Landscape ecology has been very influential in developing tools for describing both structure (e.g. the distribution and sizes of patches) and function (i.e. the flow among patches) of heterogeneous environments (Turner 1989, Turner & Gardner 1991). This approach has shown that spatial heterogeneity on a landscape level may influence many types of ecological processes (Kolasa & Pickett 1991, Wiens et al. 1993). However, it is also clear that landscape structure and function must be described from an organism-centered view (Kolasa & Pickett 1991), which invites the use of population dynamic hypotheses, and presents the challenging task of merging population ecology with landscape ecology. Standard, non-spatial, predator-prey models predict that the grazing pressure in a given area is related to primary productivity (Oksanen et al. 1981). The model assumes that the number of dynamically important trophic levels is dependent on primary productivity and, in its simplest form, it can be outlined as follows: In extremely unproductive areas (e.g. boulder-fields), plant biomass is too low to sustain mammalian herbivores. In undisturbed areas, plants will thus eventually deplete their resources and compete. In moderately productive areas (e.g.arctic and alpine heaths), plant production is high enough to sustain herbivores, albeit at low densities, lower than what is needed for efficient predators to have a positive growth rate. Uncontrolled by predation, these herbivores are predicted to exert a strong grazing pressure on the vegetation. In more productive areas (e.g. tall herb meadows), plant production is high enough to sustain both herbivores and predators. With herbivores controlled by predation, plants will experience a low grazing pressure, and competition will be an important structuring factor for the plants. According to these models, a productivity gradient from extremely barren areas to productive areas should contain a zone of strong grazing pressure at intermediate productivities. A reÂanalysis using two types of patches with different primary productivity (T. Oksanen 1990) shows that the exact predictions depend on the proportion of these two patches in the habitat. Predation pressure could be high (and thus grazing pressure low) in a patch of intermediate productivity if it is embedded in a matrix of more productive patches, and, reversely, a productive patch might have a high grazing pressure if it is embedded in a matrix of less productive patches. These predictions parallel those of the source-sink model of Pulliam (1988) where a habitat where the consumer has a high growth rate "exports" juveniles to a habitat where the consumer growth rate is lower or even negative, thus creating a higher grazing pressure in the latter habitat than would have been possible without this continuous restocking of individuals. The general conclusion from these models is that grazing pressure may vary between patches both as a consequence of differences in productivity and also because of the spatial arrangements of patches. Any comprehensive understanding of the interactions between herbivores and plants in a heterogeneous environment must thus be based on experiments and observations that explicitly take the spatial heterogeneity of the study area into account.