Analysis of spatial pattern of the seed bank of an old field in Maryland showed that the seed bank was composed primarily of annual and short—lived perennial species. Tree and shrub species were poorly represented. Seed abundance decreased with depth in the soil. The seeds of all species generally had clustered distributions in the soil. However, there was no trend in mean cluster size within or among species, or at different soil depths. The lack of definite trends indicates that the spatial distributions of seeds are governed primarily by environmental factors, and to a lesser degree by biological factors. The method of spatial pattern analysis developed for this study represents a significant improvement over traditional contiguous quadrat methods. In this method, individuals' positions are mapped in the field using cartesian coordinates, and the quadrat data are subsequently derived mathematically. The resolution of pattern is greatly improved, and several problems that reduce the effectiveness of other methods are eliminated. This method can be used to advantage in any study in which spatial patterns analysis is performed, particularly if the objects can be mapped easily. A comparison of three sampling techniques revealed that the precision of seed—number estimates can be improved by taking a large number of small samples, or by subsampling large whole units with very small subunits. The smaller number of large samples favored by many seed—bank researchers generally results in imprecise seed estimates. This result has implications for spatial studies of any clustered organism.