Question: To what extent does trait-dependent emigration bias selection estimates in a natural system? Organisms: Two freshwater cohorts of Atlantic salmon (Salmo salar) juveniles. Field site: A 1 km stretch of a small stream (West Brook) in western Massachusetts. USA from which emigration could be detected continuously. Methods: Estimated viability selection differentials for body size either including or ignoring emigration (include = emigrants survived interval, ignore = emigrants did not survive interval) for 12 intervals. Results: Seasonally variable size-related emigration from our study site generated variable levels of bias in selection estimates for body size. The magnitude of this bias was closely related with the extent of size-dependent emigration during each interval. Including or ignoring the effects of emigration changed the significance of selection estimates in 5 of the 12 intervals, and changed the estimated direction of selection in 4 of the 12 intervals. These results indicate the extent to which inferences about selection in a natural system can be biased by failing to account for trait-dependent emigration. ?? 2005 Benjamin H. Letcher.