AbstractThe rapid decline of the Snake River spring–summer‐run Chinook salmon Oncorhynchus tshawytscha evolutionarily significant unit (ESU) in the 1990s led a group of scientists to develop the Plan for Analyzing and Testing Hypotheses (PATH). Under this plan, researchers used spawner–recruit (SR) data to estimate the survival of out‐migrating smolts through eight dams of the Federal Columbia River Power System (FCRPS). Direct measurements of survival during out‐migration, known as passage survival, were not available, so the PATH scientists estimated survival for index populations using trends from Ricker‐type SR models. This modeling framework had the advantage of estimating both the direct and indirect (or latent) effects (FCRPS‐related mortalities that do not occur until the smolts have passed the FCRPS dams) on the life cycle survival of the populations. We evaluated the SR model used by the PATH scientists by examining how changes in model structure affected important inferences. We calculated condition indexes as measures of the sensitivity of the model results to perturbations in the SR data and model structure, finding that the results were highly sensitive to certain assumptions. In particular, we found that changing the Ricker α term from a population‐specific parameter to a parameter common to all of the populations in the ESU changed total passage survival from 9% to 56% and latent mortality from one‐half the total passage mortality to a value that is not significantly different from zero. Therefore, the condition indexes revealed high potential sensitivities of the SR model results to perturbations in data and model structure. Although information criteria indicated that the population‐specific model had a poorer fit than lower‐parameter models, it was impossible to resolve the question as to whether there was latent mortality.