Abstract: Although cotton (Gossypium hirsutum L.) is generally much more sensitive to potassium use deficiency (KUE) than most other crops, potassium (K) requirements of different cotton cultivars vary greatly. They are therefore critical both to identify high KUE cotton genotypes, and determine the differences in the mechanisms of K use among different cotton genotypes. In recent decades, the use of grafting to enhance cotton resistance to soil diseases has been extended to increase nutrient use efficiency (NUE). To that end, two cotton genotypes (a high KUE and yield potential 103 and a low KUE and yield potential 122 genotypes) were carefully screened from 86 cotton cultivars in the Plant Nutrition Laboratory of Huazhong Agricultural University during 2001—2005. Then a pot experiment was conducted to determine the differences between the screened genotypes in terms of KUE, yield, assimilation partitioning, root/shoot ratio, harvest index and K content/partition. Both the ungrafted (103 and 122) and self-grafted (103/103 and 122/122) cotton genotypes were investigated under different K applied 0.55 g(K₂O)·kg^-1(soil) and 0 g(K₂O)·kg^-1(soil) treatments. The results showed that grafting had different degrees of effect on dry-matter and K accumulation/distribution and yield of the two cotton genotypes under different K conditions. Dry-matter and accumulated K contents of vegetative organs were higher in 103/103 than in 103. Conversely, dry-matter and accumulated K contents of reproductive organs were lower in 103/103 than in 103. Furthermore, 103/103 yield and K use index declined relative to 103. Interestingly, all the conditions of the low KUE cotton genotype 122 completely changed after grafting. That was to say that dry-matter and accumulated K contents of vegetative organs were lower in 122/122 than in 122. However, dry-matter and accumulated K contents of reproductive organs were higher in 122/122 than in 122. K uptake efficiency varied with K application level. While K uptake efficiency was lower in 103/103 than in 103 under applied K treatment, the reverse was true under 0 g(K₂O)·kg^-1(soil) treatment. On the contrary, K uptake efficiency was higher in 122/122 than in 122 under 0.55 g(K₂O)·kg^-1(soil) treatment while the reverse was true under 0 g(K₂O)·kg^-1(soil) treatment. It suggested that grafting differently influenced different cotton genotypes, with a more balanced dry-matter distribution in grafted cotton genotypes. The results laid the theoretical basis for further insight into K uptake and use efficiency by different cotton genotypes.