Without going through this experiment it may at once be stated that the reasoning is correct, for within the limits of error of observation they do so. Having established this proposition, we can next compare _inter se_, the brightness of any or all of the rays of the spectrum by a preliminary comparison with the reflected beam of white light. As in the colour patch apparatus all colours and principal dark lines of the solar spectrum are known by reference to a scale, in making a graphic representation of the results, we first of all plot on paper a scale of equal parts, and at the scale number where a reading is made, the aperture of the sectors in degrees is set up. Thus, suppose with red light the scale number which marked the position of the slit was 59, and the aperture 10°, we should set up at that scale number on the paper a height of 10 on any empyric scale. If in the green at scale No. 38 the sectors had to be closed to 7°, we should set up 7 at that number on the scale. When observations have been made at numerous places in the spectrum, the tops of these ordinates, as they are called, should be joined, and we then get the observed curve of luminosity for the whole spectrum. For convenience’ sake we make the highest point 100, and reduce the other ordinates in proportion. For some purposes it may be advantageous to give the luminosity curve in terms of a scale of wave-lengths. For our purpose, however, it is in general sufficient to use the scale of the instrument. [Illustration: FIG. 19.] Now, if we test the vision of the various types of colour blind by this plan, we should expect to get luminosities at different parts of the spectrum which would give very different forms to these curves. We cannot hope, for instance, that a red-blind who sees no red in the extreme end of the spectrum would show any luminosity in that region, nor that the green-blind should show as much in the green part of the spectrum as those who possess normal colour vision, since one of the sensations is absent. With monochromatic vision there should be a still further departure from the normal curve. That these differences do exist is fully shown in Fig. 19. One of the most striking experiments in colour vision is to place a bright red patch on the screen, and to ask a red-blind to make a match in luminosity with the white. The latter will have to be reduced to almost darkness--a darkness, indeed, that makes the match almost seem incredible. You will notice that the places in the spectrum where the red- and green-blind see grey are by no means places of greatest luminosity. We shall find that these luminosity curves are suggestive when making another investigation into the form of the spectrum curves of the colour sensations.