Potassium deficiency is widespread in cotton growing areas around the world. Many of the cotton producing states have initiated research programs to re-evaluate current K fertilizer recommendations, with conflicting results. It is, therefore, essential that the soil scientists across the Cotton Belt work together in a unified manner, seeking the principles which are true on all soil types, and using creative approaches to address this problem. The objectives of this study were: 1) to develop a method whereby we can predict cotton response to K fertilization across the U.S. Cotton Belt using the K buffering capacity, 2) to determine the relationship of commonly measured soil physical and chemical properties to KBC for a number of diverse cotton-producing soils, and 3) to compare soil extractants and K fertilizer recommendations used across the Cotton Belt. Locations were chosen to represent the major soil types across the Cotton Belt. Variations in soil texture and clay mineralogy are represented in this selection of 45 site-years. Each cooperator used varieties and management practices common to his/her state. Yield was measured by treatment in tests with a randomized complete block design with three application rates (0, 50, and 100 lb K₂0/A) and four replicates. Soil samples were taken to 45 cm depth (in 15 cm depth increments) from each location prior to K fertilization. The following initial physical and chemical properties were determined: particle size analysis, clay mineralogy, organic matter content, CEC, soil pH, NH₄OAc-extractable K, Ca, and Mg, Mehlich-1 extractable K, Ca, and Mg, and Mehlich-3 extractable K, Ca, and Mg. Potassium fixation percentage was measured using a seven-day incubation in 10 mM KNO₃ and two sequential extractions. After clay mineralogy was determined by x-ray diffraction, the soils were grouped into seven mineralogical classes. All of the experimental sites which had yield responses to K fertilizer were in mineralogical classes with moderate to high amounts of kaolinite. When K fixation was < 100% there was a 44% probability of yield response to K across all experimental sites. When K fixation was between 100 and 110%, probability of yield response was 23%; and above 110%, probability of yield response was 0%. Potassium fixation was positively correlated with CEC, and was more highly correlated with Mehlich-3 and ammonium acetate extractable K than with Mehlich-1 extractable K. The use of K fixation percentage as a predictor of cotton yield response to K fertilization shows promise across the Cotton Belt and merits further research.