Recent neutral landscape models derived from percolation theory predict that a landscape with sparse but contagious habitat coverage is functionally equivalent to one with more abundant but randomly situated habitat patches. We tested this prediction in a field experiment that determined how habitat-patch abundance and configuration affect landscape use by animals. Using a 2 x 2 factorial design in a 25-m2 landscape, we created four treatments by varying the ratio of habitat (grass) to non-habitat (sand) patches (10%:90% vs 20%:80%) and the clustering of grass habitat patches (random vs contagious). We then followed the movements of tenebrionid beetles (Coleoptera: Tenebrionidae) through these experimental landscapes and measured their pathway characteristics. Our results were largely consistent with neutral-model predictions, in that the amount of habitat present had a greater influence than did habitat configuration and habitat abundance exerted its strongest influence on movement behaviours when habitat was sparse, regardless of spatial pattern. However, we also detected interactions between the amount and spatial arrangement of habitat patches that reflected how beetles responded to certain landscape-level properties of different abundances and configurations of habitat: average habitat-patch size and the distance between patches were better predictors of beetle responses than were the average number of patches or the amount of patch edge (perimeter) within a treatment. Because of such latent influences of landscape-level habitat properties, our results have important implications for conservation efforts that endeavour to preserve landscape function by maintaining some minimum amount of habitat coverage. Interactions between habitat abundance and configuration complicate the detection and definition of landscape function, illustrating the need for a spatially realistic approach in habitat management.