The performance is tested of 12 bed load sediment transport formulae developed for use in gravel bed channels. The formulae are applied in the manner intended by the original authors. To this end, the test data are restricted to ones obtained in approximately steady flow when the material in motion was similar to that present on the bed; that is, transport was not size selective. This represents the nearest approach to “equilibrium sediment transport” that can be realized with available data. Four sets of river data and three sets of flume data were chosen for the test, covering a range of eight orders of magnitude in unit bed load transport rate. The test data were not used in the development of the formulae. The analysis separates mean bias and local bias. No formula performs consistently well. Limitations of the test data, the constraints imposed by an operationally realistic test, and reasons based upon the physics of the transport phenomenon all may be adduced for this. To estimate the magnitude of transport with limited hydraulic information, stream power equations should be used because they provide the most straightforward scale correlation of the phenomenon. In particular, the approach of Bagnold deserves further study. When local hydraulic information is available, a formula should be selected that is sensitive to bed state or grain size distribution and, in this context, the formulae of Einstein, Parker, and Ackers-White-Day bear continued examination.