We have reported recently that the seizure model and experimental protocol may markedly influence anticonvulsant tolerance and withdrawal characteristics of benzodiazepine (BDZ) receptor ligands so that predictions on tolerance and dependence liability of novel drugs should be based on a battery of chronic experiments. In the present study, we compared BDZ receptor ligands with different intrinsic efficacy and/or gamma-aminobutyric acidA/BDZ receptor subtype selectivity in two seizure models, by using different experimental approaches to assess the tolerance and dependence liability. In one approach, mice were chronically treated with either diazepam, clonazepam, clobazam or the novel anxiolytic and anticonvulsant beta-carboline derivative abecarnil for 4 weeks, at doses which were about equipotent to increase the threshold for myoclonic seizures induced by pentylenetetrazole. Anticonvulsant activity was determined several times during the period of chronic treatment as well as up to 2 weeks after termination of treatment in the same group of animals per drug. The threshold for electroshock-induced tonic seizures was used as a second seizure model in separate groups of mice. In another approach, drug treatment protocols were the same but the seizures were induced only twice during the 4-week period of treatment to reduce the number of trials which could lead to "learned" tolerance. In additional groups of mice, the seizure thresholds were only determined before and after the period of treatment to assess whether repeated seizure induction during the chronic treatment affects the development of dependence. All four drugs lost anticonvulsant activity during the chronic treatment in the different models and experimental approaches, without indication for a significant involvement of learned tolerance. However, marked protocol-related differences were seen with respect to withdrawal symptoms, i.e., measures of physical dependence-inducing properties of the different drugs. For instance, the pentylenetetrazole model was more sensitive than the electroshock-induced tonic seizures model to detect significant decreases in the seizure threshold in the withdrawal period. Both in terms of tolerance- and dependence-inducing properties and adverse effects seen during the chronic treatment in mice, abecarnil did not seem to offer clear advantages compared to the more traditional BDZ receptor ligands. The data substantiate the importance of study design for obtaining predictable informations on the tolerance and withdrawal characteristics of BDZ receptor ligands.