We studied the clinical value of Wellcogen, a latex particle agglutination test designed to diagnose rapidly Group B streptococcal infections in 620 infants. The sensitivity and specificity were 18 of 20 (90%) and 18 of 22 (81%), respectively. False negatives were documented in 2 patients and false positive tests occurred in 4. Also 10 of 620 (1.6%) infants had detectable antigen in the absence of proved infection. In addition, in 4 other infants with Group B streptococcal bacteremia, antigenemia was absent in serum obtained on the day of hospitalization but antigen was detected in specimens obtained subsequently. We also assayed the sera of 102 randomly chosen hospitalized adults; 4 (4%) contained "antigen" but none had Group B streptococcal infection. None of 50 urines obtained from other randomly chosen hospitalized adults contained "antigen." In patients with Group B streptococcal meningitis cerebrospinal fluid and unconcentrated urine uniformly contained antigen but in bacteremia without focality, assay of blood and urine were both necessary to diagnose correctly all 7 patients with the Wellcogen test. We conclude that despite acceptable sensitivity and specificity, the interpretation of the Wellcogen test must be performed with caution and knowledge of its limitations.