Received 5 June 2007 Accepted 6 June 2007 We need a coherent, deliberate process to learn from failures and successes in stream rehabilitation projects. Insightful evaluations comparing projects in varied regions have been limited by the diversity of approaches and some scientific uncertainty as to how to best accomplish longer-term monitoring (Bernhardt et al., 2005; Giller, 2005; Palmer et al., 2005; Palmer and Bernhardt, 2005; Reid, 2001, Reid and Furniss, 2002). Addressing just the critical issue of biological assessments would take an entire paper but with hundreds of natural channel design (NCD) projects completed across the country, simply evaluating their performance and impacts on bank erosion would be a step forward. Too often critiques of NCD projects have returned continually to the same few locations, and have yet to grapple with many other projects with different designs, disturbance histories, and environments. I hope this brief review will encourage better dialogue between scientists and stream rehabilitation practitioners. NCD most often seeks to restore the dimension, pattern, and profile of a disturbed river system to emulate the natural stable river (Rosgen, 2006). A stable channel is defined as a dynamic, alluvial channel whose characteristic dimensions or features do not change over engineering time scales (Niezgoda and Johnson, 2005). Stream bank erosion is a natural process, but when accelerated by human impacts creates a disequilibrium condition, although in some cases a braided river and/or anastomozing river type is the stable form (Rosgen, 2006; Jaquette et al., 2005). With a surge in funding for stream restoration in the United States, prompted by the decline of Pacific coast salmon runs and water quality problems across the country, public and private groups have spent more than $14 billion on 37 000 stream restoration projects since 1990 (Bernhardt et al., 2005). The Chesapeake Bay watershed alone had 747 projects to reconfigure channels and reduce bank erosion (Hassett et al., 2005), most completed since 1995. In North Carolina there have been over 400 NCD projects (Miller et al., 2006). Some of the most successful stream restoration in the western United States has been the simplest and least expensive, including riparian planting and controlling livestock use of riparian areas, thus allowing bank vegetation to recover which alone can sometimes much improve channel conditions (Nagle and Clifton, 2003). Other projects using NCD are quite controversial (Malakoff, 2004), with costs from $165/m in small rangeland streams to $2300/m in urban areas. Although NCD is expensive, the more common approaches of armoring banks with rock rip-rap can cost $325/m or more in medium-sized streams while berming and channelization can cost $7400/m (Lovegreen and Petlock, 2006), with both resulting in major problems with largely negative impacts on aquatic habitat. Dave Rosgen, a former regional hydrologist for the US Forest Service Rocky Mountain Region, devised the NCD approach and is its most influential proponent. Drawing on research by Luna Leopold and many others, Rosgen distilled decades of his own field observations into