Summary. The agronomic effectiveness of directly applied North Carolina reactive phosphate rock was determined for 4 years from annual dry matter responses at 26 permanent pasture sites across Australia as part of the National Reactive Phosphate Rock Project. Fertiliser comparisons were based on the substitution value of North Carolina reactive phosphate rock for triple superphosphate (the SV50). The SV50 was calculated from fitted response curves for both fertilisers at the 50% of maximum yield response level of triple superphosphate. The reactive phosphate rock was judged to be as effective as triple superphosphate in the 1st year (and every year thereafter) at 4 sites (SV50 >0.9), and was as effective by the 4th year at 5 sites. At another 9 sites the reactive phosphate rock was only moderately effective with SV50 values between 0.5 and 0.8 in the 4th year, and at the final 8 sites it performed poorly with the 4th year SV50 being less than 0.5. Pasture environments where the reactive phosphate rock was effective in the 1st year were: (i) those on sandy, humic or peaty podsols with an annual rainfall in excess of 850 mm; (ii) those on soils that experienced prolonged winter inundation and lateral surface flow; and (iii) tropical grass pastures in very high rainfall areas (>2300 mm) on the wet tropical coast on North Queensland. The highly reactive North Carolina phosphate rock became effective by the 4th year at sites in southern Australia where annual rainfall exceeded 700 mm, and where the surface soil was acidic [pH (CaCl2) <5.0] and not excessively sandy (sand fraction in the A1 horizon <67%) but had some phosphorus (P) sorption capacity. Sites that were unsuitable for reactive phosphate rock use in the medium term (up to 4 years at least) were on very high P-sorbing krasnozem soils or high P-sorbing lateritic or red earth soils supporting subterranean-clover-dominant pasture, or on lower rainfall (< 600 mm) pastures growing on soils with a sandy A1 horizon (sand component >84%). No single environmental feature adequately predicted reactive phosphate rock performance although the surface pH of the soil was most closely correlated with the year-4 SV50 (r = 0.67). Multiple linear regression analysis found that available soil P (0–10 cm) and the P sorption class of the surface soil (0–2 cm), together with annual rainfall and a measure of the surface soil"s ability to retain moisture, could explain about two-thirds of the variance in the year-4 SV50 . The results from this Project indicate that there are a number of specific pasture environments in the higher rainfall regions of Australia where North Carolina reactive phosphate rock can be considered as an effective substitute P fertiliser for improved pasture.