In theory, the rate of interpatch dispersal significantly influences the population dynamics of predators and their prey, yet there are relatively few field experiments that provide a strong link between these two processes. In tallgrass prairies of North America, the planthopper, Prokelisia crocea, and its specialist parasitoid, Anagrus columbi, exist among discrete host-plant patches (prairie cordgrass, Spartina pectinata). In many areas, the matrix, or habitat between patches, has become dominated by the invasive exotic grass, smooth brome (Bromus inermis). We performed a landscape-level field study in which replicate cordgrass networks (identical in number, size, quality, and distribution of cordgrass patches) were embedded in a matrix composed of either mudflat (a native matrix habitat) or smooth brome. Mark–recapture experiments with the planthopper and parasitoid revealed that the rate of movement among cordgrass patches for both species was 3–11 times higher in smooth brome than in mudflat. Within three generations, planthopper and parasitoid densities per patch were on average ∼50% lower and spatially 50–87% more variable for patches embedded in a brome as compared to a mudflat matrix. A brome-dominated landscape also promoted extinction rates per patch that were 4–5 times higher than the rates per patch in native mudflat habitat. The effect was more acute for the parasitoid. We suggest that the differences in population dynamics between networks of patches in brome and those in mudflat were driven by underlying differences in interpatch dispersal (i.e., patch connectivity). To our knowledge, this is the first experimental study to reveal that matrix composition, in particular, the presence of an invasive plant species, affects the spatial and temporal dynamics of an herbivore and its natural enemy.