We can imitate even more exactly the matches that they would make by matching white light with a mixture of red, green, and violet of the proper hues, and covering up the red or green slit, and then placing the test-skein and the matches in the colour so formed. From the other skeins viewed in the same light can be picked out the matches which would be possible. There is very little chance, if any, of a mistake about them being made when this plan is adopted. CHAPTER XIV. Holmgren’s test, although a qualitative one, is most accurate in allowing a diagnosis to be formed, but it sometimes happens that a candidate is not satisfied that he has failed in passing the test, and wishes for another examination. Such a re-examination is best carried out by the spectrum method, which I will now describe. The test with the spectrum is a very decisive one, and can be carried out with the patch apparatus (Fig. 3), page 19. Personally, I like to have some idea of the kind of colour blindness, if any, which exists by first using the Holmgren test. Should these tests show that a candidate is colour blind in any degree, a very excellent beginning is to try and find his neutral point in the spectrum--if he has one. To arrive at it we place two patches of light on the screen, one of colour and the other of white, the rotating sectors being in the last-named beam, and ask him to say when the two colours appear alike. It must be remembered that white is coloured from the effect of contrast as long as the colour alongside differs from it. A good _point de depart_ is with the slit in the yellow, then to move it into the red, and then gradually to push it into the green. When here, if colour blind, he will say, “The two patches are nearly alike, but that the white is rather pink or green,” as the slit gets further towards the blue. The operator, whilst changing the colour, alters the sectors so that the luminosities are about the same. A point will be reached when the colour blind will say, “Now they are both alike, but one is rather darker than the other.” The sectors are altered until he says they are both alike, and the observation is satisfactory when he declares the two patches of light are both alike in colour and in darkness. It is curious how misleading the word brightness is to some people who are uneducated. I find it much safer to ask which is the darker colour, rather than which is the brighter. A little patience will always enable you to get a good observation. The place in the spectrum which is the neutral point is now noted. The neutral point is again found, but this time commencing in the blue. The same procedure is adopted as before, and we thus get a second reading for it, and the two will be found to be very close to one another. In difficult cases, four or five observations may be made, and the mean taken as a close approximation. So far the spectrum test has not shown whether the observer is red- or green-blind, except