1. In turtle cones the steady-state relation between the internal potential and log light intensity was much flatter in the steady state than it was at 30 msec after the beginning of a step of light; this is attributed to a desensitization which develops with a delay of 50-100 msec.2. When a weak flash was superposed on a steady background light which hyperpolarized the cone by 3-6 mV the amplitude of the linear response to a flash was reduced to 1/e and the time to maximum was shortened from about 110 to 70 msec; the response also became diphasic. With stronger background lights the flash sensitivity continued to fall, but the time to maximum did not become shorter than 40-50 msec and lengthened again with very strong lights.3. In cones the flash sensitivity S(F) was reduced to half its dark value S(F) (D) by a light intensity of 1/S(F) (D)zeta where zeta is about 20 sec/V.4. At low levels of background light, about two-thirds of the change in sensitivity was time-dependent and one-third was attributable to the ;instantaneous non-linearity' described in the previous paper.5. The reduction in time to peak and the decrease in sensitivity produced by a background light which hyperpolarized by about 3 mV was little affected by changing the diameter of the area illuminated from 12 to 800 mum.6. An experiment with a rod showed that a very weak light which hyperpolarized by only 0.5 mV decreased the linear response to 1/e and shortened the time to maximum from 300 to 180 msec.7. With weak or moderate flashes the time-dependent desensitization lagged behind the potential by 50-100 msec.8. The desensitization and shortening of time scale which persisted after a flash or step were associated with an after-hyperpolarization. The relaxation of potential, sensitivity and time scale became slower as the preceding illumination was increased from 10(3) to 10(10) photons mum(-2); the increase seemed to occur in steps involving components which relaxed with time constants of the order of 0.1, 1, 10 and 100 sec. A rebound phenomenon was observed after steps longer than 30 sec and with equivalent intensities greater than 10(5) photons mum(-2) sec(-1).9. Several of the observations are explained by a hypothesis in which the central assumption is that the particles which block the ionic channels are degraded or removed by an autocatalytic reaction.