If we wish to know which is the under and which the upper image, we have only to take a coloured medium, for then a light object reflected from the under surface is of the colour of the medium, while that reflected from the upper surface presents the complemental colour. With dark objects it is the reverse; hence black and white surfaces may be here also conveniently employed. How easily the double images assume and evoke colours will here again be striking. 226. Thirdly, the primary images may be considered as _principal_ images, while the secondary can be, as it were, annexed to these as _accessory_ images. Such an accessory image produces a sort of double form; except that it does not separate itself from the principal object, although it may be said to be always endeavouring to do so. It is with secondary images of this last description that we have to do in prismatic appearances. 227. A surface without a boundary exhibits no appearance of colour when refracted (195). Whatever is seen must be circumscribed by an outline to produce this effect. In other words a figure, an object, is required; this object undergoes an apparent change of place by refraction: the change is however not complete, not clean, not sharp; but incomplete, inasmuch as an accessory image only is produced. 228. In examining every appearance of nature, but especially in examining an important and striking one, we should not remain in one spot, we should not confine ourselves to the insulated fact, nor dwell on it exclusively, but look round through all nature to see where something similar, something that has affinity to it, appears: for it is only by combining analogies that we gradually arrive at a whole which speaks for itself, and requires no further explanation. 229. Thus we here call to mind that in certain cases refraction unquestionably produces double images, as is the case in Iceland spar: similar double images are also apparent in cases of refraction through large rock crystals, and in other instances; phenomena which have not hitherto been sufficiently observed.[1] 230. But since in the case under consideration (227) the question relates not to double but to accessory images, we refer to a phenomenon already adverted to, but not yet thoroughly investigated. We allude to an earlier experiment, in which it appeared that a sort of conflict took place in regard to the retina between a light object and its dark ground, and between a dark object and its light ground (16). The light object in this case appeared larger, the dark one smaller. 231.