The diagnosis of leptospirosis is often made using the microscopic agglutination test (MAT), in which live antigens representing >20 serogroups undergo reaction with patient serum samples to detect agglutinating antibodies. Data derived from this assay are often used to infer the identity of the infecting leptospiral serovar or serogroup; however, paradoxical reactions and cross-reactions between serogroups are common. To evaluate the usefulness of this approach, data on culture-proven cases of leptospirosis that occurred in Barbados from January 1980 through December 1998 were reviewed. A total of 151 isolates of 4 serovars were identified. The sensitivity of MAT for the prediction of the infecting serovar was determined. Overall, the predominant serogroup at a titer of >or=100 correctly predicted 46.4% of all serovars isolated. If a titer of >or=800 was used as the cutoff, sensitivity decreased slightly to 44.4%. The overall specificity for all serogroups was 64.8%. Serologic analysis appeared to be of little value for the identification of the infecting serovar in individual cases of leptospirosis in humans. Presumptive serogroup reactivity data should be used only to gain a broad idea of the serogroups present at the population level.