The report of a major triglycosyl phenol phthiocerol (phenolic) glycolipid in some strains of Mycobacterium tuberculosis that resembles the phenolic glycolipid I of Mycobacterium leprae raised the prospects of a specific serodiagnostic tool for human tuberculosis. The terminal diglycosyl unit of the M. tuberculosis product was synthesized and converted to a corresponding neoglycoprotein, the O-(2,3,4-tri-O-methyl-alpha-L-fucopyranosyl)-(1----3)-O-alpha-L- rhamnopyranosyl)-(1----9)-oxynonanoyl-bovine serum albumin, and applied, in ELISA, to sera from individuals with tuberculosis. Although the correlation coefficient between the synthetic product and the native glycolipid was excellent, the seroreactivity rate of tuberculous sera was disappointing; only 24 of 119 sera from tuberculosis patients showed evidence of anti-glycolipid antibodies. In isolates of M. tuberculosis from tuberculosis patients the glycolipid was present in only 1 of 11. A partially deglycosylated version was present in two other isolates; however, most isolates lacked the glycolipid. Accordingly, while the results, unlike those of others, do not portend a future for this form of serodiagnosis in the management of tuberculosis, they offer intriguing hints as to the basis of the variable immunogenicity and pathogenicity of strains of M. tuberculosis.