On definite Proportions, 5 



fncnt proportions, their respective quantities are ulwdys 

 indicated by some of these umpie expression'; ; 1 A ivitli 

 1 6, the niinimun), 1 A with 1| B, or perhaps rather 2 A 

 with 3 B. 1 A with 2 B, and 1 A ivith 4 B. But there is 

 no example in my experitnents ot the proporiion 1 A 

 with 3 B. 



It will further be shown, that ivhen tun lodies, A and 

 B, have loth affinities trUh tuo oiheis, C and D, the quan- 

 tities of C and D wli/.h are saturated hy A, are precisely in 

 the same proportion os the quaidili,es which ar^i ia/wated 

 ly B. Thus, since 100 parts of lead are capable of com- 

 bining in the first deiiree with 15 d of sulphur and 7*8 of 

 oxygen, and 100 parts of iron, according to the analysis 

 which will hereafter lie related, combine in the first degree 

 •with 58*8 of sulphur, the composition of the protoxide of 

 iron may be computed by the simple proportion 15'6 : 

 7*8 = 58-8 : 29*4, so that ioo parts of iron require to be 

 combined with 29-4 of oxygen. This is confirmed by the 

 experiments which I am about to communicate, and all 

 binary combinations may be determined in the same man- 

 ner. It has been ably demonstrated by Richter, that a si- 

 milar principle is applicable to the combinations of salts. 



It is obvious that the result of these calculations, sup- 

 posing them to be well founded, must be susceptible of 

 much greater accuracy and certamly than the common 

 analysis. I have endeavoured to give the greatest possible 

 precision to the analyses which I shall here relate, and I 

 have repeated the most important of them more than once, 

 before I allowed myself to depend on then). These are 

 certainly free from errors of more tnan one or two parts in 

 a thousand, and the others are withm at most one half per 

 cent, of the truth, but still only accurate enough to give 

 approximations in computation. Perhaps we shall never 

 succeed in analysing substances so accurately, as to obtain 

 results agreeing witli the proportions of the compooent 

 parts to the last place of decimals : on the other hand, it 

 will not be impossiljlc, when we have a number of very 

 accurate analyses, to correct them so by calculation, ihat 

 all the elements of the computation of a combination may 

 afford precisely the same resiih. 



I shall arrange my experiments in the order which seems 

 most convenient for the illustration of the subject, and I 

 shall totally refrain from all theorizing. How far tJie results 

 of the experiments confirm the theory, will be obvious with- 

 out particular comn»ent, and the ideas to which they lead 



A 3 will 



