Intraventricular administration of chemotherapy is one approach to overcoming the limited distribution of anticancer drugs and their active metabolites into the CNS. This form of regional chemotherapy has led to effective treatment of occult and overt meningeal leukaemia in humans. In contrast, the efficacy of this therapy is extremely limited in the treatment of leptomeningeal dissemination of various solid tumours. Pharmacokinetic studies of the commonly intraventricularly applied anticancer agents in humans have demonstrated that, using low drug doses, very high drug concentrations can be achieved in the cerebrospinal fluid (CSF) and relatively high concentrations in the leptomeninges but not in the brain tissue and the plasma. Therefore, this approach is not an effective treatment for bulky disease of brain tissue, and results in minimal systemic toxicity. In comparison with intralumbar administration, lower interpatient variability of CSF drug concentrations and improved clinical efficacy were observed. 'Concentration x time' schedules, i.e. frequent small drug doses over a short period, enable long-term CSF exposure to cytotoxic drug concentrations while avoiding excessively high and potentially neurotoxic drug concentrations. The technique of ventriculolumbar cerebrospinal perfusion delivers continuously high drug concentrations throughout the CSF for several hours, but its widespread use is limited by the technical complexities of this approach. In this article, the dosages, schedules and pharmacokinetic data of routinely used intraventricular agents in humans, e.g. methotrexate, cytarabine, glucocorticoids and thiotepa, are outlined in detail. In addition, pharmacokinetic data of investigational agents for intraventricular administration (diaziquone, DTC 101, mercaptopurine, mafosfamide, etoposide, topotecan, nimustine [ACNU] and bleomycin) are presented. Better understanding of the CSF pharmacology of these drugs is an essential prerequisite for safe, effective administration of these drugs. Investigational efforts are underway to verify the feasibility and efficacy of different dosages, schedules and combination therapies of these new intra-CSF agents. Current and future clinical research should also focus on methods allowing the delivery of tumoricidal drug concentrations for extended periods into the CSF and the brain tissue while minimising neurotoxicity and systemic toxicity (e.g. liposomal drug preparations, monoclonal antibodies, immunotoxins and gene therapy).