BACKGROUND. Detection of malignant cells on cytologic examination of the cerebrospinal fluid (CSF) is the diagnostic gold standard for leptomeningeal carcinomatosis. The absence of cells is a primary endpoint for most therapeutic trials. Unfortunately, false-negative results are common. Practical strategies are necessary to remedy this problem. METHODS. Four physician-dependent variables (CSF sample volume, site of CSF sampling, processing time, and frequency of CSF sampling) were identified, and their contributions to the false-negative rate of CSF cytology were evaluated prospectively in 39 patients with leptomeningeal carcinomatosis. Retrospective data were analyzed to estimate the importance of these variables in daily practice. RESULTS. False-negative CSF cytology results correlated with small CSF volume (P < 0.001), delayed processing (P < 0.001), not obtaining CSF from a site of symptomatic or radiographically demonstrated disease (P = 0.02), and sampling fewer than two times (P < 0.001). In 1 year, 97% of CSF specimens at the study institution were of inadequate volume; >25% were processed too slowly. CONCLUSIONS. False-negative CSF cytology results are common, but can be minimized by: 1) withdrawing at least 10.5 mL of CSF for cytologic analysis; 2) processing the CSF specimen immediately; 3) obtaining CSF from a site of known leptomeningeal disease; and 4) repeating this procedure once if the initial cytology is negative.