The durability of several common bank protection methods used as less expensive or environmentally acceptable alternatives to toe-trenched angular rock riprap has been documented during the 1980s at seven typical bank erosion sites in British Columbia. Since the sites were uncontrolled examples of actual river training works in operation, natural events hindered a precise comparison of each scheme with toe-trenched angular rock riprap protection. However, of the four alternatives investigated – gravel dykes, tree revetments, riprap with toe apron, and semi-round riprap – all but semi-round riprap performed less than satisfactorily over the years of observation. Documentation of performance is sufficient to conclude that on steep gravel-bed rivers, gravel dykes do not endure; tree revetments require constant maintenance and will not endure large floods, unless extremely well-constructed; toe aprons are not as reliable as toe trenches for the same volume of rock; and well-placed, large semi-round rock performs well under moderately severe attack. Key words: erosion, bank protection, channel stability, river training structures, gravel-bed.