Zolpidem is one of the most frequently prescribed hypnotics, as it is a very short-acting compound with relatively few side effects. Zolpidem's short duration of action is partly related to its short elimination half-life, but the associations between plasma levels and pharmacodynamic (PD) effects are not precisely known. In this study, the concentration-effect relationships for zolpidem were modelled. Zolpidem (10 mg) was administered in a double-blind, randomised, placebo-controlled trial to determine PD and pharmacokinetics (PK) in 14 healthy volunteers. Zolpidem was absorbed and eliminated quickly, with a median T(max) of 0.78 h (range: 0.33-2.50) and t(1/2) of 2.2 h. Zolpidem reduced saccadic peak velocity (SPV), adaptive tracking performance, electroencephalogram (EEG) alpha power and visual analogue scale (VAS) alertness score and increased body sway, EEG beta power and VAS 'feeling high'. Short- and long-term memory was not affected. Central nervous system effects normalised more rapidly than the decrease of plasma concentrations. For most effects, zolpidem's short duration of action could be adequately described by both a sigmoid E(max) model and a transit tolerance model. For SPV and EEG alpha power, the tolerance model seemed less suitable. These PK/PD models have different implications for the mechanism underlying zolpidem's short duration of action. A sigmoid E(max) model (which is based on ligand binding theory) would imply a threshold value for the drug's effective concentrations. A transit tolerance model (in which a hypothetical factor builds up with time that antagonises the effects of the parent compound) is compatible with a rapid reversible desensitisation of GABAergic subunits.