The decision whether to take additional remedial action (removal of soil) from regions contaminated by uranium mill tailings involves collecting 20 plugs of soil from each 10-m by 10-m plot in the region and analyzing a 500-g portion of the mixed soil for /sup 226/Ra. A soil sampling study was conducted in the windblown mill-tailings flood plain area at Shiprock, New Mexico, to evaluate whether reducing the number of soil plugs to 9 would have any appreciable impact on remedial-action decisions. The results of the Shiprock study are described and used in this paper to develop a simple model of the standard deviation of /sup 226/Ra measurements on composite samples formed from 21 or fewer plugs. This model is used to predict as a function of the number of soil plugs per composite, the percent accuracy with which the mean /sup 226/Ra concentration in surface soil can be estimated, and the probability of making incorrect remedial action decisions on the basis of statistical tests. 8 refs., 15 figs., 9 tabs.