Abstract: Detection of vegetation change in arid Australian rangelands at different scales and with different ground-based techniques was investigated. Changes caused by grazing could not be detected without stratifying for spatial patterns at fine scales. Temporal variation in rainfall and grazing was such that species numbers fluctuated markedly; numbers were best assessed when favourable conditions produced the maximum response. More heavily grazed sites supported fewer species and composition was unstable over time; there was a threshold beyond which recovery did not occur. Several species were identified as useful indicators but it was not possible to reduce them to functional groups, to simplify interpretation.