We examined the actions of a carboxylated analogue of pregnanolone ((3alpha,5beta)-20-oxopregnane-3-carboxylic acid; 3alphaCOOH5betaP) on receptors composed of glycine receptor alpha3 subunits, expressed in Xenopus oocytes. This analogue both inhibits and potentiates this receptor; potentiation increases with more negative membrane potentials while block increases with less negative membrane potentials. We used a second analogue ((3alpha,5beta)-3-hydroxymethylpregnan-20-one; 3alphaCH(2)OH5betaP) to examine the mechanism for voltage-dependent potentiation. This analogue potentiates but does not block the glycine alpha3 receptor. Steady-state responses and current relaxations following voltage jumps support the idea that the voltage dependence of potentiation indirectly arises from a voltage dependence for channel activation by glycine, rather than an intrinsic voltage dependence for potentiation. Potentiation results from a slowing of the channel deactivation rate. In the absence of steroid, at a low [glycine] current relaxations after a voltage jump show two exponential components, with a weighted average time constant of approximately 425 ms (-50 mV, 22 degrees C). The rate for channel deactivation increases at more negative potentials (e-fold per 170 mV) whereas activation decreases (e-fold per 230 mV). The probability a channel is active at a high [glycine] is greater than 0.9. The addition of 10 microM 3alphaCH(2)OH5betaP decreases the deactivation rate by 6.3-fold (-50 mV). Voltage-dependent block by 3alphaCOOH5betaP is consistent with simple open-channel block, with voltage dependence reflecting interactions of the charge on the analogue with the electrical field. Block and unblock have equal but opposite dependence on membrane potential, and the charge on 3alphaCOOH5betaP senses approximately 70% of the membrane field at the blocking site. The apparent forward rate for block, however, is very slow (2 x 10(5) m(-1) s(-1)).