1. S-potentials recorded from the excised tench retina left undisturbed in the optic cup show colour cells of the two types originally described by Svaetichin & MacNichol (1958).2. One type (green/blue, G/B) is depolarized by signals from green cones and hyperpolarized by blue cones. The other type (red/green, R/G) is depolarized by deep red and hyperpolarized by green cones.3. By superposing spectral flashes upon steady adapting lights it is possible to find a spectral range in which only one kind of cone is effective. In this range the effect of any spectral light may be matched with that of any other provided the energies are linked in a fixed ratio that defines the action spectrum of the pigment.4. The green pigment has an action spectrum with maximum at 540 nm and corresponds well with the pigment that Marks measured in ;green' cones. The blue pigment has not been measured, but it probably corresponds with that found by Marks in ;blue' cones. However, the red pigment whose action spectrum we measured had its maximum at 680 nm, whereas the difference spectrum of Marks's red cone pigment peaked at 620 nm. The 620 nm cones excite the luminosity S-units but not the R/G units.5. In the range where only one type of cone is effective the relation between the light intensity, I(0), and V(0), the S-potential generated (both expressed in suitable units), is given by equation (1) p. 545. It is the relation that would be found if cone signals increased the conductance through a polarized ;S-membrane' in proportion to the flux of caught quanta.