An experiment was conducted to determine the effect of rates and sources of Mg on yield and leaf composition of the Grand Nain banana, grown on a Corozal clay during a 3-year period. The Mg rates tested were 0, 56, 112, 168, 224 and 448 kg/ha/year. The source of Mg for treatments 2, 3 and 4 was magnesium sulfate (Epsom salt), and for treatments 5 and 6 dolomitie limestone the first year and commercial magnesium oxide (Fert-O-Mag) in subsequent years. An optimum yield response of 8.9 t/ha/ year of bananas was obtained with annual applications of about 275 kg/ha of Mg from dolomitic limestone the first year and commercial magnesium oxide in subsequent years. Considering magnesium sulfate as the source of Mg, and estimating the cost of materials and local farm gate banana prices, the optimum economic response is obtained with applications of 112 to 168 kg/ha/year of Mg. This represents a yield increase of 5.8 to 7.7 t/ha/year of fruits over the O Mg treatment. Magnesium concentration in the third youngest banana leaf averaged 0.30% in plants that received 224 and 448 kg/ha of Mg from dolomite in the plant-crop.