270 Recent Progress in Sanitary Science. 



On referring to these tables, it will be seen that the first 

 element to be determined is the amount of oxygen in the 

 atmosphere, and within what limits this amount may vary, 

 the atmosphere still remaining in a state of purity. So im- 

 portant is this determination, that many of the greatest physi- 

 cists have expended all the resources of their skill upon its 

 solution. About a century ago, Lord Cavendish made no 

 less than five hundred analyses of the atmosphere, and by the 

 method of absorption of oxygen by nitric oxide, a method 

 which now to us appears too crude to give reliable results, ar- 

 rived at the number 20-833 as representing the percentage of 

 oxygen. This result is little more than T Vthof one per cent, 

 less than 20-95, which is the number now accepted as the 

 most accurate mean of recent determinations. And yet Lord 

 Cavendish could not satisfy himself that there was any differ- 

 ence in the percentage of oxygen in London air, as compared 

 with that of air from the surrounding country. To deter- 

 mine whether the composition of the atmosphere was indeed 

 invariable, the subject was reopened by Dumas and Boussin- 

 gault, who employed in their classic research the chemical 

 attraction of copper for ox} T gen at an elevated temperature. 

 The air from the Jardin des Plantes, after purification from 

 every trace of moisture and carbonic anhydride, was passed 

 through a weighed tube containing turnings of pure copper, 

 and the residual nitrogen collected in a glass balloon, previ- 

 ously exhausted of air. Every precaution that ingenuity 

 could suggest, was used to insure the accuracy of the experi- 

 ments, which were repeated a great number of times and on 

 large quantities of air ; and yet these two illustrious chemists 

 did not venture to assume that the composition of the atmos- 

 phere was otherwise than invariable, and that the slight dif- 

 ferences in the percentages of oxygen obtained were due to 

 real differences, and not to variations within allowable limits 

 of instrumental error. 



By improved eudiometrical methods, Regnault afterwards 

 settled conclusively the fact of variations in the percentage of 



