Solitary rod inner segments were isolated from salamander retinae. Their Ca current was studied with the 'whole-cell, gigaseal' technique (Hamill, Marty, Neher, Sakmann & Sigworth, 1981). The soluble constituents of the cytoplasm exchanged with the solution in the pipette. The external solution could be changed during continuous perfusion. Membrane voltage was controlled with a voltage clamp. After permeant ions other than Ca were replaced with impermeant ions (i.e. tetraethylammonium as a cation, and aspartate or methanesulphonate as an anion), an inward current remained. It activated at approximately -40 mV, reached a maximum at approximately 0 mV, and decreased as the membrane was further depolarized. The size of the current increased when Ba was substituted for external Ca. The current was blocked when Ca was replaced with Co. The voltage at which the current was half-maximum shifted from approximately -22 to -31 mV during the initial 3 min of an experiment. The maximum amplitude of the current continuously declined during the entire course of an experiment. The time course for activation of the Ca current following a step of depolarization could be described by the sum of two exponentials. The time constant of the slower exponential was voltage dependent. Deactivation following repolarization could also be described by the sum of two exponentials. Both time constants for deactivation were independent of voltage (between -30 and 0 mV) and faster than the slower time constant for activation. When the internal Ca concentration was buffered by 10 mM-EGTA, the Ca current did not inactivate during several seconds of maintained depolarization. When the concentration of EGTA was reduced to 0.1 mM, the Ca current declined and the membrane conductance decreased during several seconds of maintained depolarization. This inactivation was incomplete and only occurred after a substantial quantity of Ca entered. Following repolarization the Ca conductance recovered from inactivation. In contrast, the continuous decline observed during the course of an experiment (item 3) was not reversible. The difference suggests that inactivation and the decline are distinct processes.