Trials of pneumococcal vaccine in healthy young adult populations suggest 75 to 95 per cent type-specific efficacy. Trials have not been done, however, in groups for which pneumococcal vaccine is currently recommended in the United States. To assess efficacy in the immunocompromised groups now receiving the vaccine, we compared serotypes of 35 isolates of Streptococcus pneumoniae isolated from blood or cerebrospinal fluid one month or longer after the patient had received commercially available pneumococcal vaccine with serotypes of 392 isolates from unvaccinated persons surveyed in a study of the nationwide distribution of pneumococcal serotypes. Proportions of infections with vaccine serotypes were unexpectedly similar. An alternative approach to estimation of vaccine efficacy suggested an average efficacy of 36 per cent in persons of all ages. The estimated efficacy was lowest in children two to 10 years old (less than or equal to 0 per cent) and in persons with preexisting diseases that are thought to predispose to pneumococcal disease (less than or equal to 0 per cent), and highest in the group over 10 years old (60 per cent). The possibility that efficacy is low in high-risk, immunocompromised populations makes it important to evaluate clinical efficacy and not just serum antibody responses in such groups.