Immediately after sampling, leukocyte counts in native cerebrospinal fluid (CSF) start to decrease rapidly. As the time lapse between CSF collection to analysis is not routinely registered, the clinical significance of decreasing cell counts in native CSF is not known. Earlier data suggest that addition of serum-containing medium to CSF directly after sampling prevents this rapid decrease in leukocyte counts and, thus, may improve the accuracy of CSF cell counting and cell characterization. Here, we prospectively examined the effect of storage time after lumbar puncture on counts of leukocytes and their major subsets in both native CSF and after immediate addition of serum-containing medium, measured by flow cytometry and microscopy. We collected CSF samples of 69 patients in tubes with and tubes without serum-containing medium and determined counts of leukocytes and subsets at 30 minutes, 1 hour, and 5 hours after sampling. Compared to cell counts at 30 minutes, no significant decrease in cell number was observed in CSF with serum-containing medium 1 and 5 hours after sampling, except for the granulocytes at 1 hour. In native CSF, approximately 50% of leukocytes and all their subsets were lost after 1 hour, both in flow cytometric and microscopic counting. In 6/7 (86%) samples with mild pleocytosis (5–15 × 106 leukocytes/l), native CSF at 1 hour was incorrectly diagnosed as normocellular. In conclusion, addition of serum-containing medium to CSF directly after sampling prevents cell loss and allows longer preservation of CSF cells prior to analysis, both for microscopic and flow cytometric enumeration. We suggest that this protocol results in more accurate CSF cell counts and may prevent incorrect conclusions based on underestimated CSF cell counts.