sectors are gradually closed, and we now come to such a reduction that the red is absolutely invisible; but the green still shines out. It is losing its colour somewhat, and appears of a bluish tint. The reason of this change of hue in the latter we shall shortly see. The sectors are withdrawn and the red re-appears, and is as bright as the green. The slit of the collimator is next opened, and there is no doubt that the red is much brighter than the green, as it was purposely made at the beginning of the experiment. The same class of experiment might have been repeated with the green and violet or the red and violet, and the same kind of results would have been obtained. The violet would have been the last to disappear when the green was so reduced in luminosity that it appeared in the ordinary brilliant spectrum to be equal to the violet ray selected. When the green was of the luminosity given by a slit equal in width to that of the violet, the violet would have disappeared first, owing to its feeble brightness to begin with. Now, if we measure a feebly illuminated spectrum we must adopt some special means to exclude all light, except that of the comparison light and the ray to be measured. This we can do by the box which is shown in the next diagram (Fig. 24). [Illustration: FIG. 24.] At one end of a box, shown in plan, is an eye-piece, E. The other end has at its centre a white square of paper of 1½-inch scale. The monochromatic beam _a_, coming from the spectrum through the slit S and the reference beam _b_ of white light, are reflected from glass mirrors M₁, M₂ to apertures in opposite sides of the box, and from close to these apertures by the right-angled prisms P₁ P₂, so as to fall on and cover S. Rods R₁, R₂ are inserted in the box in the paths of the beams, so that the opposite halos of S are illuminated. Diaphragms inside the box cut off any stray light, and rotating sectors placed at A and B regulate the intensity of the beams as required. The sector A is rotated with a previously determined-on aperture; the white light coming through B is altered till the luminosity of the two on the screen, as seen through E, are the same. Every part of the spectrum can be measured in this way; the result is shown in the diagram. Fig. 25 (the measures will be found at page 215 in the appendix). In this case the orange light at D where it fell on the screen was equal to 1/132 of an amyl-acetate light, which, in its turn, is closely ·8 of a standard candle. In the same figure the luminosity curve of the ordinary bright spectrum is given for reference, and it can be seen how the point of maximum luminosity is shifted into the green, lying almost over the E line of the solar spectrum. The maximum, of course, has been made 100 as before, for had it been drawn to the same scale as the other, the form of the curve would not have been demonstrated. There is a remarkable resemblance between it and the