Precisely similar effects must be produced by the loss of weight in the two components of a binary star. Here both components are radiating away energy, and so are simultaneously losing weight. Detailed calculation shews that they must continually recede from one another, but that the shape of their orbit will undergo no change. Neither separately nor in combination do the two effects just described explain either the shapes or the sizes of the observed orbits of binary stars as a whole. To interpret these we must call on yet a third agency, the gravitational forces from passing stars. We have already seen how these account for the statistical distribution of orbits which is actually observed. The combination of all three agencies, tidal friction, extending over millions of years, loss of weight, extending over millions of millions of years, and disturbance from passing stars, extending over a similar period, is responsible for the evolution of binary star systems. Their aggregate effect is to widen the distance between the two stars, while at the same time knocking the orbit out of shape. SUBDIVISION. While these changes are going on in the orbital arrangement of a binary system, the two components are themselves changing their physical condition on account of their continual loss of weight, and, as with the parent stars, this loss of weight will generally result in a shrinkage in the size of the star. The shrinkage of either component of the system causes its shape to run through the sequence of configurations we have already enumerated, and if the shrinkage continues for long enough, the component may end by further dividing into two separate masses. Either or both of the constituents of a binary system may subdivide into binary sub-systems in this way, resulting in a system of either three or four stars. H. N. Russell has shewn mathematically that when a binary system _P_, _Q_ divides into a triple system, _P_, _q_, _qʹ_, through _Q_ breaking up into two constituents _q_, _qʹ_, the distance between _q_ and _qʹ_ cannot be more than about a fifth of the original distance _PQ_. This theoretical law is well confirmed by observation. Fig. 13 shews a typical multiple system, and we notice that the separations in each of the various sub-systems are all quite small in comparison with those of the main systems. [Illustration: Fig. 13. A typical multiple star.] The development of the hypothetical primitive chaos has now been traced through five generations of astronomical bodies, _chaos—nebulae—stars—binary systems—sub-systems_,