598 PROCEEDINGS OF THE AMERICAN ACADEMY. 



also the part (0.06 F^ not eliminated when we average the two columns 

 to obtain column 13. 



Applying this method to all the results of Table III (with the 

 exception of those results taken with poor mirrors as shown by our 

 notes), the gas action present in the ballistic deflections " through air " 

 is 0.8 per cent. Applying the corresponding data and equations to 

 Table IV, the gas action present in the red glass values is 1.1 per cent 

 and in the water cell values, 0.3 per cent. The sign of F^ comes out 

 negative, which means that the gas action was suction. 



This reasoning assumes that the glass faces of the vanes during the six 

 seconds exposure are not warmed by absorption nor by the conduction of 

 heat through the thin glass from the silver coating. The effect of any 

 such absorption or conduction would be to diminish the computed gas 

 action. As estimated from the static observations, the gas action in the 

 ballistic measurements is comparable in magnitude with the computed 

 values obtained above, and of the same sign. Both results show that the 

 uneliminated gas action by the most liberal estimate cannot have ex- 

 ceeded 1 per cent of the radiation pressure. Because of its smallness 

 and indefiniteness no correction for gas action has been made to the final 

 pressure values. If corrections were applied its effect would be to 

 slightly reduce the observed pressures. 



Aside from the measurements of pressure and energy for which the 

 probable errors are given, the percentage accuracies in the other 

 measurements entering into the computations, and their effects upon 

 the final result follow: — 



1. Quantities which affect individual series : 



(a) Pressure values, — 



Period of balance T, accurate to 0.2% ; effect on result 0.0% 



Lever arm of balance /, " 0.1%; " " 0.0% 



Constant of galv'meter(?2, " 0.5%; " « 0.0% 



Estimate of possible error 



due to changing ratio of 



period of G<2, to length of 



exposure of bolometer 0.4%; " " 0.1% 



