We studied the mechanisms underlying CO(2)-dependent DC potential shifts, using epicranial, epidural, epicortical, intraventricular, and intraparenchymal (intraneuronal, intraglial, and field) recordings in ketamine-xylazine-anesthetized cats. DC shifts were elicited by changes in artificial ventilation, causing end-tidal CO(2) variations within a 2-5% range. Hypercapnia was consistently associated with negative scalp DC shifts (average shift -284.4 microV/CO(2)%, range -216 to -324 microV/CO(2)%), whereas hypocapnia induced positive scalp DC shifts (average shift 307.8 microV/CO(2)%, range 234 to 342 microV/CO(2)%) in all electrodes referenced versus the nasium bone. The former condition markedly increased intracranial pressure (ICP), whereas the latter only slightly reduced ICP. Breakdown of the blood-brain barrier (BBB) resulted in a positive DC shift and drastically reduced subsequent DC responses to hypo-/hypercapnia. Thiopental and isoflurane also elicited a dose-dependent positive DC shift and, at higher doses, hypo-/hypercapnia responses displayed reverted polarity. As to the possible implication of neurons in the production of DC shifts, no polarity reversal was recorded between scalp, various intracortical layers, and deep brain structures. Moreover, the membrane potential of neurons and glia did not show either significant or systematic variations in association with the scalp-recorded CO(2)-dependent DC shifts. Pathological activities of neurons during spike-wave seizures produced DC shifts of significantly smaller amplitude than those generated by hyper-/hypocapnia. DC shifts were still elicited when neuronal circuits were silent during anesthesia-induced burst-suppression patterns. We suggest that potentials generated by the BBB are the major source of epicortical/cranial DC shifts recorded under conditions affecting brain pH and/or cerebral blood flow.