Fire-induced or enhanced soil water repellency is often viewed as a key cause of the substantial increases in runoff and erosion following severe wildfires. In this study, the effects of different fire severities on soil water repellency are examined in eucalypt forest catchments in the Sandstone Tablelands near Sydney, burnt in 2001 and 2003. At sites affected by different fire severities and in long-unburnt control sites, repellency persistence was determined in situ and in the laboratory for surface and subsurface soil samples (n=846) using the Water Drop Penetration Time (WDPT) test. All long-unburnt samples were found to be water repellent, with severe to extreme persistence (>900 s) being dominant for surface (0–2.5 cm) and slight to moderate persistence (10–900 s) for subsurface (2.5–5 cm) soil, indicating naturally very high ‘background’ levels of repellency. In contrast to the generation or enhancement of repellency usually reported following forest fires of similar severity in previous studies, burning caused widespread destruction of repellency. The mineral soil depth to which repellency was destroyed (0.5–5 cm) was found to increase with burn severity. Below this charred wettable layer, persistence of pre-existing water repellency increased. Two years after the fire, the frequency of extreme repellency persistence was reduced in the surface and subsurface. However, recovery to pre-fire repellency levels had not been achieved. The associated hydrological impacts of these fire effects are more complex than simply the enhancement of overland flow, runoff and soil erosion with increasing fire severity. For forest fires sufficiently severe to remove foliage and ground litter above already repellent soil, a more severe burn, in which there is destruction of surface soil repellency, would result in lower runoff response compared to a burn insufficiently severe to destroy surface repellency. During storms intense enough to saturate the wettable surface rapidly, this layer may, however, be removed by overland flow, with potentially severe implications for soil fertility and seedbed survival, post-fire ecosystem recovery, and downstream sedimentation and water quality. The results demonstrate that existing fire severity classifications are not well suited to predicting fire impacts on soil hydrological responses and highlight the need for a new fire severity evaluation scheme. A scheme encompassing not only foliage and ground cover status, but also changes to surface and subsurface soil hydrological properties, would provide a better prediction of the immediate hydrological effects of wildfires on catchments such as flash flooding and erosion, and also of their time-to-recovery than current classifications allow. Such a scheme could prove invaluable given the future increase in fire frequency and severity predicted for many regions.