liECENT STUDIES IN GRAVITATION. 209 



set in various positions. But the variation in the attraction was 

 merel\' of the order of error of experiment. In another experi- 

 ment the attracted masses were small calc spar cr^'stal eviinders, 

 weighing a little more than one-half gm. each. But again there was 

 no evidence of variation in the attraction with variation of axial 

 direction. 



Practically the same problem was attacked in a diti'erent wa}^ by 

 Mr, Gray and nwself." We tried to find whether a ([uartz crystal 

 sphere had any directive action on another quartz crystal sphere close 

 to it, whether they tended to set with theii- axes parallel or crossed. 



It may easily be seen that this is the same problem by considering 

 what nmst happen if there is any difference in the attraction between 

 two such spheres when their axes are parallel and when they are 

 crossed. Suppose, for example, that the attraction is always greater 

 when their axes are parallel, and this seems a reasonable supposition, 

 inasmuch as in straightforward crystallization successive parts of the 

 cr3'stal are added to the existing crystal, all with their axes parallel. 

 Begin, then, with two quartz crystal spheres near each other, with 

 their axes in the same plane, but perpendicular to each other. Remove 

 one to a ver}- great distance, doing work against their nuitual attrac- 

 tions. Then, when it is quite out of range of appreciable action, turn 

 it round till its axis is parallel to that of the fixed cr3\stal. This 

 absorbs no work if done slowh'. Then let it return. The force on 

 the return journey at every point is greater than the force on the 

 outgoing journe}', and more work will be got out than was put in. 

 When the sphere is in its first position, turn it round till the axes are 

 again at right angles. Then work must be done on turning it through 

 this right angle to supply the difference between the outgoing and 

 incoming works. For, if no work were done in the turning, we could 

 go through C3'cle after cycle, always getting a balance of energy over, 

 and this would, I think, imply either a cooling of the crystals or a 

 diminution in their weight, neither supposition being admissible. We 

 are led, then, to say that if the attraction with parallel axes exceeds 

 that with crossed axes, there must be a directive action resisting the 

 turn from the crossed to the parallel positions. And, conversely, a 

 directive action implies axial variation in gravitation. 



The straightforward mode of testing the existence of this directive 

 action wouM consist in hanging up one sphere by a wire or thread 

 and turning the other round into various positions and observing 

 whether the hanging sphere tended to twist out of position. But the 

 action, if it exists, is so minute and the disturbances due to air cur- 

 rents are so great that it w^ould be extremely difficult to observe its 

 effect directlj". It occurred to us that we might call in the aid of the 



^ Phil. Trans. 192, 1899, A, p. 245. 



SM 1901 11: 



