Frequent disturbances such as fire are widely considered important drivers of plant composition and diversity in productive grassy ecosystems. Effects of fire frequency on grassland soils, however, are less well understood. We established replicated disturbance regimes in a high-quality, representative Themeda australis–Poa sieberiana-derived grassland in south-eastern Australia that had historically been burnt every 4–8 years. Effects on soil chemical, physical and biological properties were measured after 10 years of application of 2-, 4-, and 8-yearly burning, 2-yearly mowing and an undisturbed treatment. Contrary to other grassy ecosystems, there were no detectable effects of disturbance regime on total soil nitrogen and carbon, or a range of other soil chemical properties in the top 10 cm. However, a cumulative effect of burning on the grassland soil was evident from a suite of changes to soil surface properties, available nutrients and biological activity. In particular, on biennially burnt plots, reduced litter and plant protective cover were associated with increased soil surface compaction, decreased infiltration and decreased soil biological activity, which in turn were related to poor sward recovery after fire and drought. These relationships indicate potential for positive feedbacks whereby repeated removal of soil protective cover and changes to soil surface chemistry through very frequent burning ultimately lead to further reduction in soil protective cover through reduced productivity. However, this is only likely in extreme cases: data from unburnt plots indicated that soils that had historically been burnt every 4–8 years had not passed a threshold beyond which such soil changes were irreversible or damaging. Contrary to other predictions, cessation of burning for 13 years did not lead to detectable soil nutrient release through senescence of dominant grasses. Biennial mowing with slash retention was an effective alternative disturbance for maintaining sward vigour while avoiding soil surface damage.