Most ecologists are comfortable with the notion of habitats as recognizable entities and also with situations where the junction between two adjacent habitats forms a discrete edge. Such edges form naturally because of sharp changes in important edaphic, geomorphological, climatic or chemical properties to which plants, in particular, respond. Less clear is the effect of such edges on assemblages of mobile organisms, especially invertebrates that operate at relatively small spatial scales. The objective of the present study was to sample invertebrate composition across a natural edge between a well-developed riparian habitat on fluvial sands and a saltbush habitat developed on a stony gibber plain in a semi-arid region of New South Wales, Australia. A total of 150 pitfall traps on five 1-km-long transects that straddled the edge produced more than 13 000 adult specimens from 21 ordinal invertebrate taxa. A total of 10 446 beetle, ant, wasp, fly and springtail specimens were further sorted into 426 morphospecies. Comparisons and estimates of trends in abundance and richness were made, along with computation of multivariate dissimilarity and permutation statistics, to determine if the land system edge was coincident with changes in invertebrate abundance and composition. These analyses were unable to detect disjunctions in diversity coincident with the edge. The data suggest that many taxa are either present consistently in both habitats or are mostly found in one habitat but 'leak' several hundred metres across into the other. Few taxa were unique to either habitat. The result is that assemblage composition for invertebrates changes gradually over distances of up to 400 m either side of the edge and that the distance to a recognizable change in composition is taxon dependent. Even sharp habitat edges, as defined by discrete changes in soils and plants, are not edges but broad transition zones for many invertebrate taxa. There are several implications of these results, especially for landscape ecology.