1. The 2,3-benzodiazepines GYKI 52466, GYKI 53405 and GYKI 53655 antagonized AMPA-induced currents in cultured superior colliculus neurones in a non use-dependent manner (steady state IC50s: GYKI 52466 9.8 +/- 0.6 microM; GYKI 53405 3.1 +/- 0.6 microM; GYKI 53655 0.8 +/- 0.1 microM). 2. Higher concentrations of all three antagonists slowed the onset kinetics and quickened the offset kinetics of AMPA-induced currents indicative of an allosteric interaction with the AMPA recognition site. 3. Cyclothiazide (3-300 microM) dramatically slowed desensitization of AMPA-induced currents and potentiated steady state currents (EC50 10.0 +/- 2.5 microM) to a much greater degree than peak currents. Both tau on and tau off were also increased by cyclothiazide in a concentration-dependent manner (EC50: tau on 42.1 +/- 4.5 microM; tau off 31.6 +/- 6.6 microM). 4. Cyclothiazide (10-100 microM) shifted the concentration-response curves of the 2,3-benzodiazepines to the right. For example, with 10 microM cyclothiazide the IC50s of GYKI 52466 and GYKI 53405 on steady-state AMPA-induced currents were 57.9 +/- 9.5 and 41.6 +/- 1.5 microM, respectively. 5. GYKI 53405 and GYKI 52466 concentration-dependently reversed the effects of cyclothiazide (100 microM) on offset kinetics (GYKI 53405 IC50 16.6 +/- 4.2 microM). However, the 2,3-benzodiazepines were unable to reintroduce desensitization in the presence of cyclothiazide and even concentration-dependently slowed the onset kinetics of AMPA responses further (GYKI 53405 EC50 8.0 +/- 2.8 microM). 6. GYKI 52466 decreased the peak amplitude of hippocampal area CA1 AMPA receptor-mediated excitatory postsynaptic currents (e.p.s.cs) (IC50 10.8 +/- 0.8 microM) with no apparent effect on response kinetics. Cyclothiazide prolonged the decay time constant of AMPA receptor-mediated e.p.s.cs (EC50 35.7 +/- 6.5 microM) with less pronounced effects in slowing e.p.s.c. onset kinetics and increasing e.p.s.c. amplitude. 7. Cyclothiazide (330 microM) shifted the concentration-response curve for the effects of GYKI 52466 on AMPA receptor-mediated e.p.s.c. peak amplitude to the right (GYKI 52466 IC50 26.9 +/- 9.4 microM). Likewise, GYKI 52466 (30-100 microM)) shifted the concentration-response curve for the effects of cyclothiazide on AMPA receptor-mediated e.p.s.c. decay time constants to the right. 8. In conclusion, cyclothiazide and the 2,3-benzodiazepines seem to bind to different sites on AMPA receptors but exert strong allosteric interactions with one another and with other domains such as the agonist recognition site. The interactions of GYKI 52466 and cyclothiazide on AMPA receptor-mediated e.p.s.cs in area CA1 of hippocampal slices provide evidence that the decay time constant of these synaptic events are not governed by desensitization.