A landscape consists of a mosaic of different landscape elements. The structure and composition of the landscape, as well as the changes in them, influence the distribution, abundance and dynamics of different species (Morris 1995, Wiens 1995). Surprisingly, many studies assume that the preferred habitat fragments are isolated from one another by a hostile matrix, i.e. a divided landscape (Addicott et al. 1987), even though fragments are parts of the landscape mosaic and the surrounding matrix might also be used (Andren 1994). Individuals may utilize several different landscape elements during their daily activity and, thus, experience the landscape as heterogeneously undivided (Addicott et al. 1987). Therefore, if one assumes that the species perceives the landscape as divided (Addicott et al. 1987) but the species actually uses the matrix as well, then the effect of landscape changes on population size will be misinterpreted. Thus, the degree to which a species is specialized in utilizing different landscape elements will affect the numerical population response to landscape changes. Here we present alternative models to test the effect of landscape changes on different types of species. Sometimes the landscape can be described as divided (Addicott et al. 1987). The simplest explanation for the decline in population sizes as fragments become smaller and more isolated is that small fragments can be considered as random samples from larger ones (Connor and McCoy 1979, Haila 1983). However, sometimes the decline in population size is greater than predicted from the random sample hypothesis and degree of isolation and fragment size influence population size, i.e. there are true fragmentation effects (Andren 1994, 1996). There seem to be thresholds in the proportion of preferred habitat in the landscape where the change in population size is suddenly greater than predicted from the random sample hypothesis, i.e threshold for habitat fragmentation (Andren 1996). The threshold value will depend on, for example, the area requirements and dispersal distances typical for members of each species (Lande 1987, Andren 1996, Bascompte and Sole 1996, Pagel and Payne 1996). Thus, we also include the numerical population responses due to pure habitat loss and true effects of fragmentation in the model.