We applied Cormack-Jolly-Seber open population models to manatee (Trichechus manatus latirostris) photo-identification databases to estimate adult survival probabilities. The computer programs JOLLY and RECAPCO were used to estimate survival of 677 individuals in three study areas: Crystal River (winters 1977-78 to 1990-91), Blue Spring (winters 1977-78 to 1990-91), and the Atlantic Coast (winters 1984-85 to 1990-91). We also estimated annual survival from observations of 111 manatees tagged for studies with radiotelemetry. Survival estimated from observations with telemetry had broader confidence intervals than survival estimated with the Cormack-Jolly-Seber models. Annual probabilities of capture based on photo-identification records were generally high. The mean annual adult survival estimated from sighting-resighting records was 0.959-0.962 in the Crystal River and 0.936-0.948 at Blue Spring and may be high enough to permit population growth, given the values of other life-history parameters. On the Atlantic Coast, the estimated annual adult survival (range of means = 0.877-0.885) may signify a declining population. However, for several reasons, interpretation of data from the latter study group should be tempered with caution. Adult survivorship seems to be constant with age in all three study groups. No strong differences were apparent between adult survival ofmales and females in the Crystal River or at Blue Spring; the basis of significant differences between sexes on the Atlantic Coast is unclear. Future research into estimating survival with photo-identification and the Cormack-Jolly-Seber models should be vigorously pursued. Estimates of annual survival can provide an additional indication of Florida manatee population status with a stronger statistical basis than aerial counts and carcass totals.