The kinetics of the dark-adapted salamander rod photocurrent response to flashes producing from 10 to 105 photoisomerizations (Φ) were investigated in normal Ringer’s solution, and in a choline solution that clamps calcium near its resting level. For saturating intensities ranging from ~102 to 104Φ, the recovery phases of the responses in choline were nearly invariant in form. Responses in Ringer’s were similarly invariant for saturating intensities from ~103 to 104Φ. In both solutions, recoveries to flashes in these intensity ranges translated on the time axis a constant amount (τc) per e-fold increment in flash intensity, and exhibited exponentially decaying “tail phases” with time constant τc. The difference in recovery half-times for responses in choline and Ringer’s to the same saturating flash was 5–7 s. Above 104Φ, recoveries in both solutions were systematically slower, and translation invariance broke down. Theoretical analysis of the translation-invariant responses established that τc must represent the time constant of inactivation of the disc-associated cascade intermediate (R*, G*, or PDE*) having the longest lifetime, and that the cGMP hydrolysis and cGMP-channel activation reactions are such as to conserve this time constant. Theoretical analysis also demonstrated that the 5–7-s shift in recovery half-times between responses in Ringer’s and in choline is largely (4–6 s) accounted for by the calcium-dependent activation of guanylyl cyclase, with the residual (1–2 s) likely caused by an effect of calcium on an intermediate with a nondominant time constant. Analytical expressions for the dim-flash response in calcium clamp and Ringer’s are derived, and it is shown that the difference in the responses under the two conditions can be accounted for quantitatively by cyclase activation. Application of these expressions yields an estimate of the calcium buffering capacity of the rod at rest of ~20, much lower than previous estimates.