A variety of hydraulic and ecological processes in rivers depend in part on the grain size distribution of the riverbed. For this reason it is important to accurately sample the bed to determine the grain size distribution. To describe the accuracy of riverbed sediment samples, techniques based on the multinomial distribution are presented. These techniques allow confidence intervals to be calculated for grid, areal, and volume samples. Additionally, sample sizes can be estimated to achieve a desired accuracy for the three sample types. A transformation is necessary when considering areal and volume samples. This is because neither of these two sample types is well conditioned for statistical analysis. The transformation is based on the results of a previously performed grid sample. The transformation then estimates either an areal or volume sample size that is necessary to match the accuracy of the grid sample. This procedure is useful when obtaining a hybrid distribution from grid and areal samples. Having areal and grid sample results of known accuracy allows confidence intervals to be placed on the final hybrid distribution. A series of laboratory experiments is used to illustrate the various techniques. Comparisons are also made with techniques presented in the literature. ¿ 2000 American Geophysical Union