The daily oral administration of chlordiazepoxide (40 mg/kg) over 9 weeks in rats elicited full tolerance to muscle relaxant effects within 7 weeks, as revealed by twice weekly evaluations of abdominal tone myorelaxation and decreased grip strength. No full tolerance was achieved, however, during the 9 weeks of treatment in terms of ataxia. Electroencephalographic (EEG) studies showed that this tolerance to the behavioural effects was accompanied by a progressive decrease in mean power spectra, associated with a progressive decrease in the beta band, but in this case, full tolerance was reached within 4 weeks. Once weekly evaluations of the ability of chlordiazepoxide to protect the animals against pentylenetetrazole seizures revealed a similar pattern. Treatment with flumazenil (50 mg/kg p.o.) 24 h after the last chlordiazepoxide administration induced a clear withdrawal syndrome associated with EEG changes which consisted of an increase in total power spectra associated with an increase in the delta band (in comparison with chlordiazepoxide-dependent rats not given the antagonist). These findings suggest that the different kinetics of the tolerance to anticonvulsant and EEG effects in comparison to myorelaxant effects can be attributed to a different involvement of benzodiazepine receptor subtypes.