Poor yields of East African highland bananas (Musa spp., AAA-EAHB) on smallholder farms have often been attributed to problems of poor soil fertility. We measured the effects of mineral fertilizers on crop performance at two sites over two to three crop cycles; Kawanda in central Uganda and Ntungamo in southwest Uganda. Fertilizers were applied at rates of 0N-50P-600K, 150N-50P-600K, 400N-0P-600K, 400N-50P-0K, 400N-50P-250K and 400N-50P-600K kg ha_1 yr_1. In addition 60Mg-6Zn-0.5Mo- 1B kg ha_1 yr_1 was applied to all treatments, with the exception of the control plots which received no fertilizer. Fresh bunchmass and yield increased with successive cycles. Yield increases above the control ranged from 3.1 to 6.2 kg bunch_1 (average bunch weight for all treatments 11.5 kg bunch_1) and 2.2- 11.2Mg ha_1 yr_1 (average yield for all treatments 15.8 Mg ha_1 yr_1) at Kawanda, compared with 12.4- 16.0 kg bunch_1 (average bunch weight for all treatments 14.7 kg bunch_1) and 7.0-29.5 Mg ha_1 yr_1 (average yield for all treatments 17.9 Mg ha_1 yr_1) at Ntungamo. The limiting nutrients at both sites were in the orderK > P > N. Potassium,Nand P foliar nutrient mass fractions were below previously established Diagnosis and Recommendation Integrated System (DRIS) norms, with the smallest K mass fractions observed in the best yielding plots at Ntungamo. Total nutrient uptakes (K > N > P) were higher at Ntungamo as compared with Kawanda, probably due to better soil moisture availability and root exploration of the soil. AverageN, P andKconversion efficiencies for twocrop cycles at both sitesamounted to 49.2 kg finger DM kg_1 N, 587 kg finger DM kg_1 P and 10.8 kg finger DM kg_1 K. Calibration results of the model QUEFTS using data fromNtungamo were reasonable (R2 = 0.57, RMSE = 648 kg ha_1). Using the measured soil chemical properties and yield data froman experiment atMbarara in southwest Uganda, the calibrated QUEFTSmodel predicted yields well (R2 = 0.68, RMSE = 562 kg