Field studies were conducted in North Alabama to evaluate cotton (Gossypium hirsutum) response to long-term applications of K fertilizer, and to evaluate K movement and status within the soil profile resulting from these applications. The test was initiated in the fall of 1986. Potassium was applied to a Dewey silt loam (Typic Paleudult) at rates of 0, 60, 120 and 180 lb K₂O acre^-1 for the first three years of the study. No K fertilizer was applied during the fourth year of the study (1990). After the fourth year of the test, half of the plots were left in residual. The experiment had a split plot design with four replications. During 1987-1989, two varieties were compared (Stoneville 825 and Deltapine 50) and these served as whole plots. During 1991-1997, only one variety was evaluated and annual versus residual K fertilization served as whole plots. Potassium treatments have been used as subplots throughout the test. During the first three years of the test there were no consistent differences between Stoneville 825 and Deltapine 50. Lint yields were increased significantly by K fertilization in one out of every two years during this 10-year test. In non-responsive years, a lack of a response to K fertilization was attributed to poor rainfall amount and/or distribution. In good production years, K fertilization on this soil increased lint yields by as much as one bale per acre. During 1991-1996, residual K fertility treatments consistently produced lower yields as compared to annual K treatments. In this test there were no significant differences between split (fall-spring), fall or spring applied K with respect to lint yields. The yield response data from this test gives strong support to the calibration data currently used by the Auburn University Soil Testing Laboratory. For the duration of this test, applying K according to the recommendations of the Auburn University soil testing laboratory resulted in an average increased yield of 2.8 lb lint for every lb of applied K₂O. Micronaire increased with increasing K rates and was the only lint quality parameter that was consistently affected by K fertilization. Soil analysis data showed increased levels of Mehlich 1, ammonium acetate and boiling nitric acid (nonexchangeable) extractable K in the plow layer of this soil. Data collected in 1989, 1993 and 1997 indicate that there was no detectable downward movement of K in this soil.