WITHDRAWN FROM THE ACTION OF GRAVITY. 645 



would necessarily of itself induce a cause of irregularity; and 

 moreover it would not allow of a uniform distribution of the 

 movements of transport, because there would be opposition in 

 regard to these movements, at least in the terminal constrictions. 



It may therefore be regarded as very probable that the transfor- 

 mation takes place according to one or the other of the two first 

 methods, and never according to the third, i. e. that things will 

 be so arranged that the figure which is transformed may have 

 for its terminal portions either two dilatations, or one constric- 

 tion and one dilatation, but not two constrictions. In the former 

 case, as we have seen, the movement of the liquid of all the con- 

 strictions would ensue on both sides simultaneously ; and in the 

 second, this movement would occur in all in one and the same 

 direction. If this is really the natural arrangement of the phse- 

 nomenon, we can also understand how it will be preserved 

 even when it is disturbed in its regularity by slight extraneous 

 causes. Now this, as we shall see, is confirmed by the experi- 

 ments relating to the mercurial cylinder : although the trans- 

 formation of this cylinder has rarely yielded a perfectly regular 

 system of spheres, I have found in the great majority of the re- 

 sults, either that each of the solid bases was occupied by a mass 

 little less in diameter than the isolated spheres, or that one of 

 the bases was occupied by a mass of this kind and the other by 

 a much smaller mass. 



53. For the sake of brevity, let us denominate divisions of the 

 cylinder those portions of the figure, each of which furnishes a 

 sphere, whether we consider these portions in the imagination 

 as in the cylinder itself, before the commencement of the trans- 

 formation, or whether w^e take them during the accomplishment 

 of the phaenomenon, i. e. during the modifications which they 

 undergo in arriving at the spherical form. The length of a divi- 

 sion is evidently that distance which, during the transformation, 

 is comprised between the necks of two adjacent constrictions, 

 consequently it is equal to the sum of the lengths of a dilatation 

 and two semi-constrictions. Let us therefore see how the length 

 in question, i. e. that of a division, may be deduced from the 

 result of an experiment. Let us suppose the transformation to be 

 perfectly regular, and let \ be the length of a division, / that of 

 the cylinder, and n the number of isolated spheres found after 

 the termination of the phaenomenon. Each of these spheres 

 being furnished by a complete division, and each of the two ter- 



