Intracellular recordings from cones and rods in the retina of the turtle, Pseudemys scripta elegans, revealed that in darkness the cell voltage fluctuated spontaneously about its mean level. The fluctuations were reduced during bright steady illmination of the cell often to a level close to that obtained with the electrode outside the cell where the noise did not change significantly during illumination. 2. The magnitude of the intrinsic dark noise (voltage variance in darkness minus voltage variance in strong light) varied widely from cell to cell. In the noisiest cones it was about 0-4 mV2 while in quiet cones it was often as low as 0-01 mV2. The noise appeared radom and could be fitted by a Gaussian probability density function. 3. The spread of voltage in the network of coupled photoreceptors was estimated by measuring the spatial profile of the response to a brief flash of constant intensity moved across the retina. For a light stimulus in the form of a long narrow slit, the peak flash response usually decayed exponentially with displacement from the centred position. 4. For maximum responses less than about 5 mV in cones, the length constant of exponential decay, lambda, varied from less than 10 mum to greater than 35 mum, and the values obtained in opposite directions were often unequal. Background illumination did not significantly change lambda. In cells with extremely narrow spatial profiles, an exponential fit to the decay could not be made reliably. 5. Occasionally the spatial profiles had definite secondary peaks. In the most pronounced examples in a red-sensitive cone and in a rod the maxima were separated by about 20 and 50 mum respectively; for each, one peak was approximately as sharp as the optical stimulator while the second was broader. 6. Cones with short length constants displayed high dark noise while cones with long length constants were relatively quiet. 7. Three models of electrical coupling between cells were investigated: one based on a distributed network, one on a discrete square grid arrangement, and one on a discrete hexagonal array. Each model predicts a strong dependence of both noise and input resistance on length constant, and for tightly coupled cells each predicts that voltage variance is proportional to lambda-2. 8. The measured relationship between voltage variance and lambda in a large sample of cones was well described by both discrete models when the average cell spacing was taken to be approximately 15 mum. 9...
