Although seizure models using systemic administration of the chemoconvulsant pentylenetetrazol (PTZ) for induction of generalized clonic seizures in rodents are widely employed to identify potential anticonvulsants, the important role of diverse technical, biological and pharmacological factors in interpretation of results obtained with these models is often not recognized. The aim of this study was to delineate factors other than sex, age, diet, climate, and circadian rhythms, which are generally known. For this purpose, experiments with 8 clinically established antiepileptic drugs were undertaken in the following PTZ models: (1) the threshold for different types of PTZ seizures, i.e., initial myoclonic twitch, generalized clonus with loss of righting reflexes, and tonic backward extension of forelimbs (forelimb tonus), in mice; (2) the traditional PTZ seizure test with s.c. injection of the CD97 for generalized clonic seizures in mice; and (3) the s.c. PTZ seizure test in rats. In rats, in addition to evaluating drug effects on generalized clonic seizures, a ranking system was used to determine drug effects on other seizure types. When drugs were dissolved in vehicles which themselves did not exert effects on seizure susceptibility, the most important factors which influenced drug potencies were: (1) bishaped dose-response curves, i.e., a decline in anticonvulsant dose-response at high doses of some drugs, leading to misinterpretations of drug efficacy if only a single high drug dosage is tested; (2) effects of route of PTZ administration (i.v. infusion vs. s.c. injection) on estimation of anticonvulsant potency; (3) species differences in drug metabolism; (4) differences in drug potencies calculated on the basis of administered doses compared to potency calculations based on 'active' drug concentrations in plasma; (5) qualitative and quantitative species differences in drug actions; (6) endpoints used for PTZ tests; (7) misleading predictions from PTZ seizure models. Analysis of anticonvulsant drug actions indicated that myoclonic or clonic seizures induced by i.v. or s.c. PTZ might be suitable for predicting efficacy against myoclonic petit mal seizures in humans, but certainly not to predict efficacy against absence seizures. Tonic seizures induced by PTZ were blocked by drugs, such as ethosuximide, which exert no effect on tonic seizures in humans. In order to reduce the variability among estimates of anticonvulsant activity in PTZ seizure models, the various factors delineated in this study should be rigidly controlled in experimental situations involving assay of anticonvulsant agents.