Gravel-bed rivers organize to a threshold state in which the fluid stress during dominant flow conditions slightly exceeds the critical value. These channels change fairly abruptly to sand-dominated beds, facilitated by a paucity of sediment in the range of 1–10 mm. Eolian sediments are also bimodal; sand seas often transition rapidly downwind to silty loess deposits. In this paper we hypothesize that both transitions result from the same processes: (1) production of fine sediment by spallation from saltation abrasion of coarse grains, and (2) segregation of fine and coarse modes by suspension and saltation transport, respectively. We demonstrate that gravel rivers and eolian dune fields are dynamically equivalent in terms of threshold transport (Shields stress) and collision dynamics (Stokes number). While abrasion may not be the dominant cause of downstream fining of fluvial gravels, the reduction of collision efficiency with grain diameter sets a lower limit on gravel size, leading to a grain-size gap. The collision efficiency hypothesis potentially reconciles conflicting studies on the importance of sorting and lithology in downstream fining. Shields and Stokes similitude in rivers and deserts may be used to design appropriately scaled experiments to test and refine our hypothesis.