1. Properties of the light-sensitive voltage noise in cones in the retina of the turtle, Pseudemys scripta elegans, have been studied by intracellular recording.2. Suppression of the noise by light was a function of the hyperpolarizing response of a cone but not of the size or pattern of illumination.3. Power density spectra of the noise were fitted in many cones by the product of two Lorentzians with characteristic time constants tau(1) and tau(2) averaging 40 and 7 msec respectively. The spectra of some cells were peaked and could be fitted by a resonance curve.4. Spectra in dim light exhibited decreased low frequency power. They could often be fitted by a product of two Lorentzians using the same value of tau(2) as used in darkness but decreasing tau(1) and the zero frequency asymptote. An e-fold reduction in tau(1) occurred with lights which hyperpolarized by 4-7 mV.5. Injection of hyperpolarizing currents of about 0.1-0.2 nA into weakly coupled cones reduced the noise, and also reduced the sensitivity to dim flashes.6. The variance-voltage relation during steady illumination of different intensities differed from cone to cone. Dim lights increased the noise in some cells and decreased it in others, but moderately bright lights which gave steady responses of more than about one third maximal reduced the noise in all cells.7. When the cell was transiently depolarized during the differentiated component following steady illumination, the noise was less than it was after prolonged darkness.8. In the after-effect of bright light, the time course of recovery of noise was the same as that of flash sensitivity and voltage. The noise was reduced e-fold for hyperpolarizations averaging 3 mV while for sensitivity this reduction occurred for 1.3 mV. For a given hyperpolarization the noise was lower during the after-effect than during steady dim illumination.9. When a series of dim flashes was delivered to a cone, no significant increase in variance over the dark noise was detected during the photo-response. This implies that each photoisomerization evokes no more than about 1.5 muV at the peak of the response in a coupled cone, corresponding to about 50 muV in an isolated cone.10. The elementary shot events underlying the noise are about 100 muV in amplitude in an isolated cone, have a characteristic time constant of 16-60 msec and reflect unit conductance fluctuations of about 16 pS (S, Siemen identical with Omega(-1)).11. It is concluded that the noise source is internal to the cones. We postulate that the noise arises from opening and closing of the light-sensitive ionic channels in the outer segment, and that in darkness there is a residual concentration of the blocking substance which on average closes up to about one third of the channels. It seems likely that the unit event involves a considerable number of blocking molecules and ionic channels.