In South Africa there is widespread reliance on kikuyu grass as a source of summer forage for intensive dairy, beef and sheep enterprises. There is, however, a dearth of reliable information on the fertility requirements of kikuyu pastures. Responses of previously productive and unproductive kikuyu pastures in the Mistbelt and Highland Sourveld of KwaZulu‐Natal to top‐dressed nitrogen (N) and phosphorus (P) fertilizers were studied infield cutting trials conducted during two consecutive seasons. Significant dry matter (DM) yield responses accompanied incremental N applications of up to 500 kg ha−1. First‐season DM yields of previously productive pastures at zero applied N were in the range 6 to 9 t ha−1, while yields of previously unproductive pastures at zero N were 0 to 2 t ha−1. By the second season, yields at zero N were uniformly low on the various pastures, reflecting the limited extent of plant‐available N reserves in soils under productive kikuyu. In both the Mistbelt and Highland Sourveld, efficiency of response to applied N in seasons when rainfall was similar to the long‐term mean was approximately 30 kg DM per kg N over the N application range of 0 to 250 kg N ha−1. When rainfall was substantially below average, N efficiencies decreased markedly. Application of N in six rather than three dressings during the growing season resulted in the N efficiency increasing by 3.5 kg DM per kg N. In contrast to the enormous responses to N, there was little or no response to P in the various trials. An optimum P test (NH4HCO3/ NH4F / EDTA) for high clay content (>50% clay) Mistbelt soils of approximately 10 mg L−1is suggested from yield‐soil test calibrations. Data highlighting the impact of soil sampling depth on P test values are presented, and this information serves to underline the importance of standardizing sampling procedures in the soil testing of pastures for fertility purposes.