with these three colours, and finally divide the luminosity curve into the _colour luminosity_. In all these matches the violet luminosity is very small indeed compared with the red or green. A match with white is now made by a mixture of all these colours, and you will see, from the images of the slits on the screen, that the _luminosity_ of the violet is almost a negligible quantity compared with the others. We may, therefore, as a first approximation, divide up the luminosity curve into two parts, one being the luminosity of the green in the different colours and the other of the red. The green, however, is made up of red, of violet, and of an excess of green sensation, which in this case comes practically to a mixture of white with the green sensation. How can we tell how much is green and how much is white? Suppose I, as a normal-eyed person, compare the luminosity of the colour coming through the red slit with that coming through the green slit, and then get the green-blind to do the same, it is evident that any excess in the luminosity as measured by myself over that measured by the green-blind must be due to the green sensation, and we can also see how much red and violet make up his white. We shall not be far wrong, then, in apportioning the constituents of the white thus found between the green and the red; the violet being, for the time being, negligible. We must subtract the red sensation from the green _colour_ curve and add it to the red colour curve: the two curves will then be very closely the curves of the _red and green sensations_. By causing the green-blind to make mixtures of red and violet for all the colours of their spectrum, we can arrive at what must be finally taken away or given to these curves, though such addition or subtraction of violet will be small when the luminosities are considered. The accompanying figure (Fig. 22) gives an idea of the shape and general features of these curves. It may be remarked that we can check the general accuracy of the measures of the colour mixtures by calculating or measuring the areas of the two colour curves, the red and the green. If accurate, they should bear the same ratio that the _luminosities_ of the two colours bear to each other (when mixed with a little violet, which is practically negligible) to form white light. So far, then, we can utilize the luminosity methods to calculate and to trace the sensation curves for the normal eye. It will not escape your notice that the maximum heights of these two component curves are nowhere near the parts of the spectrum where the colour is the purest. Another check to these curves may also be made by taking the difference in the ordinates of the luminosity curves of the colour blind and the normal eyed. Too much stress must not, however, for the moment, be laid on this, as this method depends on the absolute correctness of the scale of the ordinates in the curves. It must be recollected that to the former