To evaluate the relative merits and deficiencies of Millipore filter and cytocentrifuge preparations in the detection of central nervous system (CNS) acute leukemia in pediatric patients, 300 cerebrospinal fluid (CSF) specimens from 17 patients were prepared by both methods. The 17 patients studied were all diagnosed and treated for CNS leukemia. Leukemic blast cells were found by at least one method in 91 CSF specimens, and the results of both techniques were positive in 77 (85%) of 91 specimens. Of the 14 specimens in which the results of only one method were positive, seven yielded positive results only by the cytocentrifuge method, and seven yielded positive results only by the Millipore filter method. In 12 of the 14 discrepant specimens, the paired specimen whose results were not interpreted as positive was technically unsatisfactory (nine specimens) or had cells suspicious for blast cells rather than unequivocal blast cells (three specimens). Blast cells were identified in specimens with low nucleated cell counts (less than or equal to 5/mm3) by both methods and usually were immediately preceded or followed by CSF specimens showing florid disease. We conclude that performance of both methods is unnecessary for routine surveillance if processing techniques yield quality preparations. Cytocentrifuge preparations stained by Wright's method allow better morphologic correlation with bone marrow blast cells and allow easier identification of blood or bone marrow contamination.