This study tests the hypothesis that if cerebrospinal fluid (CSF) has a nucleated blood cell count (NucBC) of less than 6/mm3, CSF tests other than bacterial culture need not be performed to exclude the diagnosis of bacterial meningitis in patients not receiving antimicrobial agents. The results of tests performed on the first specimen of CSF obtained for a given hospital visit from children younger than 3 years of age, exclusive of newborn infants admitted to the hospital on their date of birth, were analyzed. Of 3356 CSF specimens evaluated, 122 were from patients with bacterial meningitis; 460 specimens were analyzed separately because the erythrocyte count was greater than 1000/mm3. A negative CSF screening test result was defined as a CSF NucBC less than 6/mm3. In facilitating the diagnosis of bacterial meningitis, this screening test had a sensitivity of 98.4%, a specificity of 75.2%, and a negative predictive value of 99.9%. The other CSF tests varied widely in screening effectiveness: a Gram-stained smear had a sensitivity of 53% and a specificity of 97%. Receiver operating characteristic curve analysis was used to assess the screening relevance of CSF tests. The CSF NucBC and CSF segmented NucBC performed indistinguishably and superiorly compared with the CSF protein or glucose concentration and the ratio of CSF glucose to serum glucose concentration. Logistic regression analysis showed that the NucBC alone is superior to any combination of the other CSF tests. In a prospective study of 215 children younger than 3 years of age undergoing a lumbar puncture in our emergency department, 85% had empiric criteria identifying them as appropriate for an abbreviated CSF evaluation. The CSF NucBC was less than 6/mm3 in 70% of the 181 patients who would have been eligible for an abbreviated CSF evaluation. These data suggest that a strategy for the sequential testing of CSF could be adopted that would exclude unnecessary determinations and thereby save time, effort, and health care dollars.