We evaluated the survival of juvenile salmon through turbines in Columbia River dams and found no differences between two operations but strong evidence of delayed mortality from turbine passage. After tagging with a passive integrated transponder (PIT) tag and a radio tag, yearling Chinook salmon Oncorhynchus tshawytscha were released at McNary Dam on the Columbia River through a turbine operating both within 1% of peak efficiency (a discharge rate of 317 m(3)/s) and outside the 1% range at the maximum blade angle (464 m(3)/s). Estimated relative survival to a detection array 15 km downstream was 0.871 at 317 m(3)/s and 0.856 at 464 m(3)/s and 0.858 and 0.814, respectively, to an array 46 km downstream. The highest point estimates of survival occurred under the lower discharge, suggesting that operating turbines within 1% of peak efficiency is a useful guideline for fish protection at McNary Dam. In a concurrent evaluation using balloon tags, estimated mean direct survival ranged from 0.930 to 0.946. Radio tag estimates were significantly lower than balloon tag estimates under both operations. Based on these differences, we estimated that delayed mortality comprised from 46% to 70% of total estimated mortality. We reviewed the literature and concluded that delayed mortality was caused by sublethal impacts to fish sensory systems, which increased vulnerability to predation in the tailrace. We recommend that future research to improve turbine designs and operations for fish passage focus on this major component of mortality