The objective of the present study was to examine the accuracy of using unbound brain concentration determined by a brain homogenate method (C(ub)), cerebral spinal fluid concentration (C(CSF)), and unbound plasma concentration (C(up)) as a surrogate for brain interstitial fluid concentration determined by brain microdialysis (C(m)). Nine compounds-carbamazepine, citalopram, ganciclovir, metoclopramide, N-desmethylclozapine, quinidine, risperidone, 9-hydroxyrisperidone, and thiopental-were selected, and each was administered as an intravenous bolus (up to 5 mg/kg) followed by a constant intravenous infusion (1-9 mg/kg/h) for 6 h in rats. For eight of the nine compounds, the C(ub)s were within 3-fold of their C(m); thiopental had a C(m) 4-fold of its C(ub). The C(CSF)s of eight of the nine compounds were within 3-fold of their corresponding C(m); 9-hydroxyrisperidone showed a C(CSF) 5-fold of its C(m). The C(up)s of five of the nine compounds were within 3-fold of their C(m); four compounds (ganciclovir, metoclopramide, quinidine, and 9-hydroxyrisperidone) had C(up)s 6- to 14-fold of their C(m). In conclusion, the C(ub) and C(CSF) were within 3-fold of the C(m) for the majority of the compounds tested. The C(up)s were within 3-fold of C(m) for lipophilic non-P-glycoprotein (-P-gp) substrates and greater than 3-fold of C(m) for hydrophilic or P-gp substrates. The present study indicates that the brain homogenate and cerebral spinal fluid methods may be used as surrogate methods to predict brain interstitial fluid concentrations within 3-fold of error in drug discovery and development settings.