VII] OF BENARD'S EXPERIMENT 503 



are similar ^to but more elegant than James Thomson's tesselated 

 patterns, and both of them are in their own way still more curious 

 than M. Leduc's; for the latter depend on centres of diffusion 

 artificially inserted into the system and determining the number 

 and position of the "cells," while in the others the cells make 

 themselves. In Benard's experiment a thin layer of Hquid is 

 warmed in a copper dish. The hquid is under peculiar conditions of 

 instability, for the least fortuitous excess of heat here or there would 

 suffice to start a current, and we should expect the whole system to 

 be highly unstable and unsymmetrical. But if all be kept carefully 

 uniform, small disturbances appear at random all over the system; 

 a current ascends in the centre of each; and a "steady state," if 

 not a stable equilibrium, is reached in time, when the descending 

 currents, impinging on one another, mark out a "cellular system." 

 If we set the fluid gently in motion to begin with, the first "cell- 

 divisions" will be in the direction of the flow; long tubes appear, 

 or "vessels," as the botanist would be apt to call them. As the 

 flow slows down new cell-boundaries appear, at right angles to the 

 first and at even distances from one another; parallel rows of cells 

 arise, and this transitory stage of partial equilibrium or imperfect 

 symmetry is such as to remind the botanist of his cambium tissues, 

 which are, so to speak, a temporary phase of histological equilibrium. 

 If the impressed motion be not longitudinal but rotary, the first fines 

 of demarcation are spiral curves, followed by orthogonal inter- 

 sections. 



Whether we start with liquid in motion or at rest, symmetry and 

 uniformity are ultimately p,ttained. The cells draw towards uni- 

 formity, but four, five or seven-sided cells are still to be found 

 among the prevailing hexagons. The larger cells grow less,, the 

 smaller enlarge or disappear; where four partition- walls happen to 

 meet, they shift till only three converge; the sides adjust themselves 

 to equal lengths, the angles also to equahty. In the final stage the 

 cells are hexagonal prisms of definite dimensions, which depend on 

 temperature and on the nature and thickness of the liquid layer; 

 molecular forces have not only given us a definite cellular pattern, 

 but also a "fixed cell-size." 



Solid particles in the fluid come to rest in svmmetrical positions. 

 If they be heavier they accumulate in little isolated heaps, each in 



