A three-year experiment to compare foliar application of five potassium sources was conducted on Loring-Calloway silt loam in 1991-1993 at Marianna, Arkansas. The K salts used were sulfate, chloride, thiosulfate, carbonate and nitrate. Bi-weekly solutions of these salts were applied four times in 1991 and 1993 and five times in 1992 during the 2nd through the 8th weeks of flowering. Each application consisted of the equivalent of 10 lb K nitrate in 10 gals of solution/A. The check and treatments other than K nitrate received N as urea at the rate supplied by the K nitrate treatment. The experimental site was relocated in 1992 and again in 1993 to decrease soil physical variation. Composite soil samples of 0-6-in and 7-12-in soil zones were taken before planting each year. In 1991, Melich 3 test K levels were 179 and 176 lb/A in the 0-6- and 7-12-inch depths, respectively. The corresponding values were 212 and 157 in 1992 and 252 and 236 in 1993. The K levels of 212-252 in the 0-6-in zone of this soil are considered medium for cotton. The K levels of 176 and 236 in the 7-12-in zone are higher than normal. The soil was fertilized with 60 lb/A of K₂O in 1991 and no K in 1992 and 1993. The foliar treatments affected the total cotton yield only in 1992 when the overall annual yield was the highest. The 1992 yield produced by the K nitrate treatment was significantly higher than those of the check and K carbonate treatment. The 1992 yield produced by K carbonate was significantly lower than that of the check. In 1993, the K nitrate treatment produced significantly more yield at second harvest than all other treatments. In 1992 the K nitrate treatment produced a significantly higher boll weight than that of the K carbonate. In 1993 the K chloride treatment produced a significantly higher boll weight than that of the K carbonate treatment. The % soluble K in petioles of the most recently mature leaves of check plots at 3-8 weeks after first flower ranged from 6.0 to 3.7 in 1991, 5.3 to 3.5 in 1992 and 5.5 to 4.3 in 1993. At flowering weeks 10 and 11 in 1992, % petiole K dropped to 2.0 and 1.3, respectively. Extreme soil variation limited precision in measuring treatment effects in 1991. Moisture deficiency and unusually high temperatures limited yields in 1993. Differences in seasonal conditions among the three years may have contributed to the differing treatment responses. In another 1991 field experiment on a non-irrigated farm field, four bi-weekly applications of 4.4 lb each of N and K₂O as urea-KNO₃ and as urea-KCl produced equally significant yield increases compared with the check. The 0-6-in soil zone in that field contained 170 lb/A Melich 3 soil test K. In two other 1991 field studies, (1) no important leaf-burn resulted from foliar treatment at first flower with KNO₃ at per A rates and solution volumes of 10 lb in 5, 7.5 and 10 gal and 20 lb in 10, 15 and 20 gal. (2) No important leaf-burn resulted from foliar application of mixtures of urea, KNO₃ and boron when applied in 5 or 10 gal of solution/A at 4 rates/A of N, K₂O and B respectively as follows: 13-6-0.2, 10-10-0.3, 15-8-0.3 and 3-11-0.3.