A mean field, metapopulation model of a predator-prey interaction is developed in order to understand the consequences of habitate destruction at different trophic levels. Such a model allows is to explore different ecological scenarios (donor control vs. top-down control) by changing a single parameter. The response to habitat destruction is qualitatively the same for both predator and prey, although there are interesting differences linked to the trophic position. A similar decrease in the colonization rates affects the two species quite differently. Predators diminish faster than prey, and furthermore, the fraction of occupied sites decreases more sharply as colonization rates are lowered, i.e. there is a nonlinear relationship between regional abundance and colonization rate. There is a well-defined threshold in the colonization rate below which the predator becomes extinct. Thus, dispersal rate is critical for predicting the consequences of habitat destruction. Finally, these results are compared with the behavior of a spatially explicit simulation. The only difference between the analytical model and the simulation is that colonization is no longer a global phenomenon in the latter, but it is a local process, constrained to the nearest patches. The bulk of the results are similar to the mean field behavior, and we comment on some differences related with non-homogeneity and real space. Some general implications for conservation biology and biological control are outlined. Copyright 1998 Academic Press