568 PROCEEDINGS OF THE AMERICAN ACADEMY. 



measurements of radiation pressure than the early measurements of 

 the present writers. The principal difference between the methods 

 employed by him and by the writers for determining the pressure was 

 that he used very thin metallic vanes surrounded by gas at extremely 

 low pressures, thus following Maxwell's suggestion literally, while the 

 writers used silvered glass vanes and worked at large gas pressures for 

 which the gas action had been carefully and exhaustively studied and 

 found to be negligibly small for short exposures. From our knowledge 

 of the variation of gas action in different vacua, we feel sure that our 

 method would not have been successful in high vacua because of the 

 relatively large gas action. Professor Lebedew's own results, with 

 blackened vanes of lower heat conductivity, show that his success in 

 eliminating gas disturbance was due to the high heat conductivity of 

 thin vanes rather than to the high vacua employed. 



Professor Lebedew's * estimate of the accuracy of his work is such as 

 to admit of possible errors of twenty per cent in his final results. An 

 analysis of Professor Lebedew's paper and comparison with our prelimi- 

 nary experitneuts seems to show that his accidental errors were larger 

 than ours, but through the undiscovered false resistance in the bolometer 

 our final results were somewhat further from the theory than his. Either 

 of the above researches would have been sufficient to establish the exist- 

 ence of a pressure due to radiation, but neither research offered, in our 

 judgment, a satisfactory quantitative confirmation of the Maxwell-Bartoli 

 theory. 



Later Pressure Measurements. 



Description of Apparatus. — The Torsion Balance. 



The form of suspension of the torsion balance, used to measure radia- 

 tion pressure in the present study, is seen in Fig. 2. The rotation axis 

 a b was a fine rod of drawn glass. A drawn glass cross-arm c, bent down 

 at either end into a small hook, was attached to the axis. The surfaces 

 G and D, which received the light beam, were circular microscope cover- 

 glasses, 12.8 mm. in diameter and 0.17 mm. thick, weighing approxi- 

 mately 51 mgs. each. To distinguish the two vanes from each other, in 

 case individual differences should appear in the measurements, and also 

 to mark the two faces of each vane for subsequent recognition, a letter O 

 was marked on one, and Don the other by diamond scratches. Through 

 each glass, a hole 0.5 mm. or less in diameter, was drilled near the edge, 



* P. Lebedew, Ann. Pliys., VI. 457 (1901). 



