Averages of measurements made in three rivers characterized by a high availability of sediment in relation to runoff, and comparable data that represent the high end of the range of transport rates observed during three sets of laboratory experiments, confirm there is an upper, particle-size-dependent limit to bed-load transport efficiency. Incorporating the regression relation derived from these six diverse and unrelated data sets into R. A. Bagnold's classic formulation yields ib ؍ [0.0115⅐D50 ؊0.51 ]͞0.63. This straightforward scale correlation can be used to estimate the potential rate of bed-load transport (for average conditions) in gravel-bed rivers when sediment transport is constrained neither by the supply of sediment to nor by the amount of sediment available in the channel. The independent application of this empirical limit formula to two rivers with applicable bed-load transport regimes reveals a good (؎10%) correspondence between average observed and predicted transport rates.bed-load transport efficiency ͉ gravel-bed rivers ͉ sediment transport B ed-load transport provides the major process linkage between the hydraulic and material conditions that govern river-channel morphology, and knowledge of bed-load movement is required not only to elucidate the causes and consequences of changes in fluvial form but also to make informed management decisions that affect a river's function. Unfortunately the collection of high-quality bed-load transport data is an expensive and time-consuming task, and for many practical purposes recourse is made to a bed-load transport formula (1). The ability of any formula to predict the bed-load transport rate under given flow conditions is predicated on the assumption that it is possible to describe the rate at which bed load is transported in terms of measurable hydraulic and sedimentological quantities (1). Even assuming there are no limitations on sediment supply, the task is complicated by the realization that, at all but the highest flows, the movement of heterogeneous sediment is governed by absolute and relative size effects, so that the local transport rate depends on the population of particles immediately available at the bed surface (2). However, little is known about how the composition of the bed surface changes over time, and despite more than a century of effort it is not yet possible to make reliable predictions of bed-load transport rates. One way to address this impasse is to remove the restrictions imposed by sediment supply and availability from the problem and determine how much bed load a river is capable of transporting, rather than how much it actually transports. That is, to articulate a limit formula for potential transport. I accomplish this by incorporating an empirical, particle-size-dependent term for bed-load transport efficiency into an existing formulation that has been shown to be relatively successful in predicting bed-load transport in gravel-bed rivers (1, 2), where the dominant bed material size ranges from 0.002 to ϳ0.2 m.G. K. ...