(A left) Equivalent circuit of a recorded cone (orange) coupled to neighboring rods (blue) via GJs (OS, outer segment; EOS, reversal potential of the rod/cone light sensitive conductance; GJ, gap junction; iGJ, junctional current; ipip, pipette current; VRod and VCone, rod/cone membrane potential). (A right) When the cone is depolarized, either in current clamp (by constant current injection) such that VCone(DARK) > EOS, or in voltage clamp such that VCone(HOLD) > EOS, a junctional current iGJ will flow into the rods and depolarize them beyond, at, or below EOS. Each of these three possible outcomes is expected to lead to the indicated different combinations of response polarities when delivering dim and bright flashes. (B) A reduced version of the spectral protocol (sequences 1–2 in Figure 1D) was delivered with a cone recorded in current clamp in control conditions (VCone(DARK) = −40 mV) or during depolarization by constant current injection beyond the reversal potential of its light-sensitive conductance (VCone(DARK) = +66 mV). While brightb flash responses reversed polarity, dim flash responses became smaller but did not reverse. This is not compatible with an origin of the dim and bright flash responses in the same electrotonic compartment, and matches one of the predicted outcomes for coupled cones (panel A, VRod(DARK) < EOS). The moderate shift toward G in the spectral preference displayed by this cone (second brightb flash) could be explained by cone–cone coupling and/or by initial recovery from saturation of its coupled rods (see ‘Discussion’). (C) The same experiment as in panel B but performed in voltage clamp in a different cone (VCone(HOLD) = −40 mV and +60 mV). Dim flash responses could not be detected above noise when the cone was depolarized, despite the presence of large inverted brightb flash responses. This matches a different predicted outcome for coupled cones (panel A, VRod(DARK) = EOS). The slight shift toward G in the spectral preference displayed by this cone (second brightb flash) is likely explained by a slow ‘bump’ in the plateau displayed after brightb flashes, present only at −40 mV (not shown, but observed in sequence three of the spectral protocol). The outcome predicted for the case of VRod(DARK) > EOS in panel A was never observed. Records are averages of 3–4 sweeps.