Few estimates of migration rates or descriptions of behavior or survival exist for wild populations of out-migrating Pacific salmon smolts from natal freshwater rearing areas to the ocean. Using acoustic transmitters and fixed receiver arrays across four years (2010-2013), we tracked the migration of > 1850 wild sockeye salmon (Oncorhynchus nerka) smolts from Chilko Lake, British Columbia, to the coastal Pacific Ocean (> 1000 km distance). Cumulative survival to the ocean ranged 3-10% among years, although this may be slightly underestimated due to technical limitations at the final receiver array. Distinct spatial patterns in both behavior and survival were observed through all years. In small, clear, upper-river reaches, downstream migration largely occurred at night at speeds up to 50 km/d and coincided with poor survival. Among years, only 57-78% of smolts survived the first 80 km. Parallel laboratory experiments revealed excellent short-term survival and unhindered swimming performance of dummy-tagged smolts, suggesting that predators rather than tagging effects were responsible for the initial high mortality of acoustic-tagged smolts. Migration speeds increased in the Fraser River mainstem (~220 km/d in some years), diel movement patterns ceased, and smolt survival generally exceeded 90% in this segment. Marine movement rates and survival were variable across years, with among-year segment-specific survival being the most variable and lowest (19-61%) during the final (and longest, 240 km) marine migration segment. Osmoregulatory preparedness was not expected to influence marine survival, as smolts could maintain normal levels of plasma chloride when experimentally exposed to saltwater (30 ppt) immediately upon commencing their migration from Chilko Lake. Transportation of smolts downstream generally increased survival to the farthest marine array. The act of tagging may have affected smolts in the marine environment in some years as dummy-tagged fish had poorer survival than control fish when transitioned to saltwater in laboratory-based experiments. Current fisheries models for forecasting the number of adult sockeye returning to spawn have been inaccurate in recent years and generally do not incorporate juvenile or smolt survival information. Our results highlight significant potential for early migration conditions to influence adult recruitment.