1. The spread of activation and background desensitization in rods was studied by recording membrane current from single outer segments in pieces of isolated toad retina.2. Flash sensitivity changed slightly along the outer segment, falling by about 30% from base to tip.3. When only the distal half of an outer segment was in the recording pipette, illumination of the unrecorded part elicited little or no photocurrent at the recorded part, indicating that a photoisomerization does not cause activation of the entire outer segment.4. With diffuse illumination of an outer segment fully drawn into the pipette, the intensity-response relations at fixed times were invariant in form for most of the rising phase of the flash response and were considerably steeper than the Michaelis relation. The observed relation was consistent with a model in which a photoisomerization blocks all channels over a short region of the outer segment.5. With illumination restricted to a narrow transverse slit, the intensity-response relations at fixed times were much less steep, as would be expected for a very limited longitudinal spread of activation. An upper limit for the effective longitudinal diffusion coefficient of the internal transmitter was estimated to be about 3 x 10(-7) cm(2) sec(-1). This corresponds to a space constant for longitudinal spread of transmitter of about 3 mum at the time of the dim response peak.6. The time course of flash responses elicited with light positioned either on the edge or on the centre of the outer segment was very similar.7. Desensitization resulting from steady illumination by a transverse slit was also localized longitudinally. A linear desensitization parameter T, defined in Results, decayed approximately exponentially along the outer segment, on either side of the site of photoisomerization, with a space constant of about 6 mum.8. Transverse spread of desensitization was more effective than longitudinal spread.9. After turning off a dim diffuse background light, the decay of T was roughly exponential with a time constant of several seconds. From this, and the steady state space constant of 6 mum, it is estimated that the effective longitudinal diffusion coefficient for a ;desensitizing substance' would also be about 10(-7) cm(2) sec(-1).10. The restricted longitudinal spread of activation and desensitization may be explained by the barrier to diffusion presented by the stacked membranous disks in the outer segment. This baffling reduces the effective longitudinal diffusion coefficient to about 1/50 that of ordinary aqueous diffusion, but it does not significantly affect transverse spread.