Soybean (Glycine max(L.) Merrill) has been shown to have a great capacity to take K from soil, suggesting that it might absorb K from non‐exchangeable forms. In this paper, the effect of level of K fertilization on soil exchangeable K content and K uptake by soybean are discussed. The experiment was conducted on a Typic Haplortox (sandy loam), fertilized with 0, 40, 80, 160 and 240 kg K2O/ha as KC1 or K2SO4. During five years before the experiment, half of the plots received those K rates annually and the other half only in the first three years, providing an opportunity to study the residual effect of applied K. Plant samples were taken at pod filling and at harvest. Soil cores were collected in 20 cm increments down to 80 cm deep at plant emergence, pod filling and after harvest. There was a residual effect of K, and 240 kg K2O/ha applied in a 3‐year term led to the same yield and K uptake as 80 kg ICO/ha applied annually for 3 years. Fertilized plants absorbed 160% more K than unfertilized ones, but soil exchangeable K accounted for less than 50% of plant uptake; therefore the exchangeable pool must have been replaced in time for soybean uptake. On the other hand, the K recycled from the plant to the soil was not found in the exchangeable form. There was evidence of K leaching deeper than 80 cm, and in addition, the K recycled from the plants may have turned into non‐exchangeable forms in the soil.