Island biogeography theory, and the 50/500 rule of genetics, have effectively devalued small habitat fragments for species conservation. Metapopulation theory has given new value to small remnants but data on species persistence are scarce. This study examined the capacity of very small and sheep-grazed remnants of eucalypt woodland in agricultural Western Australia to support remnant-dependent terrestrial arthropods. We surveyed 53 sheep-grazed remnants of wheatbelt wandoo Eucalyptus capillosa for the presence of four species of arthropod with different dispersal strategies (terrestrial versus aerial) and diet (predaceous vs. herbivorous): the harvester and mound-building termite Drepanotermes tamminensis, the wood-eating and mound-building termite Amitermes obeuntis, the predaceous and burrowing scorpion Urodacus armatus and the predaceous ‘bull’ ant Myrmecia nigriceps. All species with the exception of the scorpion disperse aerially, and all construct above-ground structures that are easily recognized. Remnants ranged in size from 50 m2 to 21 000 m2 (mean 1791 m2), in spatial isolation (distance to the nearest vegetation remnant) from 10 m to 500 m (mean 123 m) and in a length-to-width ratio (shape) from circular (mean ratio 1.0) to linear (mean ratio 4.0). Observations in small and grazed remnants were compared with observations made in six wandoo woodland sites within a large (1040 ha) and ungrazed remnant. The total number of target species was highly correlated with remnant area (r = 0.68). Remnant isolation and remnant shape had no apparent influence on the total number of target species. The minimum area of grazed remnants in which individual species were recorded followed the large predator Urodacus armatus (4515 m2) > smaller predator Myrmecia nigriceps (300 m2) > harvester termites Drepanotermes tamminensis (102 m2) > wood-eating termites Amitermes obeuntis (50 m2). With the exception of U. armatus which occurred only in three of the four largest grazed remnants, the occurrence of all other species increased from small to large grazed remnants, suggesting a remnant-size effect for all species. Remnant isolation or remnant shape had no apparent influence on the occurrence of any one species. The terrestrially dispersing scorpion persisted in remnants despite their isolation from other remnants from 200 m to 500 m. For both termite species, mound heights were significantly greater in large, ungrazed woodlands than in small and grazed woodlands. The incidence of mound abandonment in smaller and grazed remnants was considerably higher for harvester than for wood-eating termite colonies. This suggests differences in spatial requirements and possibly diet-related susceptibilities to fluctuations in food availability. The diameter of Myrmecia nigriceps nests showed no relationship with remnant size or isolation. This study demonstrated that even very small remnant woodlands on farms may play an important role in sustaining small native animals, either as stepping-stones for dispersing individuals (termites, ants) or in providing adequate habitat to sustain populations for longer periods (all four species).