16 PROCEEDINGS OF THE AMERICAN ACADEMY. 



very slightly with the temperature. Hence we may calculate the value 

 of P from one case, and see if the same value applies in others. If for 

 convenience we take as our unit for 'p the pressure exerted by a gram 

 equivalent of the anion when dissolved in a litre of water, it is obvious that 



for normal solutions » = a, for decinormal solutions » = — , and for 



// 



centinormal solutions » = . Thus with a centinormal solution of 



^ 100 



potassic chloride, where ninety-four* per cent of the salt is dissociated, 



we have 



log P = — + log 



dT 0.000198 ° 100 



.00094 ^ j^g 0^ ^ 4 747 _ 3 _^ ^g^g ^ 2.720. 



.000198 ° 100 



Hence P= 525. lu a similar way we may find for sodium P— 585 ; 

 evidently 550 represents a fair average value which we may adopt for 

 this quantity upon the scale of measurement used for p. The centi- 

 normal solutions were chosen in this case because here the disturbing 

 reaction HgoCL = HgCL + Hg is the least active. In order to get an 

 idea of the probable error of our many hypotheses, it is well to calculate 

 backwards and obtain a theoretical value for the temperature coefficient for 

 other centinormal solutions besides potassic chloride from the equation 



^ ^0.000198 log ^i^^^li^ 

 dT ^ a" 



in which the above named values of P and p have been substituted. 

 Performing these substitutions, and, taking for the values of the electrical 

 conductivities the data obtained by Kohlrausch at 18". we obtain the 



following table. 



* This value for a" is of course onls' approximate ; but the approximation is 

 sufficiently accurate, for a large difference in a makes only a very small ditTerence 

 in the temperature coefficient. 



