Soil, pasture and animal responses to 2 levels of superphosphate were measured within a farmlet trial, at Ballantrae hill country research area, near Woodville, during 1975-80. Soil and pasture measurements were also made for 3 years before the trial started. Pastures, which were dominated by low-fertility-tolerant grasses (LFTC;), moss and flatweeds, were oversown with 4 legumes before the trial started. Soil Olsen P level was 5 under untopdressed pasture in 1973, and 9 in 1975 after application of 500 kg/ ha superphosphate. Superphosphate rates were 120 (on low fertiliser areas q LF) and 640 (on high fertiliser areas = HF) kg/ ha/ yr average during 1975/80. Annual pasture production (adjusted for variable summer rainfall) was 7.1 t DM/ ha from untopdressed pasture and 8.1 t following 500 kg superphosphate/ ha in 1973-74. LF production rose to 8.5, and HF to 12.0, in 1980. Legume contribution rose from 5% in untopdressed pasture to 18% and 23% in LF and HF respectively in 1975/76. By 1979/80 legume contribution had steadily declined to 10% on both treatments, probably due to a measured increase in soil N availability and increased competitiveness of associated grasses. Ryegrass content rose at both fertiliser levels, while LFTG content fell; these trends were greatest under HF. Winter stocking rate was increased from 6 to 10.9 and 14.9 su/ha on LF and HF-respectively. Per animal performance did not decrease. A range of techniques which could increase efficiency of superphosphate use in hill country is discussed. These techniques include: selective application to responsive pastures; spring application; grazing management to encourage legume growth.