Can we in any way find from these methods the colour sensation curves? I think we can. Suppose we have a second instrument exactly like the first placed side by side with it, we can then throw two patches of colour on the two adjacent white surfaces, and we can mix with either, or both of them, as much white light as we choose. From the second instrument let us throw all the spectrum colours in succession on to the one surface, and on to the other the three primary colours mixed in such proportions as to match them accurately. This plan is, I venture to think, a better way of obtaining the value of colours in terms of standard colours than that adopted by Maxwell. This method gives the values directly, and not by calculation from matches with white. Let us place one slit near each of the extreme ends of the spectrum; that in the red near the red lithium line, and another a little beyond G in the violet of the spectrum, whilst the third slit should be in the exact position in the green, where the _green-blind sees grey_. Now it might be a matter of dispute as to whether one was entitled to make this last one of the positions for the slits, for we use it entirely on the assumption that two of the colour sensations which we suppose we possess are identical with those of the green-blind. This might be, or might not be, the case; but I think it can be shown very easily that the assumption we are making is more than probably exact. Having the slits in these positions, we may endeavour to match the spectrum orange. We mix the red and the green lights together, and find that the best mixture is always paler than the orange, but by adding a small quantity of white to the orange we at once form a match. In the same way if we have a greenish-blue to match, we shall find that we can only make the match when we add a little white to the simple colour. Now let us shift the position of the slit in the green just a little--a very little--towards the blue, and again try to match orange. Do what we will we cannot find apertures to the slits which will give us the colour, though it be diluted with white. It will be too blue or too red, but never exactly orange. This tells us that there is too much blue in the green we are using. Next, shift the slit a little towards the red below our fixed position, and endeavour to match the blue. We shall find that this, too, becomes impracticable. The blue is either too green or too violet, telling us that our mixture contains too much green. As the neutral point of the colour blind is the only position for the green slit which enables us to make a good match to both the orange and the blue, it follows that this must be the point where these two colour sensations are so arranged as to be in the proportions required to form white when green is added; that is, that there is neither an excess of red nor an excess of violet. To come back to our measures of colour. We can make up every spectrum colour