The supposition amounts, in fact, to the highly improbable view that we are looking here at a range of stars extending in a cylindrical stratum directly from the eye—a stratum whose section is so very minute in comparison with its breadth that, whereas the whole field within which the spot is included is but small, the distance separating the nearest parts of the group from the furthest is equivalent to the immense distance supposed to separate the sphere of seventh magnitude stars from the extreme limits of our Galaxy. And the great improbability of this view is yet further increased when it is observed that within this spot there is to be seen a very marked tendency to the formation of minor streams, around which the milky light seems to cling. It seems, therefore, wholly improbable that the cluster really has that indefinite longitudinal extension suggested by Professor Nichol. In fact, it becomes practically certain that the milky light comes from orbs really smaller than the seventh magnitude stars in the same field, and clustering round these stars in reality as well as in appearance. The observations applied to this spot may be extended to all clusters of globular form; and where a cluster is not globular in form, but exhibits, on examination, either (1) any tendency within its bounds to stream-formation, or (2) a uniform increase in density as we proceed from any part of the circumference toward the centre, it appears wholly inconceivable that the apparent cluster is not really a cluster, but a long range of stars extending to an enormous distance directly from the eye of the observer. When, in such a case, many stars of the higher magnitudes appear within the cluster, we seem compelled to admit the probability that they belong to it; and, in any case, we can not assign to the furthest parts of the cluster a distance greatly exceeding (_proportionally_) that of the nearest parts. Of a like character is the evidence afforded by narrow streams and necks within the Galaxy itself. If we consider the convolutions over Scorpio, it will seem highly improbable that in each of these we see, not a real convolution or stream, but the edge of a _roll_ of stars. For instance, if a spiral roll of paper be viewed from any point taken at random, the chances are thousands to one against its appearing as a spiral _curve_, and, of course, the chance against several such rolls so appearing is very much greater. The fact that we are assumed to be not very far from the supposed mean plane of the Milky Way would partly remove the difficulty here considered, if it were not that the thickness and extent of the stratum, as compared with the distances of the lucid stars, must necessarily be supposed very great, on the assumption of any approach to uniformity of distribution.