1. The mechanism of action of picrotoxin on GABA receptor channels in rat sympathetic neurones has been investigated with whole-cell clamp. In addition, the action of picrotoxin on single GABA channels has been examined in outside-out membrane patches from these cells. 2. Picrotoxin, at concentrations which dramatically reduced the amplitude of whole-cell GABA currents, did not alter the spectral time constants or single-channel conductance estimated by analysis of GABA-activated current noise. This was observed at potentials both negative and positive to the GABA reversal potential (i.e. for both inward and outward GABA currents). In control conditions, the slow and fast time constants from GABA noise were 40 +/- 14 ms and 2 +/- 0.4 ms, while the estimated single-channel conductance was 14 +/- 2 pS. In the presence of picrotoxin, the time constants and estimated single-channel conductance were 41 +/- 5 ms, 2.7 +/- 0.6 ms and 15 +/- 2.3 pS. 3. Picrotoxin did not alter the shape of the whole-cell GABA current-voltage relationship, indicating that the steady-state block was not voltage dependent. The lack of effect of picrotoxin on the GABA noise spectra and the lack of outward rectification makes it unlikely that picrotoxin acts by a simple voltage-dependent (or voltage-independent) channel blocking mechanism. In the presence of picrotoxin the reversal potential for GABA remained at approximately 0 mV in symmetrical chloride. 4. Distributions of total burst durations, obtained from single-channel records with low concentrations of GABA, were fitted with three or four exponential components. Picrotoxin had no consistent effect on the time constants of the total burst length distributions. It also did not alter the amplitude of the main conductance state. However, picrotoxin did reduce the frequency of channel openings. 5. The application of brief ionophoretic pulses of GABA, to cells under whole-cell voltage clamp, revealed that the rate of onset of block by picrotoxin was accelerated in the presence of GABA. In the absence of agonist, picrotoxin produced a more slowly equilibrating block. 6. Our data are consistent with a mechanism whereby picrotoxin binds preferentially to an agonist bound form of the receptor and stabilizes an agonist-bound shut state. This could, for example, mean that picrotoxin enhances the occurrence of a desensitized state or an allosterically blocked state.