In 12-year (1975–87) fertilizer trial on a pearl millet-wheat rotation, there were no responses to applied K and exchangeable K decreased from 620 to 200 kg/ha. In spite of progressively diminishing exchangeable K, removal of K by crops was independent of soil K status, indicating that exchangeable K had no relation to K uptake in this soil. Thus, nonexchangeable K made a major contribution to plant K, up to 90% in untreated plots and > 70% in K-treated plots. A significant (r = -·708) relationship between nonexchangeable K and K fixation capacity was found. Nonexchangeable K was also significantly but negatively correlated (r = -0·583) with free energy exchange (ΔGr). Potassium supply values, expressed by pK-½p (Ca + Mg), were positively and significantly correlated with K removal by crops in 1987 and with the K saturation values of the exchange phase, so that both these measures can be successfully used to test soil K availability. The results of this study suggested that on this soil, under a pearl millet-wheat crop system, inclusion of K in the fertilizer programme may soon be needed.