Due to its robustness and versatility, several variations of the blood–brain barrier specific parallel artificial membrane permeability assay (PAMPA-BBB) have been reported in the central nervous system (CNS) drug discovery practice. In this study, the impact of the main assay parameters on the predictive power of PAMPA-BBB was thoroughly investigated with 27, passively BBB-transported drug molecules with in vivo logBB data. The single and combined effects of the following variables were systematically studied and optimized: incubation time and temperature (4 vs. 18h, RT vs. 37°C), type of the read-out (UV-reader vs. HPLC), solvent composition (n-dodecane/n-hexane), lipid concentration (0–10w/v % PBLE), cholesterol content (0–1.66w/v %), and thickness of the lipid membrane, and the DMSO cosolvent content (5–20v/v %), respectively. Based on our results, solvent-driven and lipid-driven mechanisms of diffusion were identified in different assay conditions. Moreover, the analysis of membrane retention (MR%; the mole fraction of solute “lost” to the membrane) data obtained at various membrane compositions (volume of solvent and concentration of phospholipids) revealed the compound-specific nature of this parameter. The optimized conditions for the PAMPA-BBB were the following: 4h incubation at 37°C, detection by HPLC-DAD, iso-pH conditions (pH=7.4) with 5v/v % DMSO content in buffer solutions, and PBLE (10w/v %; without cholesterol) as membrane dissolved in the mixture of n-hexane:n-dodecane 3:1.