OF ARTS AND SCIENCES. 157 



smaller weight from the weight of the water, then a small error in the 

 observed weight of the hydrogen would have no appreciable effect on 

 the weight of the oxygen. 



Dumas fully recognized the source of error to which we have re- 

 ferred, and in his paper on the subject wrote what may be translated 

 as follows : — 



" Of all analyses presented to a chemist, that of water is the one 

 which offers the greatest uncertainty. Indeed, one part of hydrogen 

 unites with eight parts of oxygen to form water, and nothing would 

 be more exact than the analysis of water, if we could weigh the 

 hydrogen as well as the water which results from its combustion. 

 But the experiment is not possible under this form. We are obliged 

 to weigh the water formed, and the oxygen which was consumed in 

 producing it, and to determine by difference the weight of the hydro- 

 gen which has entered into combination. Thus an error of ■$%■$ in the 

 weight of the water, or of F £^ in the weight of the oxygen, is equiva- 

 lent to an arror of fa or fa in the weight of the hydrogen. Let 

 these two errors be in the same direction, and the total error will 

 amount to 5^." 



In the second place, however carefully the exterior surface of the 

 combustion tube may be guarded, it is impossible that the contents of 

 the tube should bear the same relations to the surrounding atmosphere 

 before and after the combustion. We begin with a tube containing 

 cupric oxide in different states, and we end with it containing reduced 

 copper, whose condition will vary more or less with the character 

 of the oxide employed; and the power of these materials to occlude 

 air or hydrogen is an unknown quantity in our experiment. That 

 it is an appreciable quantity is evident from several incidental 

 observations. 



Dumas endeavored to avoid any source of error arising from this 

 cause by exhausting the combustion tube before weighing it, but he 

 himself expresses a doubt whether a trace of hydrogen might not have 

 been left. Erdmann and Marchand in part of their experiments re- 

 sorted to the same expedient, but their results obviously vary with the 

 condition of the cupric oxide employed; and the following remarks of 

 Schutzenberger, in a discussion of the variability of the law of definite 

 proportions before the Chemical Society of Paris in 1 883, as quoted 

 in the " American Journal of Science," 3d series, Vol. XXVI. page 65, 

 have an important bearing on the same point: — 



" When water is synthesized by reduction of a known weight of 

 CuO, by weighing the reduced Cu and the water formed, it is found 



