Information on fertilizer response in cassava in Africa is scarce. We conducted a series of on-farm and on-station trials in two consecutive years to quantify yield responses of cassava to mineral fertilizer in Kenya and Uganda and to evaluate factors governing the responses. Average unfertilized yields ranged from 4.2 to 25.7 t ha-1 between sites and years. Mineral fertilizer use increased yields significantly, but response to fertilizer was highly variable (- 0.2 to 15.3 t ha-1). Average yield response per kg applied nutrient was 37, 168 and 45 and 106, 482 and 128 kg fresh yield per kg of applied N, P and K, respectively in 2004 and 2005. Fertilizer response was governed by soil fertility, rainfall and weed management, but was not influenced by variety, pest and disease pressure and harvest age. Relative N and K yields were positively correlated to SOC and exchangeable K, while response to fertilizer decreased on more fertile soils. Still, fertilizer response varied widely on low fertility soils (e.g. on soils with <10 g kg-1 SOC, responses ranged from - 8.6 to 24.4 t ha-1), indicating strong interactions between factors governing fertilizer response. Response to fertilizer was reduced if total rainfall <1,500 mm or rainfall from 0 to 3 months after planting <400 mm. Fertilizer application promoted plant growth and resulted in a better soil coverage and reduced weed competition. Yields in fertilized fields were independent of weed management, unless growing conditions were unfavourable.