Dryland salinity in southern Australia has been caused by inadequate water use by annual crops and pastures. The purpose of this study was to compare the water use of annual pastures and Medicago sativa L. ( lucerne) with Dorycnium hirsutum ( L.) Ser., a potential new perennial forage species. The soil water dynamics under bare ground, annual legume-, lucerne-, and D. hirsutum-based pastures were compared at 2 sites in the low- ( Merredin) and medium- ( New Norcia) rainfall wheatbelt of Western Australia between September 2002 and February 2005.Soil under D. hirsutum was drier than under annual pastures by 8 - 23mm in Year 1, 43 - 57mm in Year 2, and 81mm in Year 3. Lucerne used little additional water (< 19 mm, n.s.) compared with D. hirsutum and pro. le soil water content was similar under both species throughout the experiment. At Merredin, annual pastures used water to a depth of 1.0 m, whereas under both D. hirsutum and lucerne in the first 3 years after establishment the successive maximum depth of water use was 1.0, 1.8, and 2.2 m. At New Norcia, additional soil water was extracted by lucerne and D. hirsutum at depths < 1.0m and no difference between treatments was detected below 1.0 m. Biomass of D. hirsutum pasture harvested in autumn contained minimal annual components and was 15 - 50% of that produced by lucerne- or annual legume- based pastures. D. hirsutum and lucerne plant density declined each summer ( 25 80%), but D. hirsutum density was lower than lucerne due to poorer establishment. Nonetheless, the comparable water use of lucerne and D. hirsutum suggests that D. hirsutum could make reductions in recharge similar to those of lucerne in the Western Australian wheatbelt.