The presence of host plants was found to be a key determinant of soil infectivity for endophytes of epacrids growing in south-west Australian jarrah forest carrying a sparse epacrid understorey. A comparable Banksia woodland with higher density of epacrids showed more uniform levels of infective endophytes, presumably because of closer overlap of host-supported patches of endophyte. Disturbed sites in jarrah forest, in which topsoil had been recently returned following bauxite mining, initially showed extremely low endophyte infectivity, probably due to disruption of the hyphal network, absence of host plants and/or increased competition from soil micro-organisms antagonistic to the endophytes. As the system stabilized and epacrids recruited and grew, inoculum potential of the soil re-established and by 12 yr it equalled that of adjacent undisturbed native forest. At disturbed sites with sparsely distributed adult members of the Epacridaceae, endophyte inoculum proved to be adequate only directly adjacent to the epacrids and declined steeply to negligible levels at only 40 cm radially distant from a host plant. The significance of live epacrid root systems on survival of endophyte inoculum in Banksia woodland was examined following removal of shoots of adult plants of the endemic epacrid, Leucopogon conostephioides DC. Coincident with demise of roots of the detopped plants, endophyte inoculum potential of closely adjacent soil declined by 50 % during the growing season. However, the effect was short lived and infectivity rose the following year to equal that of undisturbed adjacent woodland as mycelial matrices supported by neighbouring epacrids invaded the depleted study areas.