Previous experimental studies on abrasion of gravel showed diminution rates much lower than those typical of gravel-bed rivers. This might be one reason many studies of downstream fining in rivers stress hydraulic sorting as the dominant fining mechanism. Previous experiments, however, did not replicate well the particle-to-particle impact between the bed-load gravels and bed gravels in rivers on alluvial fans during floods. The purpose of this study was to reexamine the effect of abrasion on downstream fining to determine if abrasion is responsible for longitudinal changes in gravel size in rivers on alluvial fans. A rotating drum with three interior vanes was used to produce repeated collisions among test gravel particles in water. The drum was only partially filled with water in order to simulate the impact velocity of particle collisions during a flood. The test particles were slightly weathered andesite and chert taken from the bed of the Watarase River, Japan. Experiments were conducted using three unisize grains-L, -7.0 phi (128 mm) to -6.5 phi (91 mm); M, -5.5 phi (45 mm) to -5.0 phi (32 mm); S, -4.0 phi (16 mm) to -3.5 phi (11 mm)-and mixtures of two different sizes from the three size classes. There are six principal results from the experiments. (1) Test gravel particles break frequently and decrease in weight rapidly. (2) Abrasion of chert produces mostly gravel, while andesite produces mostly sand and silt. (3) L-size chert cobbles decrease in weight rapidly as a result of being broken into smaller pieces, while andesite cobbles break so rarely that their weight decreases very slowly. (4) S-size andesite pebbles decrease in weight more rapidly than chert. (5) Abrasion depends strongly on the presence of different sizes, because smaller fragments are crushed by larger gravel particles. (6) Diminution coefficients of andesite and chert obtained from the experiment are in the range of 10 (super -3) -10 (super -1) km (super -1) . Diminution coefficients from this experiment are one to two orders of magnitude larger than those from previous experiments on abrasion. This is mainly because of the larger impact velocities in this experiment. Diminution coefficients from the experiment are also consistent with those obtained from many Japanese rivers on alluvial fans (10 (super -2) -10 (super -1) km (super -1) ). This result suggests that downstream fining of river-bed gravel on alluvial fans can be explained by abrasion.