It will be noticed, however, that the curve for the white light, while nearly equal to that of the average for the colored lights at the maximum point, nevertheless falls considerably below it at each end. This may possibly be due to the fact that with white light it was only necessary to use an 8-candle-power lamp as a source of light, so that when pieces of ground glass were interposed in order to reduce the intensity of this light, very much greater reduction would occur with this comparatively weak source than would take place with an objective source of light of far greater brilliancy, as was the case with the colored lights. Hence there would be a greater difference in absolute intensity between intensities 1 and 3 with the white light than between intensities 1 and 3 with any of the colored lights, or that represented by their average. Thus the falling of the curve of the white light at each end may possibly be due to the fact that there is a greater difference in intensity represented by these parts of the curve in the case of the white light than is represented by the analogous portions of the curve of the average of the colored lights. It will be remembered that these measurements were obtained when the image was upon the fovea, so that the white obtained was "cone white," and not due in any way to the functioning of the rods. It is interesting to note that the curve of the white is very near that of the average curve of all the colors, though I should hesitate to draw the conclusion from this that "cone white" is due to a mixture or fusion of all the excitations corresponding to the different colors. In regard, however, to relations of the amounts of curvature of the images, there are several further considerations which ought to be noted. In the first place all three measurements were made when the images were entirely on the fovea. In the fovea there are no rods, so, whatever the connection of these facts with Purkinje's phenomenon, it is one which has to do with the functions of the cones alone.