1. The effects of (-)-baclofen, muscimol and phaclofen on endogenous gamma-aminobutyric acid (GABA) release from rat cortical slices, spinal cord slices and entire retinas were studied. 2. The spontaneous resting release of GABA from the three tissues was 3 to 6 pmol mg-1 wet wt 10 min-1. Depolarization of cortical slices with KCl (50 mM) (high-K) produced an 8 fold increase in GABA release but high-K did not evoke an increased release of GABA from spinal slices or retinas. 3. When rats were injected with gamma-vinyl-GABA (250 mg kg-1 i.p.) (GVG) 18 h before death, the tissue GABA stores were increased 3 to 6 fold and high-K then evoked striking Ca-dependent releases of GABA from all three tissues. Thus, in subsequent experiments, unless otherwise stated, the nervous tissues were taken from GVG-treated rats. 4. (-)-Baclofen (10 microM) significantly reduced the K-evoked release of GABA from cortical and spinal slices but retinal release was not affected, even at a concentration of (+/-)-baclofen of 1 mM. For cortical slices, the IC50 for baclofen was approximately 5.2 microM. The inhibitory effect of baclofen on GABA release from cortical slices also occurred in slices prepared from saline-injected rats, indicating that GVG treatment did not qualitatively affect the results. 5. The inhibitory effect of (-)-baclofen on the K-evoked release of GABA from cortical and spinal slices was antagonised by phaclofen (500 microM), confirming that baclofen was producing its effects by acting at the GABAB-receptor. 6. Phaclofen (500 microM) increased the spontaneous resting release of GABA from cortical slices taken from GVG-treated rats but not from saline-injected rats. Phaclofen did not increase GABA release from spinal slices or retinas taken from GVG-treated rats. 7. Baclofen (10 microM) significantly reduced the K-evoked release from cortical slices of glutamate, aspartate, glycine and taurine. 8. Muscimol (10 microM) and delta-aminolaevulinic acid (10 microM) had no effect on either the resting or Kevoked release of GABA from cortical slices prepared from saline-injected or GVG-treated rats. 9. The results obtained with cortical and spinal slices are consistent with the presence of inhibitory GABAB-autoreceptors. The phaclofen-induced increase in GABA release from cortical slices taken from GVG-treated rats, but not from saline-injected rats, implies that under conditions of high GABA release, considerable feedback inhibition is occurring via activation of the GABAB inhibitory autoreceptors. No evidence was found for GABAB-autoreceptors on retinal GABAergic amacrine cells or for GABAA-autoreceptors in cortical slices or spinal cord slices.