Central nervous system (CNS) infections caused by bacteria with reduced sensitivity to antibacterials are an increasing worldwide challenge. In successfully treating these infections the following conditions should be considered: (i) Antibacterials do not distribute homogeneously in the central nervous compartments [cerebrospinal fluid (CSF), extracellular space of the nervous tissue, intracellular space of the neurons, glial cells and leucocytes]. Even within the CSF, after intravenous administration, a ventriculo-lumbar concentration gradient is often observed. (ii) Valid parameters of drug entry into the CSF are the CSF: serum concentration ratio in steady state and the CSF: serum ratio of the area under the concentration-time curves (AUCCSF/AUCS). Frequently, the elimination half-life (t1/2 beta) in CSF is longer than t1/2 beta in serum. (iii) For most antibacterials, lipophilicity, molecular weight and serum protein binding determine the drug entry into the CSF and brain tissue. With an intact blood-CSF and blood-brain barrier, the entry of hydrophilic antibacterials (beta-lactam antibacterials, glycopeptides) into the CNS compartments is poor and increases during meningeal inflammation. More lipophilic compounds [metronidazole, quinolones, rifampicin (rifampin) and chloramphenicol] are less dependent on the function of the blood-CSF and blood-brain barrier. (iv) Determination of the minimal inhibitory concentrations (MIC) of the causative organism is necessary for optimisation of treatment. (v) For rapid sterilisation of CSF, drug concentrations of at least 10 times MIC are required. The minimum CSF concentration: MIC ratio ensuring successful therapy is unknown. Strategies to achieve optimum antibacterial concentrations in the presence of minor disturbances of the blood-CSF and blood-brain barrier include, the increased use of low toxicity antibacterials (e.g., beta-lactam antibiotics), the use of moderately lipophilic compounds, and the combination of intravenous and intraventricular administration. Antibacterials which do not interfere with bacterial cell wall synthesis delay and/or decrease the liberation of proinflammatory bacterial products, delay or inhibit tumour necrosis factor release, and may reduce brain oedema in experimental meningitis. Conclusive evidence of the reduction of neuronal damage by this approach, however, is lacking.