EVOLUTION OF THE STELLAR SYSTEMS. 231 
revolving at the quickest in some hundreds of thousands of years. If the paths be 
elliptical, the major axes of these ellipses would be of the same order of magnitude as 
the distance which separates us from a Centauri; while if the paths be parabolic or 
hyperbolic, the two objects would pass and then separate forever. On the other hand we 
can conceive of nothing which could diminish the dimensions of a very long ellipse, 
unless it be something analogous to a resisting medium. Such a medium to be effective 
in reducing the size of the orbits would have to act for a great period of time, and 
besides would probably be visible in space as diffused nebulosity. No nebulosity is 
observed about revolving double stars, nor is there any evidence of a sensible resisting 
medium either among the stars or in our own solar system. We may therefore reject 
the idea that the dimensions of the orbits were originally very large, and have since been 
diminished. As the orbits are now of the size of those of our greater planets, and there- 
fore comparatively small, it follows that the stellar systems have originated by some 
process other than by the union of separate stars. 
As a nebula is a very rare and expanded mass, and is yet held in equilibrium by 
the pressure and attraction of its parts, it necessarily rotates very slowly ; and hence 
when it divides into two parts under the acceleration of rotation due to secular condensa- 
tion, the orbit pursued by the detached mass must be of small eccentricity. For even if 
the forces producing separation could be exerted suddenly to produce a violent rupture, 
the detached mass in pursuing its eccentric orbit would again come to periastron, where it 
would encounter resistance in its orbital motion, and the result of the grazing collision 
would be a diminution of the size of the orbit, and consequently an exaggeration of the 
resistance at the next periastron passage; in this way the system would very soon 
degenerate into one mass. On the other hand were the initial eccentricity small, the 
newly-divided masses would pass freely, and when the orbit eventually became highly 
eccentric the secular contraction in the size of the masses would prevent disturbance at 
periastron. Subsequent collision could not possibly occur, because the periastron distance 
would steadily though perhaps only slowly increase as the stars are pushed asunder and 
the orbit is rendered constantly more and more eccentric. 
It follows therefore that in the beginning the orbits are only slightly eccentric, and 
that the eccentricity is developed gradually as the result of secular tidal friction working 
through immense ages. Accordingly in the elongation of the orbits now observed we 
see the trace of a cause which has been working for millions of years. The existence of 
this cause and its effects on stellar cosmogony could probably never be inferred except 
in the manner by which I approached the problem. On the one hand it appears that 
we have inferred the true cause of the expansion and elongation of the stellar orbits, 
while on the other the trace left by this cause has enabled us to detect the existence of 
A, BsS:—-VOW. XUX. 2D: 
