The pharmacokinetics of sodium thiopental was studied in 9 normal dogs of various ages by the usual bolus method. Data were fittable by a two compartment open model in 8, three compartments were needed in one. Constant-rate infusions, as used in cerebral resuscitation, were performed in 6 of them. As expected, constant blood levels could not be obtained during the distribution phase which lasted about 6 hours. From the individual pharmacokinetic constants, the time equation of infusion rate intended to exactly compensate for distribution was calculated. The resulting "modulated infusions" succeeded in reducing the blood level trough from 50% to 20%. After the end of infusions, the decrease of plasma levels exhibited a non linear pattern which had not been noticeable in the elimination phase after boluses. Such unknown non linearity may lead to systematic errors when calculating pharmacokinetic parameters: this could explain why a complete correction was not obtained with modulated infusions. A better parametrization method in the case of non-linear elimination is under study.