>_> iiKvsiTiis i>v (i\'v<;kx and hvdkockn 



5. The elevation of llie cistern of the nianobaronieter above tlie centre of the 

 Ljldlte while the latter was filling with oxygen; 



(i. 'I'hc correction to the length of the scale of my niann-liaiiMiieter ; and 



7. The force of gravity at my lahoiatory. 



In the second series, the first three of these seven factnis were eliminated, for 

 thermometers were not used, and the expansion <>f the gl<»l)e in which the oxygen 

 had its pressure measured was compensated Ijy the equal exj>ansi(>n of the similar 

 globe containing the standard volume of hydrogen. In tlie thiid series, the cistern 

 of the niaiio-barometer, which was then used as a syphon barometer, was at the level 

 of the centre of the globe, so that the observations were reduced by taking account 

 simply of the length of my scale and of the force of gravity at my laboratory. 



In the forty-one determinations which are included in the three series, eight 

 different globes were used, in order to eliminate the effect of accidental errors in 

 the determinations of their cai)acity. 



2. MEASUKEMENT Ol'' (A I'AcrillCS OK CI.OUKS. 



My measurements of the capacities of the globes used I'oi' weighing oxygen in- 

 volved three processes, as follows : 



1. The determination of the weight of the glolic in aii-; 



•_'. The determination of the weight required to be added to maintain the 

 trlolie in equilibrium when immersed in water of known temperature ; and 



'A. The determination of tlie loss of weight of the globe when full of water 

 a!i<l immersed in water of the same temperature. (loml»ining the first determination 

 with the second, we get the external volinuc of tlie globe ; from the third and first 

 determinations we get the solid contents of the material of w liicii the globe is com- 

 posed. Tlie difference. bet ween the external volume an<l the solid contents gives 

 the ca[)acity. 



It is obvious that a determination of the weight of the water recpiired to fill 

 the globe at a known temperature would form a moi-e direct and a more accurate 

 determination of its capacity ; but this would have re(piired the use of a balance 

 carrying twenty-five kilogrammes in each pan, which I should have had to jiur- 

 clia.se for the purpose. As the ol lier expenses of the investigation were a heavy 

 burden, the method just described was used ; it is liopcd thai the results of the (h-- 

 teiniinatioiis will be found to have iieeii sullieieiitly acciii'ate. >\s the erinrsof tiie 

 calibrations seem to have lieeii made niueh less than the accidental errors of single 

 exjierimeiits, as the numbei- of calibrations of each globe was not too small, and as 

 so many globes wei'e caiil)rated and used in the determination of density, no error 

 of sensible magnitude is likel\- to ha\'e been caused. 



