We compare the results obtained by counterimmunoelectrophoresis in samples of serum and cerebrospinal fluid with microbiologic methods for 3,298 patients suspected of bacterial meningitis and/or septicemia at Instituto Adolfo Lutz, São Paulo, in a retrospective study of the period from July 1998 to July 1994. Of the 415 patients (12.6% of the total cases studied) who were positive by the serum test, only 249 (7.6% of the total cases studied) were also positive when cerebrospinal fluid was assayed. Thus, 40% of the positives (5.6% of the total) were identifiable by analysis of serum but not of cerebrospinal fluid. Neisseria meningitidis accounted for 77.7% (129) and Haemophilus influenzae for 22.3% (37) of the positive results obtained only when serum was examined. These data show that although sensitivity and specificity of serum counterimmunoelectrophoresis are relatively low compared to cerebrospinal fluid counterimmunoelectrophoresis, the serum test is necessary to complement cerebrospinal fluid counterimmunoelectrophoresis data.