Animal interpatch movement and spatial distribution are known to be influenced substantially by the composition of the landscape matrix, but little is known about the underlying mechanisms. In previous mark–recapture experiments we have found that the rates of emigration and immigration for the planthopper Prokelisia crocea are greater within a matrix composed of the introduced grass smooth brome (Bromus inermis) than a mudflat matrix. Additionally, census data indicated that individuals aggregate near the edge of host-plant patches (prairie cordgrass; Spartina pectinata) bordered by mudflat, but not in patches bordered by nonhost grasses such as brome. Here, we investigate the mechanistic basis of these matrix effects by tracking the individual movements of planthoppers released at the edge of brome- and mudflat-bordered cordgrass patches, and within homogeneous habitats of each type (cordgrass, brome, and mudflat). We found that patch edges bordered by brome were three times more permeable to emigration than mudflat-bordered edges. Also, planthoppers exhibited no tendency to avoid edges by moving away (i.e. towards the patch interior). Within homogeneous habitats, comparison of the fractal dimension of movement paths revealed that movement was more linear in mudflat than in brome or cordgrass. In addition, planthoppers exhibited greater step lengths (distance moved per 10-min interval), shorter residency times (duration of pauses between movements), and greater rates of net linear displacement in mudflat than brome and cordgrass. We attribute the planthopper's distributional patterns within patches to the lower permeability of mudflat than nonhost grass edges and the absence of edge–avoidance behavior. Contrary to conventional wisdom that low-resistance matrix types (e.g. those that promote high displacement rates) enhance interpatch dispersal rates, dispersal success may be higher in brome matrix because tortuous movement through this matrix increases the planthopper's rate of encounter with cordgrass patches.