effects are not really exceptional, though they are seldom noticed; but for the purpose of art, what appears to be exceptional must be definitely regarded as so, and for this reason discarded by the artist who desires to paint a good picture. Generally then, the value of colour lies firstly in its correspondence with nature, for upon this depends its harmony and the assistance it lends to the recognition of the beauty in the whole composition. Beyond this it may or may not have an ephemeral value according to local conditions. In any case colour must ever be subordinated to design in a picture, and this is what Poussin meant when he said that particular attention to colour is an obstacle to the art student. BOOK II ILLUSIONS IN THE PAINTER'S ART INTRODUCTORY The painter is occupied in a perpetual struggle to produce an illusion. He does not directly aim for this, but except in the very highest art where ideals are realized, the better the picture he paints, the greater the illusion. The natural test of the value of his work is its correspondence with nature, and the nearer it so corresponds, the more complete the illusion. But the whole picture is never an illusion (we leave out of consideration those instances where artificial devices are used to conceal the surroundings of the actual painted surface), for the frame and other material evidence inform us of the art. The illusion, when it exists, is forced upon our minds from moment to moment as our eyes travel over the work. It occurs to us perhaps that a face "lives," that the drapery is true to life, that the tones are real, and so on, and obviously these circumstances cannot impress us in this way unless we are momentarily deceived. And it is a sign of good quality in the work when we are so struck. This does not mean that the closer the imitation, the better the picture: on the contrary it is rare to find a good work of art produced by an exact imitator. The duty of the artist is to generalize everything that can be generalized without departing from the character of the thing represented. True there are degrees of generalization which depend on the nature of the design, the size of the work, the accessories, and other matters, but if a just balance of generalization be secured throughout, then the imitation is better than a closely detailed reproduction, because a work is always involuntarily judged from general, and not from particular, experience. A portrait for instance is a much better work of art if we can say of it "This is a good portrait of a man," than if we are compelled to confine ourselves to "This is a good portrait of Mr. Jones," even if the lineaments of the particular countenance are better defined in the latter example than in the former. The illusion would be stronger, for we are more intimately acquainted with "a man" than with "Mr. Jones." And so with accessories. An exceptionally fine rose or cabbage is never so good in a painting as