Despite nearly 200 years of study, the mechanisms contributing to the maintenance of endemic cholera and the causes of periodic epidemics remain poorly understood. To investigate these patterns, cholera data collected over 33 years (1966-1998) in Matlab, Bangladesh, were analyzed. Time-lagged autocorrelations were stratified by Vibrio cholerae serogroup, serotype, and biotype. Both classical and El Tor biotypes alternated and persisted between 1966 and 1988; the classical biotype disappeared by 1988, and the O139 serogroup first appeared in 1993. Both the Ogawa and Inaba serotypes circulated the entire time. The autocorrelations revealed that both Inaba and Ogawa epidemics were followed 12 months later by epidemics of the same serotype. Ogawa epidemics, however, were also followed by further Ogawa epidemics only 6 months later. Thus, epidemics of Inaba may selectively confer short-term population-level immunity for a longer period than those of Ogawa. These observations suggest that the Inaba antigen should be maximized in cholera vaccine designs.