We estimated apparent annual survival probability of adult Purple Martins (Progne subis) using a 14-year data set from a population in northwestern Pennsylvania. We modeled age, sex, and year effects on adult survival using known-age birds (585 males and 392 females) first banded or resighted as one-year-olds in two core breeding colonies. We resighted banded birds at (1) the two core breeding colonies, (2) other breeding colonies within 50 km of the core colonies, and (3) a large premigratory roost that attracts adult Purple Martins from colonies ≤200 km away. Apparent annual survival probability for one-year-olds increased by 0.08 when we included encounters outside the core colonies. Survival probability for older males and females was largely unaffected, a result of their high breeding-site fidelity. The model with age-specific survival (1 year old, 2–4 years old, ≥5 years old) and an age*sex interaction had the strongest support when all encounters were included. Apparent annual survival of one-year-old females (0.48 ± 0.03 [SE]) was lower than that of one-year-old males (0.59 ± 0.02), but we detected no sex differences in survival among older birds (2–4 years old: females, 0.64 ± 0.02; males, 0.62 ± 0.02). We found evidence of senescence, because birds at least five years old had lower apparent survival (females: 0.51 ± 0.06; males: 0.52 ± 0.05). Fully time-dependent models had low support, including those evaluating the relationship between annual variation in prevalence of West Nile virus and the El Niño Southern Oscillation Index, a key climate variable.