But how are we to measure this extinction of light at the different parts of the spectrum? This is a problem which I have attacked during the last few years by a variety of methods; but as is the case with almost every scientific problem, when the mode of attack is reduced to its simplest form, it yields the more readily to solution. If we have a box, like that figured in Fig. 27, and combine it with our colour patch apparatus, the problem is solved. B B is a closed box 3 feet long and about 1 foot high and wide, having two similar apertures 1½ inch in diameter in the positions shown. The aperture at the side is covered on the inside by a piece of glass _a_, ground on both sides, and a tube T is inserted, in which diaphragms, D, of various apertures can be inserted at pleasure. The most convenient form of diaphragm is that supplied with photographic lenses--an iris diaphragm. E is a tube fitted at the end of the box through which the screen S is viewed. S is black except in the centre, where a white disc is fastened to it. A mirror, M, placed as shown, reflects the light scattered by the ground glass on to the screen S. The rotating sectors are placed where shown, and are in such a position that they can be readily adjusted by the observer. The patch of any desired colour of the spectrum is thrown on _a_, and an appropriate size of diaphragm used, so that when the sectors are not less than 5° to 10° open, the light totally disappears. We can now make observations throughout the whole spectrum, and knowing the value of the different apertures of the diaphragm and the angular opening of the rotating sectors, we can at once find the amount of reduction of the particular part of the spectrum that is being required in order to just extinguish all traces of light from the white disc at the end of the box. From these measures we can readily construct a curve or curves which will graphically show the reduction given to the different parts of the spectrum. Fig. 28 gives the curve of extinction for ordinary normal colour vision. The spectrum was of such a brilliance that the intensity of the square patch of light formed on _a_ of the orange light (D) was exactly that of an amyl-acetate lamp, placed at one foot distance from the receiving screen. Knowing this, the actual luminosity of all the other rays of the spectrum can be derived from the curve of luminosity (see Fig. 20). Extinguishing the various parts of the spectrum by this plan, it is found that the red rays cease to stimulate the retina sufficiently to give any appearance of light long before the green rays are extinguished. It is only the rays in the extreme violet of the spectrum, and which consequently possess very feeble luminosity, that make any approach towards requiring the same amount of reduction as the red rays. [Illustration: FIG. 28.]