136 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 54 



jdelding results more nearly in harmony with the chemical methods of 

 determination, the general mean for sulphur dioxide gives 802 = 64.069, 

 ±.0014, and 8 = 32.069, ±.0014. 



Another value for the atomic weight of sulphur is derivable from the 

 density of hydrogen sulphide, as determined by Baume and Perrot.^ 

 Their crude values for the weight of a litre of the gas are as follows: 



Corrected to the usual standards, the weight of the normal litre be- 

 comes 1.5392 grammes. With the critical constants determined by Olzew- 

 ski, ao = . 01438, and &o = . 00240. Hence the molecular weight of H2S 

 is 34.0893, and 8 = 32.074, ±.0030. This, combined with the value de- 

 duced from the density of sulphur dioxide, gives a general mean of 8 = 

 32.070, ±.0013. 



GENEEAL DISCUSSION. 



There are now before us, as developed in the preceding pages, 55 

 ratios, from which the atomic weights of ten elements are to be computed. 

 These elements are hydrogen, silver, chlorine, bromine, iodine, nitrogen, 

 carbon, sulphur, sodium and potassium. The first twelve " ratios " are 

 really positive values, referred to = 16, which can be regarded as first 

 approximations to the true quantities. These values are applicable to the 

 reduction of the remaining ratios, by which they are themselves to be 

 adjusted in turn. 



The rigorous method of dealing with such a mass of data is well under- 

 stood.^ The several ratios should be transformed into linear equations, 

 and each one weighted inversely as the square of its " probable error." 

 The 55 equations should then be combined into 10 normal equations, 

 which, when solved, would give the 10 atomic weights now under con- 

 sideration. But that method of reduction is exceedingly laborious, and 

 would possibly be premature. There is great activity at present in the 

 measurement of fundamental ratios, and for that reason the rigorous dis- 



1 Journ. Chim. Phys., 6, 610. 1908. Baume and Perrot reduce their data with the aid of the 

 constant 22.41<1, instead of the 22.412 adopted here. Leduc's single determination of the density 

 (Ann. Chim. Phys. (7), 15, 3i) may be neslected. His gas was not certainly pure. 



-See Clarke, Am. Chem. Journ., 27, 321. 1902. 



