1. After a consideration of the existing data and of the sources of error involved, an arrangement of apparatus, free from these errors, is described for measuring the relative energy necessary in different portions of the spectrum in order to produce a colorless sensation in the eye.
2. Following certain reasoning, it is shown that the reciprocal of this relative energy at any wave-length is proportional to the absorption coefficient of a sensitive substance in the eye. The absorption spectrum of this substance is then mapped out.
3. The curve representing the visibility of the spectrum at very low intensities has exactly the same shape as that for the visibility at high intensities involving color vision. The only difference between them is their position in the spectrum, that at high intensities being 48 µµ farther toward the red.
4. The possibility is considered that the sensitive substances responsible for the two visibility curves are identical, and reasons are developed for the failure to demonstrate optically the presence of a colored substance in the cones. The shift of the high intensity visibility curve toward the red is explained in terms of Kundt's rule for the progressive shift of the absorption maximum of a substance in solvents of increasing refractive index and density.
5. Assuming Kundt's rule, it is deduced that the absorption spectrum of visual purple as measured directly in water solution should not coincide with its position in the rods, because of the greater density and refractive index of the rods. It is then shown that, measured by the position of the visibility curve at low intensities, this shift toward the red actually occurs, and is about 7 or 8 µµ in extent. Examination of the older data consistently confirms this difference of position between the curves representing visibility at low intensities and those representing the absorption spectrum of visual purple in water solution.
6. It is therefore held as a possible hypothesis, capable of direct, experimental verification, that the same substance—visual purple—whose absorption maximum in water solution is at 503 µµ, is dissolved in the rods where its absorption maximum is at 511 µµ, and in the cones where its maximum is at 554 µµ (or at 540 µµ, if macular absorption is taken into account, as indeed it must be).