In order to predict the occurrence of water repellency, which is a labile property, from field survey data obtained throughout the year, it is necessary to identify predictive relationships between water repellency and commonly measured soil properties. This paper evaluates these relationships for diverse soil assemblages.These soil assemblages include a set of reference soils from the south-west of Western Australia (an area of 250 000 km(2)), more intensively sampled suites of soils in several smaller soil-landscape associations within the south-west of Western Australia (congruent to 1000 km(2)), soils from single farms (1-10 km(2)) and transects (congruent to 0.001 km(2)), and single soil profiles (congruent to m(2)). The severity of water repellency was assessed by measuring water drop penetration time in seconds (WDPT) and was related to intrinsic properties of soils using log-transformed data. For the set of soils from the West Midland Sandplain the type of land use was also considered as a variable.There is a general tendency for WDPT to increase as organic matter content increases and decrease as the content of fine mineral material increases (clay, silt, very fine sand). However, there is no single soil property that is able to predict WDPT adequately. Furthermore, reliability of prediction decreases as the area of sampling increases. There appear to be no systematic differences in the capacity of organic matter from pasture or crop to induce water repellency, but increments of organic matter under bush increase water repellency at a greater rate than does organic matter from crop or pasture.