Random sampling of surface rocks on a gravel bar is biased toward larger sizes which, because of their area, are more likely to be picked up. Weighting can eliminate this bias. Data on average weight of a single rock are used to change numbers of rocks to weights, thus yielding size frequency data in general agreement with a sieved and weighed sample. The question of what to sample depends on the use to which the data are to be put, and is not treated in detail in this paper.