From 1986 to 1995, over US$2.5 million has been spent or allocated for projects to modify channel conditions to improve spawning habitat for Chinook salmonOncorhynchus tshawytschain the Merced, Tuolumne, and Stanislaus rivers, tributaries to the San Joaquin River, California. We evaluated the planning, design and performance of the Riffle 1 B reconstruction on the Merced River. This is typical of the nine individual riffle reconstructions completed to date, involving excavation of the existing channel bed (here, to 0.6 m) and back-filling with smaller gravels believed to be more suitable for salmon spawning. We reviewed project documents, interviewed agency staff, and conducted field surveys to document channel conditions in 1994 for comparison with the project as constructed in 1990. The project planning and design did not consider the site's geomorphic context nor processes of erosion and sediment transport under the current flow regime. As a consequence, spawning-sized gravel placed in the channel was scoured and transported through the site at a flow with a return period of 1.5 years. The need for spawning habitat enhancement in the Merced River is questionable, but if such projects are to be built, we recommend that the project planning and design consider the site's geomorphic context and acknowledge the need for and provide funds for project maintenance, and that the performance of completed projects be systematically monitored and evaluated.