New Theories of Solution. 



357 



proved by the agreement between experimental results and 

 the values calculated directly or indirectly by its aid. Mr. 

 Pickering, however, has made a number of accurate observa- 

 tions on the freezing-points of dilute solutions, and states that 

 the numbers obtained by him are in discordance with those 

 demanded by the law. The solvent he employed was water : 

 the substances dissolved were sulphuric acid, calcium nitrate, 

 calcium chloride, and alcohol. A table of the deviation of 

 these solutions from regularity, i. e. from Blagden's Law, is 

 given bv Mr. Pickering in l Nature/ vol. xlii. p. 629 (also in 

 B. A. Eeports, 1890, p. 316). It will be observed that only 

 the solution of alcohol is a non-electrolyte, and therefore it 

 only can be expected to give results in conformity with the 

 law in question. Now in Mr. Pickering's table for dilute 

 solutions the freezing-point of the aqueous alcohol shows a 

 maximum deviation from regularity of 0°'0035, a result which 

 might indeed seem to confirm the theory rather than contradict 

 it. Mr. Pickering, however, estimates his mean error at 

 o- 0005, a seventh part of the above amount. This is some- 

 what strange ; for on the same page where the claim to such 

 accuracy is first made (Chem. Soc. Journ. lvii. p. 335) he 

 compares two series of observations, made with different 

 instruments, with the following result * : — 





Freezing-point. 





P. c. H 2 S0 4 . 







Difference. 









Series I. 



Series II. 





005 



-0°-0260 



-0°-0263 



+•0003 



010 



•0515 



•0492 



-•0023 



0-20 



•0899 



•0911 



4-0012 



050 



•2135 (?) 



•2054 



4- -0009 



1-00 



•4090 



•4018 



-•0072 



1-50 



•5896 



•5846 (?) 



-•0059 



The values for the two series are deduced from the 

 " smoothed " curves drawn to represent them, i. e. errors of 

 individual experiments are already as far as possible eliminated. 

 Yet if we add up the differences without regard to sign, and 

 divide by their number, we find a mean difference of o, 003. 

 The result is quite the same if we compare the two series at 

 closer intervals, twenty corresponding points on the two curves 

 giving a mean difference of o, 0033. Thus the only evidence 

 that Mr, Pickering places before us from which we can form 



* There are evidently some misprints in this table ; I have taken the 

 differences as being correct. 



