Landscape simplification has been clearly demonstrated to have negative impacts on the in-crop density and biological-control activity of natural enemies in agricultural landscapes. The role of spatial arrangement of the landscape, however, has not been investigated in agroecosystems. We applied cost–distance modeling to investigate the relationship between the in-crop density of natural enemies and the structural connectivity of non-crop land uses surrounding crops within Australian cotton landscapes. We further compared the explanatory power of this approach with the more commonly used spatially specific proportional-area approach, which considers landscape composition in terms of the proportional area of a given land use within a given radius. Cost–distance metrics offered a more significant explanation of incrop density for the predatory beetle Dicranolaius bellulus (Coleoptera: Melyridae) than did the proportional-area approach. The in-crop density for this species was positively and significantly correlated with the connectivity of wooded land uses within a 3000 m radius. However, for natural enemy taxa that responded to landscape characteristics at smaller spatial scales (within a 750 m radius), namely Oxyopes spp. (Araneae: Oxyopidae) and Trichogramma spp., (Hymenoptera: Trichogrammatidae), the proportional-area approach gave a more significant explanation of in-crop density. Herbivore taxa responded weakly to proportional area at all scales and showed no correlation to cost–distance metrics. Findings indicate potential for simplified agricultural landscapes to be 'selectively' manipulated to enhance colonization of the crop by natural enemies, but not herbivores, by improving connectivity between crops and non-crop resources, through the presence of woody vegetation.