1. The electrical coupling between cones of known spectral sensitivity in the peripheral part of the turtle's retina was studied by passing current through a micro-electrode inserted into one cone and recording with a second micro-electrode inserted into a neighbouring cone. 2. Spatial sensitivity profiles were determined by recording flash responses to a long narrow strip of light which was moved across the impaled cones in orthogonal directions. These measurements gave both the length constant lambda of electrical spread in the cone network and the separation of the two cones. 3. The cone separation determined from the spatial profiles agreed closely with that measured directly by injecting a fluorescent dye into two cones. 4. The length constant lambda varied from 18 to 39 micron with a mean of 25 micron for red-sensitive cones and 26 micron for green-sensitive cones. 5. The majority of cone pairs studied were electrically coupled provided they had the same spectral sensitivity and were separated by less than 60 micron: thirty-two out of thirty-six red-red pairs, two out of two green-green pairs, none out of eight red-green pairs: no blue cones were observed. 6. The strength of electrical coupling was expressed as a mutual resistance defined as the voltage in one cell divided by the current flowing into the other. Mutual resistances decreased from a maximum value of about 30 M omega at separations close to zero to 0.2 M omega, the lower limit of detectable coupling at separations of about 60 micron. Mutual resistances were always positive and were independent of which cell was directly polarized. The coupling seemed to be ohmic and any rectification or non-linearity probably arose in the cone membranes rather than in the coupling resistances. 7. The results were analysed in terms of the Lamb & Simon (1977) theories of square and hexagonal lattices, which approximate to the continuous sheet model except in the case of the cone to which current is applied. 8. The total membrane resistance of a single cone was estimated as 100--300 M omega and the connecting resistances as 100 M omega for a square array and 170 M omega for a hexagonal array. The input resistance of a cone in the network was 25--50 M omega. Lower values were often obtained but may be due to injury by the micro-electrodes. 9. The time constant of an isolated cone was estimated as about 20 msec and the capacity as about 100 pF. 10. Discrepancies between experimental findings and theoretical predictions of the hexagonal or square array models were tentatively attributed to an overestimate of lambda resulting from light scattering.