AND HATU) OK TIIKIU ATOMIC VVKIOIITS. 



OXYGEN FROM I'OTASSIUM CHI.dKATE. 



Sc = "S -Sec ^rfJ! 



|-5 I il .-aa-i III 



1 2 727.04 12.0179 [.42920 



2 I 746.93 12-3951 1.42860 



3 2 769.76 12.7227 1.42906 



4 I 773-22 12.8400 '.42957 



5 I 772.22 12.8192 7.42910 



6 2 778.68 12.8745 1.42930 



7 2 778.04 12.8628 1.42945 



OXYGEN 1!Y ELECTROLYSIS. 



8 I 774.39 12.8572 1.42932 



9 2 750.12 12.3983 1.42908 



10 I 769.83 12.7796 1.42910 



11 8 76535 23.7671 1-42951 



12 3 761-52 28.7134 1-42933 



13 9 77498 21.9675 1-42905 



14 5 772.55 31-3039 1-42914 



15 7 747-S8 21.6150 1.42849 



16 7 754-99 21.8274 1.42894 



17 7 763.80 22.0808 1.42886 



If we increase tlie mean by one tliirty-thousantltb,* we get 



B = 1.42!)17 gr. ± 0.000048. 



40. O.XYOKN. KIN.U, IIESULT KOIJ THK DENSITY. 



Tlie values foiiiicl liy the different methods used are as follows: 



By use of thermometer and manometer D = i .42879 ± .00003 4 



By compensation D = i .42887 ± .00004 8 



By use of ice and barometer D = r .42917 ± .00004 '^ 



The combination of these results into a final mean nuist be left mostly to the judg- 

 ment of those interested in the matter. The probable errors of the three means 

 would indicate that the first method should have double weiglit, which would be 

 very improper, in my judgment. This method involves more easy manipulation, 

 but gives less security against constant error. So also with the second method. 

 The manipulation of the third involves more accidental errors, but involves no con- 

 stant errors which are not common to the other methods, while it avoids sonje. I 

 shall therefore compute a mean in which the third result is given double weight; 

 from which we get 



Do = 1.42900 ± 0.000034. 



* See note, page 2S. 



