The fact that red, yellow, and blue cannot be primary colours has been mentioned, as even now it is sometimes taught that they are so. As long as the theory of colour principally lay with artists there was reasonable ground for their assumption, since they worked with impure colours, viz. those of pigments; and as we shall see later on the truth of the assumption agreed with such experiments as they would make. When, however, the question was taken up by the physicist with more exact methods of experimenting, and with pure colours, the falsity of the old triad was soon capable of proof. To return from our digression: how it is that three mixed colours can give the sensation of white light is at first sight hard to understand; but a reference to the action of light on a photographic salt helps us in some degree. In the case of a sensitive salt, such as the bromo-iodide of silver, we find that a chemical decomposition is caused by the violet end of the spectrum, and is only feebly affected by any other part, though with prolonged exposure even the red will cause it. The annexed figure (Fig. 33) gives the idea of the relative action of different parts of this violet portion. Fig. 33.--Curve of Sensitiveness of Silver Bromo-iodide. The height of the curve shows the relative effects produced. Now this curve is not symmetrical, but has a maximum effect nearer to the violet end of the spectrum than to the red. The atomic composition of the silver bromo-iodide is probably two atoms of silver and one of bromine and one of iodine oscillating together, and we can conceive of some one atom, the period of whose swings in its molecule is isochronous with some wave-length of light. Further, we can conceive that, like a pendulum whose vibrations are increased in magnitude by well-timed blows, the swing of the atom is also increased, and that eventually it gets beyond the sphere of the attraction of its parent molecule, leaves it, and is attracted to some neighbouring molecule of different constitution, and that thus a chemical change is induced. This we can conceive, but how can other waves, which are not isochronous with the rhythmic swing of the atoms, alter the composition of the molecule? If we have an impulse given to a pendulum exactly timed with the period of oscillation, there is no doubt that the swing is increased. If we have one nearly in accord, it will be found that though the swings are not increased in amplitude so greatly as when there is perfect accord, yet an increased swing is given, and as exact accord is removed further and further, so the increase in the swing of the pendulum gets smaller and smaller. In somewhat the same manner it is possible that many series of waves, differing in wave-length, and therefore in periods of oscillation, may be capable of increasing the amplitude of a swing, and with the photographic salt this probably occurs, with the result which we see in the above figure. Suppose in the eye we have three such