14 ANALYSIS OF 



this, whenever the bubble of air is let oiit, and (he 

 velTel is inverted, another bubble is inrtantly formed, 

 fometimes indeed fo fmall as not to be vifible to the 

 naked eye, but always by means of a magnifying glafs, 

 and this very fmall quantity v/ill not be abforbed by the 

 water till the veflel has been laid in an inclined pofition 

 foiT!c hours. If the veflel be placed perpendicularly, the 

 bubble will come to be of a confiderable lize. Still how- 

 ever it will not increafe beyond a certain quantity, 

 though it remain in that pofition ever fo long. 



I have tried every method that I could think of to de- 

 prive water of its power of producing air, but without 

 efFedt. Heat I found of no ufe but to affift in expelling 

 the air originally contained in it, and freezing had no 

 more effedl than heat. 



When I publiflied the pamphlet above mentioned I 

 had not procured from water any other kind of air than 

 fuch as was, in a greater or lefs degree, purer than that 

 of the atmofphere, and therefore 1 imagined that this 

 might have been the origin of all the air in the atmo- 

 fphere. But I have fince found that though tlie firft 

 quantity of air that is expelled from water is much purer 

 than that of the atmofphere, the next is lefs pure, and at 

 laft it is wholly phlogifticated. This I could not difcover 

 while I made ufe of fmall bulbs ; but when I ufed large 

 ones, containing from fifty to an hundred ounces of 

 water, it was afcertained with the greateft certainty. 

 From this fad it may be inferred, either that the air pro- 

 duced from water is not that which had been imbibed 

 from the atmofphere, or that, though it imbibes mofl 

 readily that which is the purefl, it retains with the great- 

 eft obilinacy that which is leafl: pure, which is analogous 

 to other chemical affinities. If the air thus produced be 

 really generated from the water, or rather vapour, it mufl 

 be wholly phlogifticated, and afterwards purified by the 



procefs 



