 are made much less nearly con- 

 stant, and attain the enormous proportions of from 72 to 260. 

 This tends to show that there is a want of agreement between 

 the values calculated for mv 2 from the depression of the 

 freezing-point and the rate of diffusion. 



K2 



