1. S-potentials were recorded in fish from units which never responded by depolarization. These hyperpolarizing units are the L-units of Svaetichin & MacNichol (1958).2. Figure 5 shows some sets of action spectra from a single unit. For each curve the criterion of action was hyperpolarization to a fixed level, by lights of various wave-lengths. When these lights fell upon zero background (circles) the curves show that two kinds of cone contribute to the action spectrum, one with the 620 nm pigment of Marks and one with the 680 nm pigment of Naka & Rushton (1966a).3. When the lights fell upon (i) a fixed green background (triangles, Fig. 5), or (ii) a fixed red one (squares), the action spectra changed in a way that indicated greater prominence of (i) the 680 nm system (ii) the 540 nm green system that was not conspicuous without adaptation to red.4. These observations (on the tench Tinca) are contrary to the conclusions of Svaetichin & McNichol (on Gerridae) that the action spectrum is unaltered in shape by adaptation to coloured lights. The contribution of the green cones, for example, was actually absolutely greater under deep red adaptation.5. It is concluded that L-units receive signals from 680, 620, 540 nm and possibly also the blue cones, that the quantum catch in all these contribute to the hyperpolarization produced, but their interaction is more complicated than the simple addition of independent cone effects.