From 1987 to 1992, we evaluated a fish bypass system at Bonneville Dam Powerhouse 2 on the Columbia River. The survival of subyearling Chinook salmon Oncorhynchus tshawytscha released into the system ranged from 0.774 to 0.911 and was significantly lower than the survival of test fish released into turbines and the area immediately below the powerhouse where bypass system flow reentered the river. Yearling and subyearling Chinook salmon and yearling coho salmon O. kisutch released into the bypass system were injured or descaled. Also, levels of blood plasma cortisol and lactate were significantly higher in yearling and subyearling Chinook salmon that passed through the bypass system than in fish released directly into a net located over the bypass exit. This original system was then extensively modified using updated design criteria, and the site where juvenile fish reentered the river was relocated 2.8 km further downstream to reduce predation on bypassed fish by northern pikeminnow Ptychocheilus oregonensis. Based on studies conducted from 1999 to 2001, the new bypass system resulted in high fish survival, virtually no injuries to fish, fish passage times that were generally similar to water travel times, and mild stress responses from which fish recovered quickly. The mean estimated survival of subyearling Chinook salmon passing through the new bypass system was 0.946 in 2001, which was an usually low-flow year. Survival, physical condition, passage timing, and blood physiological indicators of stress were all useful metrics for assessing the performance of both bypass systems and are discussed. The engineering and hydraulic criteria used to design the new bypass system that resulted in improved fish passage conditions are described.