Spatial variation of soil test phosphorus (P) was studied by measuring soil‐test P for individual soil samples collected from uniform areas in south‐western Australia. The soil samples were collected using standard 10 cm deep by 2.5 cm diameter samplers. The Colwell alkaline bicarbonate soil test, the standard reagent adopted in Western Australia, was used. There were two parts to this study. In Part 1, 36 individual soil samples were collected on 20 m by 20 m grids within a uniform 100 m by 100 m area in paddocks on 16 different lateritic soil types. Superphosphate had been applied each previous year, using normal farmer practices, for either nine years at two sites or >20 years at the other sites. For pastures, fertiliser is usually applied to the surface. For crops, it is drilled with the seed at about 5 cm depth. In Part 2, 10 to 20 individual soil samples were collected at random locations within experimental plots (either 2 or 4.8 m wide by 30 m long) of five long‐term field experiments that were treated once only with different amounts of superphosphate applied from two to eight years previously. The P was applied to the soil surface and either incorporated through the soil with a rotary hoe or when sowing crops at 5 cm depth with tined implements, or the P was placed at 5 cm depth in bands that were 180 mm apart and the soil was not disturbed thereafter. For each site in Part 1, or each plot of each experiment in Part 2, soil‐test P for the individual soil samples varied, often markedly. Coefficient of variation was large. It was up to 56% for Part 1. For Part 2, it was up to 82% when P was incorporated into the topsoil, and up to 210% when P was banded in the soil. Spatial variation of soil‐test P is attributed to natural variation of soil physical, chemical, and mineralogical properties. Some of the variation may also be due to heterogenous mixing through the topsoil of fertiliser P applied in recent years. Soil‐test P is usually larger for recently applied fertiliser P.