I estimated fall survival and determined the magnitude and sources of mortality for American woodcock (Scolopax minor) using radio telemetry in Michigan, Minnesota, and Wisconsin during 2001-2004. In all 3 states woodcock were radio-marked on paired study areas; one of which was open to hunting and the other of which was either closed to hunting (Michigan and Minnesota) or was relatively inaccessible to hunters (Wisconsin). I used program MARK to estimate fall survival, and to construct a set of candidate models to examine the effects of hunting and several effects of covariates (sex, age, year, state, size) on survival. Fall (10 September-8 November) survival estimates based on data pooled among years and states were 0.784 (95% CI 0.746-0.817) in the hunted areas and 0.881 (95% CI 0.824-0.921) in the non-hunted areas. Hunting accounted for 48% of the 147 woodcock deaths in the hunted areas, followed by predation (32%) and various other sources of mortality (20%). The 66 woodcock deaths that occurred in the non-hunted and lightly-hunted areas were caused by predators (58%), various other sources (24%) and hunting (18%). Akaike's Information Criterion model selection indicated that fall survival varied by treatment (i.e., hunted versus non-hunted) and year. The estimate of the treatment effect was 11.6% (95% CI 0.045-0.187). Survival estimates did not vary by age, sex, bill length, or weight. The harvest rate when data were pooled among states and years was 13.0%, and the kill rate due to hunting was 14.5%.