Plasma and salivary caffeine concentrations were measured by gas-liquid chromatography in 6 healthy caffeine-free volunteers following oral administration of 50, 300, 500 and 750 mg caffeine. Caffeine was also given to a single subject intravenously in doses of 300, 500 and 750 mg. Caffeine was rapidly absorbed and was completely available at all doses. The apparent first-order elimination rate constant decreased linearly with dose and was 0.163 +/- 0.081 h-1 for 50 mg and 0.098 +/- 0.027 h-1 for 750 mg. The total body clearance was unaffected by dose and was 0.98 +/- 0.38 ml/min/kg. There was a trend towards increasing apparent volume of distribution with increasing dose. A linear relationship existed between the area under the plasma concentration, time curve and dose and dose-normalised plasma concentration, time plots were superimposable. These findings suggest that caffeine obeys linear pharmacokinetics over the dose range investigated. Despite significant inter-individual differences in pharmacokinetic parameters there was good reproducibility within 5 subjects given 300 mg caffeine orally on 3 occasions. Salivary caffeine levels probably reflect the unbound plasma caffeine concentration and can be used to estimate the pharmacokinetic parameters of the drug. Overall the saliva/plasma concentration ratio was 0.74 +/- 0.08 but within subjects some time-dependence of the ratio was found with higher ratios initially (even after intravenous administration) and lower ratios at longer time intervals after the dose. Urinary elimination of caffeine was low and independent of dose: 1.83% of the dose was eliminated unchanged.