A number of geological inferences are made from studies of the geographic distributions of sediment textural variables. Evaluation of such distributions requires that information on the total within-station variability be available. In order to obtain information on the relative magnitudes of the within-station and among-station variabilities, bottom sampling on the continental shelf of Washington was done in a stratified-multistage scheme on traverses placed by a systematic random scheme along the coast. Duplicate samples were taken at each of 450 stations and duplicate analyses of nine textural variables were made on each sample. Complete analyses are available for 130 stations. This sampling design leads directly to a three-level, hierarchical analysis of variance in which the F-ratios of the mean square among stations to that within staitons are measures of the reliability of distributions. The most diagnostic variables for differentiating stations are percentage of sand, percentage of silt, sand/mud ratio, and median phi size. The least diagnostic is percentage of clay.