We used a fibre electrode in the lower conjunctival sac of the human eye to record the a-wave of the photopic electroretinogram elicited in response to dim red flashes, delivered in the presence of a rod-saturating blue background, before and after exposure of the eye to bright white illumination that bleached a significant fraction of cone photopigment. Responses were recorded from two normal subjects whose pupils were maximally dilated. A range of intensities of bleaching light were used, from 500 to 3000 photopic cd m(-2), and exposures were made sufficiently long in duration to achieve a steady-state bleach. In addition, responses were also recorded following shorter durations of exposures to the highest intensity (3000 cd m(-2)); these durations ranged from 5 to 60 s. The amplitude of the a-wave response to dim flashes was reduced following the exposures, with brighter or longer exposures causing greater reduction. The amplitude then recovered within about 4 min to the prebleach level. The amplitudes measured at ca 15 ms after the flash were used to derive the effective intensity of the flashes, thereby quantifying the fraction of photopigment available at the time of delivery of each flash. Recovery from all exposures in both subjects followed a common time course, which could be described well by a model of pigment kinetics based on rate-limited regeneration, where the initial rate of recovery following a total bleach was ca 50% of the total pigment per minute, and the residual pigment level for half the maximal rate was ca 20% of the total pigment. The same parameters, together with a fixed photosensitivity, could account for the steady-state pigment levels seen at each bleaching intensity, and also for the fraction of pigment bleached following exposures of different duration at the highest intensity. The dim-flash ERG thus provides a novel method for assessing pigment regeneration in vivo. Our finding that pigment regeneration follows rate-limited kinetics may explain previous reports of pigment regeneration deviating from first order kinetics. We present a model of regeneration in which the rate limit arises from a limitation in the delivery of 11-cis-retinoid to the photoreceptor outer segments.