I continued to hold this amended doctrine for some years. It is true that from time to time I observed that the tone of the neutral tint varied considerably in different shadows; but still the divergencies were not such as to shake my faith in the dogma. By-and-bye, however, in a popular work on Optics, I met with the statement that the colour of a shadow is always the complement of the colour of the light casting it. Not seeing the wherefore of this alleged law, which seemed moreover to conflict with my established belief, I was led to study the matter as a question of causation. _Why_ are shadows coloured? and what determines the colour? were the queries that suggested themselves. In seeking answers, it soon became manifest {166} that as a space in shadow is a space from which the _direct_ light alone is excluded, and into which the _indirect_ light (namely, that reflected by surrounding objects, by the clouds and by the sky) continues to fall, the colour of a shadow must partake of the colour of everything that can either radiate or reflect light into it. Hence, the colour of a shadow must be _the average colour of the diffused light;_ and must vary, as that varies, with the colours of all surrounding things. Thus was at once explained the inconstancy I had already noticed; and I presently recognized in Nature that which the theory implies—namely, that a shadow may have any colour whatever, according to circumstances. Under a clear sky, and with no trees, hedges, houses, or other objects at hand, shadows are of a pure blue. During a red sunset, mixture of the yellow light from the upper part of the western sky with the blue light from the eastern sky, produces green shadows. Go near to a gas-lamp on a moonlight night, and a pencil-case placed at right angles to a piece of paper will be found to cast a purple-blue shadow and a yellow-grey shadow, produced by the gas and the moon respectively. And there are conditions it would take too long here to describe, under which two parts of the same shadow are differently coloured. All which facts became obvious to me as soon as I knew that they must exist. Here, then, respecting certain simple phenomena that are hourly visible, are three successive convictions; each of them based on years of observation; each of them held with unhesitating confidence; and yet only one—as I now believe—true. But for the help of an hypothesis, I should probably have remained in the common belief that shadows are black. And but for the help of another hypothesis, I should probably have remained in the half-true belief that they are neutral tint.