In the human eye, domination of the electroretinogram (ERG) by the b-wave and other postreceptor components ordinarily obscures all but the first few milliseconds of the rod photoreceptor response to a stimulating flash. However, recovery of the rod response after a bright rest flash can be analyzed using a paired-flash paradigm in which the test flash, presented at time zero, is followed at time t by a bright probe flash that rapidly saturates the rods (Birch et al., 1995). In ERG experiments on normal subjects, the hypothesis that a similar method can be used to obtain the full time course of the rod response to test flashes of subsaturating intensity was tested. Rod-only responses to probe flashes presented at varying times t after the test flash were used to derive a family of amplitudes A(t) that represented the putative rod response to the test flash. These rod-only responses to the probe flash were obtained by computational subtraction of the cone-mediated component of each probe flash response. With relatively weak test flashes (11-15 scot-td-s), the time course of the rod response to the test flash derived in this manner was consistent with a four-stage impulse response function of time-to-peak approximately 170 ms. A(170), the amplitude of the derived response at 170 ms, increased with test flash intensity (Itest) to a maximum value Amv and exhibited a dependence on Itest given approximately by the relation, A(170)/Amo = 1 - exp(-kItest), where k = 0.092 (scot-td-s)-1. In steady background light, the falling (i.e. recovery) phase of the derived response began earlier, and the sensitivity parameter k was reduced several-fold from its dark-adapted value. As the sensitivity, sensitivity, kinetics, and light-adaptation properties of the derived response correspond closely with those of photocurrent flash responses previously obtained from isolated rods in vitro, it was concluded that the response derived here from the human ERG approximates the course of the massed in vivo rod response to a test flash.