Contingent tolerance to the anticonvulsant effects of carbamazepine on amygdala kindled seizures develops when the drug is repeatedly given prior to but not after the electrical stimulation. Such tolerance can be reversed by kindling the rats for several days without drug or even by continuing to give the drug but after each seizure has occurred. Contingent tolerance can be slowed by reducing the electrical stimulus intensity and by chronic continuous (as opposed to repeated paired) drug administration. Contingent cross-tolerance has been demonstrated from carbamazepine to PK11195 (a drug active at peripheral-type benzodiazepine receptors) and valproate, but not to clonazepam and diazepam (two drugs active at central-type benzodiazepine receptors) or phenytoin. Endogenous physiological changes occur in conjunction with contingent tolerance, exemplified by the decrease in seizure threshold that returns to normal upon reversal of tolerance. We suggest that contingent tolerance is associated with a loss of seizure-induced adaptations, since many biochemical changes that occur following seizures (or in non-tolerant animals given drug after seizures) are not observed in tolerant animals. These include a loss of seizure-induced up-regulation of GABAA receptors and a loss of increases in mRNA expression for corticotropin-releasing-factor (CRF), thyrotropin-releasing-hormone (TRH), neuropeptide Y (NPY), glucocorticoid receptors and brain-derived neurotrophic factor (BDNF). Thus, several putative seizure-induced anticonvulsant adaptations, such as increases in GABAA receptors and TRH and NPY mRNA fail to occur in tolerant animals. These findings are consistent with the novel observations that, paradoxically, seizures themselves appear to facilitate the anticonvulsant effects of carbamazepine or diazepam on amygdala kindled seizures. That is, animals given a 'vacation' from seizures show a decreased response to these agents, a phenomenon we have called the 'time-off seizure' effect. Thus, seizures are postulated to induce adaptive changes that influence seizure thresholds and potentiate the anticonvulsant effects of exogenously administered drugs such as carbamazepine and diazepam. Taken together, these data suggest that seizures are associated with endogenous adaptations lasting days to weeks and that a selective failure of some of these to occur during contingent drug administration may underlie the development of contingent tolerance. These observations suggest tht endogenous illness-related mechanisms may participate both in the therapeutic responses of some agents and that their failure to occur could relate to loss of drug efficacy via tolerance; these processes may reveal new potential targets for therapeutic intervention.