Although fentanyl is a short-acting narcotic analgesic, unexpected respiratory depression has been seen in the postanesthetic period several hours after the last dose. This study has defined the pharmacokinetic characteristics of fentanyl in cerebrospinal fluid (CSF) of the dog and determined their relationships to ventilatory depression. 3H-fentanyl citrate (10 μg/kg) was injected intravenously in dogs anesthetized with enflurane-O2. Arterial plasma and cisternal CSF were analyzed for concentration of unchanged 3H-fentanyl [F] and for total radioactivity (3H). Maximum [F] in CSF occurred 2–10 min after injection and declined at the same rate as [F] in plasma. Values of [F] in CSF averaged approximately 46 per cent of those in plasma, and reflected the binding of fentanyl to plasma proteins. There was a dose-dependent decrease in minute ventilation, and end-tidal CO2 increased. The extent of ventilatory depression correlated closely with the log [F] in arterial plasma and in cisternal CSF (i.e., Δ PETCO2vs. log [F]CSF, r = 0.97, P < 0.01). An early phase in the recovery of ventilation paralleled the initial, rapid elimination of fentanyl from CSF and plasma. Complete recovery was protracted, and the terminal elimination phases of fentanyl from both CSF and plasma were prolonged (t1/2 = 171 ± 8 and 201 ± 25 min, respectively). It is concluded that fentanyl, a highly lipophilic drug, equilibrated rapidly between plasma and CSF, and that there was a close correlation between the concentrations of fentanyl in plasma and CSF and the intensity of respiratory depression. Recovery from the ventilatory effects of fentanyl paralleled the initially rapid elimination of the unchanged drug from CSF and plasma. However, low levels of ventilatory depression and of [F] persisted. Repeated injections of the narcotic analgesic led to the accumulation of fentanyl and increased ventilatory depression.