Despite knowledge of the bleeding hazard to thrombocytopenic cancer patients undergoing lumbar puncture (LP), a retrospective analysis of physician behavior at one hospital revealed no consistent use of platelet transfusions in patients with less than 20,000 platelets/mm3 on the day of LP. A review of the literature and laboratory cerebrospinal fluid (CSF) data in two institutions, and the performance of an LP experiment revealed that: (1) Batson's epidural venous plexus is an unlikely source and spinal radicular vessels are the most probable source of needle-induced blood in lumbar puncture; (2) the frequency of encountering needle-induced blood at LP is high, 73% (3) the frequency of brushing a nerve root, with the associated risk of lacerating the radicular artery or vein on its surface with the bevel of the LP needle, is high and may be on the order of 26%; and (4) while the passage of an LP needle, obturator in place, through a blood filled vein may carry red cells into a red cell-free medium, this does not always occur. These new considerations argue for more consistent adherence to the already published recommendation of platelet transfusion immediately prior to LP in patients with low platelets. This issue is of particular relevance to the rapidly growing population of thrombocytopenic cancer patients with extended survival on multiple chemotherapeutic regimens requiring lumbar puncture.