Soil samples were collected from 45 sites in 15 field sites in the fall of 1982 and again in 1983. In a bioassay procedure, fungi were isolated from necrotic hypocotyls of cotton seedlings grown in the samples, and disease severity index values were determined. The samples were also characterized as to textural class, pH, and organic matter content. In 1984, cotton was planted at the sites from which samples were previously taken, and seedling disease determinations were made. Isolations of Pythium spp. and disease indices from the bioassay procedure were positively correlated with disease severity in the field plantings. Soil clay content was inversely correlated with disease severity. From stepwise multiple regression analyses severe seedling disease was related to a high soil bioassay disease index value and high isolation frequencies of Pythium spp. and Thielaviopsis basicola (but not Rhizoctonia solani), along with low soil clay and high organic matter content. Use of the- stepwise regression models could aid in predicting cotton seedling disease severity in these loessial soils.