cigars fills the middle air. In painting this scene it would be usual
to prepare the tone on the palette, and the preparation would be
somewhat after this fashion: ochre warmed with a little red--a pale
violet tinted with lake for the smoke of the cigars.

But such a method of painting would seem to Seurat and Signac to be
artless, primitive, unscientific, childish, _presque du Louvre_--above
all, unscientific. They would say, "Decompose the tone. That tone is
composed of yellow, white, and violet turning towards lake"; and,
having satisfied themselves in what proportions, they would dot their
canvases over with pure yellow and pure white, the interspaces being
filled in with touches of lake and violet, numerous where the smoke is
thickest, diminishing in number where the wreaths vanish into air. Or
let us suppose that it is a blue slated roof that the dottist wishes
to paint. He first looks behind him, to see what is the colour of the
sky. It is an orange sky. He therefore represents the slates by means
of blue dots intermixed with orange and white dots, and--ah! I am
forgetting an important principle in the new method--the complementary
colour which the eye imagines, but does not see. What is the
complementary colour of blue, grey, and orange? Green. Therefore green
must be introduced into the roof; otherwise the harmony would be
incomplete, and therefore in a measure discordant.

Needless to say that a sky painted in this way does not bear looking
into. Close to the spectator it presents the appearance of a pard; but
when he reaches the proper distance there is no denying that the
colours do in a measure unite and assume a tone more or less
equivalent to the tone that would have been obtained by blending the
colours on the palette. "But," cry Seurat and Signac, "an infinitely
purer and more beautiful tone than could have been obtained by any