We compared several methods of estimating chick survival in the endangered Roseate Tern (Sterna dougallii) at two contrasting colonies. At Bird Island, Massachusetts, no chicks were taken by predators, but many dispersed into dense cover and could not be found. Chicks could not be fenced at this colony without danger of causing mortality. In these circumstances, the best estimates of chick survival were obtained by a detailed study that yield individual estimates of survival probability, taking into account brood order and mass at last encounter. The number of chicks known to have reached the age of 5 days provided a reasonable estimate of the number that survived to fledging. At Cedar Beach, New York, chicks could be fenced into enclosures, but some escaped and others were taken by predators. In these circumstances, the best estimates of chick survival were obtained using experienced judgment, but these estimates were subject to errors that may have ranged up to 30%. Selection of sample plots for study within the large colony at Bird Island may have led to sampling errors of up to 10% in estimates of productivity. Similar issues are likely to arise in estimating chick survival and productivity in other colonial species, in curcumstances where some chicks disperse or otherwise disappear before reaching fledging age. We recommend that future studies of productivity in colonial waterbirds should include measurement of spatial and temporal variability, and should attempt to estimate the magnitude of sampling errors.