The Murex Cryptococcus Test was compared with the Cryptococcal Antigen Latex Agglutination System (CALAS) for detecting cryptococcal polysaccharide in 173 cerebrospinal fluid (CSF) specimens and 117 serum samples with 99% and 97% concordance, respectively. Eighteen CSF samples and 17 serum samples were positive in both assays, and 249 were negative. The sensitivity and specificity of the Murex relative to the CALAS were 90% and 100%, respectively, for CSF, and 81% and 100%, respectively, for serum. Six discrepancies were arbitrated by retesting, using a third analytic method, review of other laboratory and clinical data, or both. The reaction in 1 CSF specimen was considered false positive by the CALAS, and the reactions in 2 serum samples were false negatives by the Murex. For 3 patients with previous cryptococcal meningitis but no active disease, only the CALAS detected antigen, suggesting that the Murex has less analytic sensitivity in this context. Titer differences dictate that direct comparisons between the 2 tests are not feasible. There were no false-positive reactions in limited testing with either method using specimens from patients with concurrent noncryptococcal infections or in rheumatoid factor-positive serum samples. Infections caused by Cryptococcus neoformans serotypes A or AD were detected equally by both assays. Based on our study, we have elected to continue to use the CALAS for routine testing for cryptococcal antigen.