A bivalent vaccine consisting of Salmonella typhi strain Ty21a containing the 120 MDa plasmid of Shigella sonnei and expressing both S. typhi and S. sonnei lipopolysaccharides (LPS) on its surface was previously shown to protect significantly against S. sonnei disease in experimental challenge studies. However, protective efficacy could not be reconfirmed in volunteers with five subsequent lots of vaccine. One vaccine lot which resembled the initial protective lots of vaccine in biochemical and serological tests, and by electron microscopy, was administered to 16 volunteers who ingested three doses of 10(9) organisms each. Antibody secreting cells (ASC) specific for S. sonnei LPS were detected in the blood of 100% of vaccines, but no protection of these vaccines was demonstrated during a S. sonnei challenge study. To assess the ability of the volunteer model to detect infection-derived immunity, six volunteers who had had clinical shigellosis due to S. sonnei two months earlier were rechallenged with wild-type S. sonnei, together with 12 controls. Prior infection provided 100% protection against febrile illness (p = 0.05) and diarrhea (p = 0.04), thereby validating the volunteer model for assessing Shigella vaccines.