Tundra swans (Cygnus columbianus ) of the eastern population were neckbanded in Maryland, North Carolina, and Alaska from 1966 through 1990. These swans were resighted and recaptured during autumn, winter, and spring, 1966-1990. Although the original motivation for this study involved swan movements, we wanted to use the resulting data to test hypotheses about sources of variation in swan survival rates. Recaptures of legbanded and neckbanded swans permitted us to estimate neckband loss rates, which were found to vary with age and sex of swans, and number of years since initial application. Estimates of annual neckband retention rate ranged from about 0.50 for adult male swans greater than or equal to 2 years after initial neckbanding to > 0.96 for immature swans and adult females the first year following neckbanding. This variation in neckband loss rates prevented the simple correction of survival estimates to account for such loss. Consequently, we developed a series of multinomial models parameterized with survival, sighting, and neckband retention probabilities for use with the recapture and resighting data.