BRUSH— A KINETIC THEORY OF GRAVITATION. 175 



Large weighings like these are made with a 2-kilogram Becker 

 balance of the most improved type, easily sensitive to a tenth milli- 

 gram with full load. 



The members of each pair of cylinders are made so closely alike 

 in weight because they are to be superposed on the pendulum rods; 

 and any difference in weight would shift the center of gravity of the 

 pair. 



The purpose of the second pair of bobs is, first, to permit com- 

 parison of bobs of the same metal — I am often asked to do this ; and, 

 second, by superposing the zinc or the bismuth, to provide bobs of 

 double the former weight and very different in shape. It was not 

 expected, however, that even such large variation of weight and shape 

 of bobs would throw any new light on the subject; and it did not, as 

 will appear. This will be gratifying to such of my friends as have 

 been bothered with doubts in these respects. 



Next, the old and new bismuth bobs were mounted on the two 

 pendulum rods ; synchronized, exchanged, and period difference 

 measured ; again synchronized, exchanged, and period difference 

 measured, all as formerly done with ::inc and bismuth bobs. In 

 addition to this, each bob, after exchanging, was turned over and 

 tested again so as to average out any small errors that might have 

 been left when center of gravity adjustments were made. In these 

 experiments the small paper vanes on the old bismuth cylinder were 

 eliminated in effect by turning them into the line of swing. 



In every instance the new bismuth gained over the old bismuth, 

 but not much. Averaging all the measurements, the mean gain was 

 about .6 mm. per hour measured in middle of swing with a reading 

 telescope. Whole amplitude was about 11.4 cm. Hence somewhere 

 about 190 hours would have been required for a full beat gain, such 

 as takes place in about 5 hours when zinc and bismuth are compared. 

 Therefore, we may call the new bismuth-bismuth effect about 2^ 

 per cent, as large as the normal zinc-bismuth effect. 



The above shows too much difference in behavior between old and 

 new bismuth cylinders to be attributed to experimental errors. It 

 seems not unlikely, then, that there is a real mass-weight difference 

 between the old and new bismuth cylinders. It is believed that this 

 difference is due to difference in crystalline structure. If this is the 



