1. GYKI 52466 is a benzodiazepine derivative that has muscle relaxant and anticonvulsant properties thought to be mediated by highly selective, noncompetitive antagonism of non-NMDA receptors. However, recent electrophysiological data showed that, in addition to non-NMDA receptors, the GABAA-receptor associated benzodiazepine site is involved in the depressant effect of GYKI 52466 on spinal reflex transmission. In view of the structural similarities between the 2,3 benzodiazepine derivative GYKI 52466 and 1,4-benzodiazepines such as diazepam, the benzodiazepine site of GABAA receptor complex could also be involved in the anticonvulsant activity of GYKI 52466, which has not yet been proven. This prompted us to study the effect of the benzodiazepine receptor antagonist, flumazenil, on anticonvulsant and adverse effects of GYKI 52466 in different seizure models in mice. The non-NMDA antagonist, NBQX and diazepam were used for comparison. 2. Seizure threshold models for different types of generalized seizures were used. The threshold for maximal (tonic) electroshock seizures (MES) was significantly increased by GYKI 52466 (10-20 mg kg-1), NBQX (80-120 mg kg-1) and diazepam (5 mg kg-1) shortly after i.p. drug administration. The same dose-range of the non-NMDA antagonists also significantly increased the threshold for myoclonic and clonic seizures induced by i.v. infusion of pentylenetetrazol (PTZ), although the magnitude of threshold increases obtained with the respective drugs, differed, at least in part, from that seen in the MES experiments. GYKI 52466 was clearly less potent in increasing PTZ thresholds for myoclonic and clonic seizures than on the MES threshold, while NBQX exerted about the same potency in both models. In contrast to the non-NMDA antagonists, diazepam was capable of increasing themyoclonic and clonic PTZ seizure threshold at much lower doses than the MES threshold. The PTZ threshold for tonic seizures was markedly increased by GYKI 52466, while NBQX and diazepam were clearly less potent in this respect.3. With respect to adverse effects, GYKI 52466 and NBQX induced significant seizure threshold increases in the different seizure models only at doses which caused sedation and ataxia, while diazepam increased the myoclonic and clonic PTZ seizure threshold at doses below those inducing motor impairment.4. Flumazenil (5-20 mg kg-1) antagonized the anticonvulsant and adverse effects of diazepam but not GYKI 52466. Instead, the anticonvulsant effect of GYKI 52466 was potentiated by flumazenil in some experiments. The anticonvulsant activity of NBQX was slightly reduced by flumazenil in the MES model but not in the PTZ test.5. The data indicate that the GABAA receptor-associated benzodiazepine site is not critically involved in anticonvulsant or adverse effects of GYKI 52466. However, both GYKI 52466 and NBQX were unable to increase seizure thresholds at doses below those inducing sedation and motor impairment,thus demonstrating that non-NMDA antagonists lack a selective anticonvulsant action in standard models of generalized seizures.