Several previous studies have shown that competitive and uncompetitive N-methyl-D-aspartate (NMDA) receptor antagonists are ineffective as anticonvulsants in fully amygdala kindled rats, i.e. a widely-used model of complex-partial seizures (limbic epilepsy). In the present experiments, different categories of ligands for the NMDA receptor-associated glycine modulatory site were evaluated in the kindling model in rats. For this purpose, the 'silent' glycine antagonist 7-chlorokynurenic acid (7-CKA), the low efficacy glycine partial agonist (+)-HA-966 (R(+)-3-amino-1-hydroxypyrrolid- 2-one), and the high efficacy glycine partial agonist D-cycloserine (D-CS) were used. In view of the poor brain penetration of some of these compounds after systemic administration, all drugs were injected bilaterally into the lateral ventricles of fully kindled rat. Anticonvulsant activity was evaluated by determination of the threshold for induction of focal afterdischarges in the amygdala (ADT) and by monitoring seizure parameters (seizure severity, seizure duration, afterdischarge duration) at ADT current. In addition to anticonvulsant activity, the behavioural adverse effects of test drugs were determined. All three drugs increased the focal seizure threshold, although at different potencies. In terms of injected dosages and percentage ADT increase, the most potent compound was 7-CKA, followed by (+)-HA-966 and D-CS. None of the drugs induced behavioural adverse effects or motor impairment (measured in the rotarod test) at anticonvulsant doses. However, D-CS produced proconvulsant effects at higher doses. The data demonstrate that pharmacological manipulation of the glycine modulatory site of the NMDA receptor is an effective means of increasing seizure threshold in amygdala-kindled rats.(ABSTRACT TRUNCATED AT 250 WORDS)