SUMMARY1. The common dark adaptation curve exhibits two branches; the course of the rod branch cannot normally be 'measured at early times since it lies above the observed cone thresholds. In this paper we measure it.2. This is done by observing the negative after-image against a uniform background critically adjusted in luminance.3. Adjacent to the bleached area to be studied is a second area more strongly bleached. If the background intensity is below threshold for the less bleached area it will not be seen there; but if the background is above that threshold, this area will be seen brighter than the other.4. The dark adapted threshold on the less bleached area is therefore the background luminance which just permits the two areas to be distinguished in the after-image.5. After 5 min cones have quite recovered, and thus have no afterimage to contaminate the rod image.6. The rod curve measured by after-image is traced over 5 units of log threshold: it is an exponential with half life of 4*5 min, and coincides with the time course of regeneration of rhodopsin in man.