Visual transduction in rods of the cynomolgus monkey, Macaca fascicularis, was studied by recording membrane current from single outer segments projecting from small pieces of retina. Light flashes evoked transient outward-going photocurrents with saturating amplitudes of up to 34 pA. A flash causing twenty to fifty photoisomerizations gave a response of half the saturating amplitude. The response-stimulus relation was of the form 1-e-x where x is flash strength. The response to a dim flash usually had a time to peak of 150-250 ms and resembled the impulse response of a series of six low-pass filters. From the average spectral sensitivity of ten rods the rhodopsin was estimated to have a peak absorption near 491 nm. The spectral sensitivity of the rods was in good agreement with the average human scotopic visibility curve determined by Crawford (1949), when the human curve was corrected for lens absorption and self-screening of rhodopsin. Fluctuations in the photocurrent evoked by dim lights were consistent with a quantal event about 0.7 pA in peak amplitude. A steady light causing about 100 photoisomerizations s-1 reduced the flash sensitivity to half the dark-adapted value. At higher background levels the rod rapidly saturated. These results support the idea that dim background light desensitizes human scotopic vision by a mechanism central to the rod outer segments while scotopic saturation may occur within the outer segments. Recovery of the photocurrent after bright flashes was marked by quantized step-like events. The events had the properties expected if bleached rhodopsin in the disks occasionally caused an abrupt blockage of the dark current over about one-twentieth of the length of the outer segment. It is suggested that superposition of these events after bleaching may contribute to the threshold elevation measured psychophysically. The current in darkness showed random fluctuations which disappeared in bright light. The continuous component of the noise had a variance of about 0.03 pA2 and a power spectrum that fell to half near 3 Hz. A second component, consisting of discrete events resembling single-photon responses, was estimated to occur at a rate of 0.006 s-1. It is suggested that the continuous component of the noise may be removed from scotopic vision by a thresholding operation near the rod output.