Febkuaey 23, 1900.] 



SCIENCE. 



283 



fairly well with the mean of all. For in- 

 stance, Koyes' results show that degree of 

 concordance which would justify us in ex- 

 pecting that, if he were to make 100 or 

 1000 experiments, his final mean would be 

 as likely as not to be larger or smaller by 1 

 part in 9500, and his result differs from 

 that which we accept by 1 part in 900. So 

 Cooke and Eichards assign a value which 

 is just as likely as not to be within 1 part 

 in 8000 of the result which they would 

 have obtained by multiplying observations ; 

 and it differs from that which we accept by 

 1 part in 1500. But do we know that their 

 means, and the means of all published re- 

 sults taken together, are not in error by 1 

 part in 900 ? The concordance of the re- 

 sults of a single experimenter, and the 

 agreement of different experiments, does not 

 justify us in asserting that we do. 



In determining the atomic weight of 

 oxygen, it has been, somewhat difficult to 

 determine directly all three of the quanti- 

 ties involved, and so to make what Stas 

 called a complete synthesis. Berzelius, 

 Dumas, and Erdmann and Marchand, 

 weighed oxygen and weighed water, thus 

 determining hydrogen by difference. More 

 recently, Dittmar and Henderson and Leduc 

 used the same method. Cooke and Rich- 

 ards, and Keiser weighed hydrogen and 

 weighed water, while Eayleigh and ISToyes 

 weighed hydrogen and weighed oxygen. 

 Any proceeding which weighs hydrogen 

 directly has a great advantage in precision ; 

 different determinations in a given series 

 agree better among themselves, and the 

 series of different experimenters also agree 

 better. But there is also a second, more 

 important advantage. We have reason to 

 believe that the constant errors involved in 

 weighing hydrogen are small, for it is pos- 

 sible to obtain hydrogen with less than 

 TTTTiTT or even ^^-jf of its weight of im- 

 purity. There is no difficulty in weighing 

 oxygen or water with accuracy, so if we 



weigh hydrogen and also weigh either oxy- 

 gen or water, we may hope for a near ap- 

 proximation to the true value of the ratio 

 sought. 



We may hope, but we cannot hnow. We 

 may believe that our hydrogen is pure, and 

 that there was no error through leakage. 

 But an unsafe stop-cock might make the 

 apparent weight of the hydrogen in a series 

 of experiments seem always smaller than 

 the fact, and might yet leave the individual 

 experiments so concordant with each other 

 as to seem trustworthy. 



If, however, we can weigh hydrogen and 

 can weigh oxygen, and then combine them 

 and weigh the water produced, we can at 

 least give a better reason for our hope, if 

 we find that the product is nearly equal to 

 the sum of the components. The manipu- 

 lation in this case is costly, and is so diffi- 

 cult, and involves so many minute details, 

 that not many have patience and time 

 sufficient for it, so that no great number of 

 such complete syntheses has been made, 

 and these few were made in conditions but 

 little varied. When such complete syntheses 

 shall have been made by different observers, 

 with those variations of apparatus and 

 method which may seem wise to them, we 

 shall be able to judge of the magnitude of 

 the errors to be feared. If such results are 

 not concordant, we shall have much to 

 learn as to sources of error ; but we now 

 see some reason to expect that they will 

 not be discordant. However, even if they 

 are not discordant, we are not to excuse 

 ourselves from further study of sources of 

 error. 



Before repeating determinations so trou- 

 blesome, and before studying unknown 

 sources of errors not yet detected, the ex- 

 perimenter should receive all possible assist- 

 ance from chemical theory and from criti- 

 cism. For some, that criticism may be most 

 profitable which is friendly and sympathetic; 

 but, for the experimenter, the value of the 



