It remains to account for the oxygen. There were 250 grains of nitric acid at first in the nitre; of which 200 grains were oxygen and 50 azote, nearly. One-fifth part of the oxygen = 40 grains, corresponds to 1 atom of oxygen. Now the whole of the oxygen derived from the nitre in the course of the experiment, seems to be 30 grains in gas, 7 grains in the carbonic acid, and 19 grains in the iron oxide, together equal to 56 grains. Now the azote and oxygen in the gas collected, were very nearly in the proportion of those elements in nitric acid; so that a portion of the acid (about ⅙) might be considered as completely decomposed, whilst the rest was only losing a small part of its oxygen: this is remarkable, and I think indicates that the carbonic acid (formed from the carbon of the retort, or from the adhering carbon) unites to the potash, expelling the _nitrous_ acid, which is immediately decomposed into its elements azote and oxygen. This would not however account for the whole of the azote: for, 40 grains of nitric acid would be united to 37 potash; whereas we find only 21 potash with carbonic acid; and I cannot believe that an error in the estimate of carbonate of potash could exist to that amount. The fact, however, was, that the elements of 40 grains of nitric acid were found in the evolved gas, and hence we have to account for the remainder 210 grains. From this there appears to have been expelled 26 grains of oxygen, nearly 19 and 7 as related above; of which the 19 grains cannot be correctly estimated by reason of the uncertainty as to the real quantity of oxide formed during the operation: there might be some left adhering to the retort, or on the other hand there might be more than the due share, derived from former experiments. Supposing then, that 26 grains of oxygen were extracted from the nitric acid, the remaining acid would require the same to be added to re-form the nitric; but by the experiments with oxymuriate of lime it seemed to require 32 grains of oxygen. This difference wants an explanation; I believe the greater error must belong to the 26 grains; perhaps the truth might be approximated best by supposing both to be 30 grains.