Existing theoretical frameworks suggest three predictions relevant to grazing effects in Australian aridlands: grazing has a negative but moderate effect on plant species richness; a separate "state" resulting from degradation caused by extreme grazing will be evident; some plant species will have a strong association with grazing relief refuges that have only ever been subject to light grazing. These predictions were examined in the dune swales of an Australian desert, with data on herbaceous species collected along transects up to 14 km from artificial water points between four and 33 years old. A cumulative grazing index was constructed utilizing both the spatial occupation patterns of cattle and the length of exposure. Despite restricting sampling to a narrow habitat, silt/clay content and soil pH influence floristic patterns independent of grazing. The analysis of quadrat data in relation to grazing revealed almost no patterns in plant cover, species richness (at two different scales), or abundance across plant life-form groups. Five species had an increasing response, and seven a decreasing response, while the only species restricted to areas of extremely low grazing pressure was sufficiently rare that it could have occurred there by chance. The dominant annual grass, the most common shrub, and a perennial tussock-forming sedge all decrease with high levels of grazing. Most species exhibit an ephemeral life strategy in response to unreliable rainfall, and this boom and bust strategy effectively doubles as an adaptation to grazing. After 20 years of exposure to managed grazing with domestic stock in Australian dune swales, patterns in species richness have not emerged in response to grazing pressure, the ecosystem has not been transformed to another degradation "state," and there is no evidence that grazing relief refuges provide havens for species highly sensitive to grazing.