In a field experiment in Western Australia, six different levels of three different phosphorus (P) fertilizers (triple superphosphate, TSP; Queensland (Duchess) rock phosphate, QRP; North Carolina rock phosphate, NCRP) were applied at the start of the experiment in 1984. Grain yield of triticale (X Triticosecale) was measured from 1984 to 1988. In February-March of each year from 1985 to 1988, soil samples were collected to measure soil extractable P (soil test values) using four reagents (Bray 1, calcium acetate lactate (CAL), Truog and Colwell). Soil test values were related to triticale grain yields, determined either as absolute yield or percentage of the maximum yield, produced later on in each year. The relationship differed with fertilizer type, reagent and year. All four soil test reagents were equally predictive of yield. It is concluded that these soil P tests provide crude predictions of plant yield regardless of the reagent used.