The structure of ecological communities has been widely and intensively debated (Strong et al. 1984, Diamond and Case 1986, Wiens 1989). Null models, for example, have often been used to test whether the species composition of islands or habitat fragments differs from random samples drawn from large areas (see several papers in Strong et al. 1984). The random sample hypothesis states, among other things, that the species-area relationship arises because large areas sample more individuals than small areas and therefore have more species (Connor and McCoy 1979, Haila 1983). The random sample hypothesis predicts, for statistical reasons, that common species are the only to be found on small island, whereas rare species will mainly occur on large islands (Wright 1991). Thus, the probability of finding a species on an island of a certain size depends on its relative abundance in the source area. Patterson and Atmar (1986) developed a method to test if species composition on islands in an shows a nested subset pattern that was different from a random distribution of species. A completely nested subset pattern means that a species found on an island with n species will be found on all islands with n+ 1 species. They went on to describe two different simulation models. Firstly, they used a model in which individual species were drawn at random (without replacement) from a species pool (RANDOMO), such that all species have the same probability to be incorporated in the sample. Secondly, they biased the sample of species with respect to their actual frequencies of occurrence in the archipelago (RANDOM 1), i.e. affording common species a disproportionally high probability to be drawn as compared to rare species. The statistic used to test the distribution pattern is the number of unexpected absences, or holes, in the in the simulation compared with the observed number of holes. An unexpected absence of a species is when a species is found on an island with n species and is missing on an island with n + 1 species, i.e. the number of holes measures the deviation from a completely nested subset pattern. The nested subset pattern has been used to study the composition in granivorous assemblages (Patterson and Brown 1991), the temporal development of species composition (Patterson 1990), as well as the implications of nested subset patterns in conservation biology (Patterson 1987). I used two published studies with suitable raw data, to test whether the bird community in habitat patches in the boreal part of Scandinavia might display a nested subset pattern or not. The first study concerns bird communities on raised peat-bogs in the county of Varmland in Sweden (county S in Bostrom and Nilsson 1983), i.e. a naturally patchy habitat. These authors concluded, based on population density differences, that the bird community on small bogs was not a random sample of the communities on large bogs. The nestedness was tested using the Monte Carlo simulation RANDOM1, devised by Patterson and Atmar (1986). The bird communities on bogs did not show a significant nestedness. The observed number of holes was 645, while the simulated mean number of holes was 694.6 ? 40.2 SD, i.e. the observed value lies 1.23 standard deviations below the simulated (p = 0.11). However, two species that were only found on two small bogs, one each, made up about 15% of all the holes. One of the species (rustic bunting, Emberiza rustic) has its normal distribution range further north and is very rare in the part of Sweden where Bostrom and Nilsson (1983) collected these data. The other species (white wagtail, Motacilla alba) is not really a bird belonging to the bird community of bogs. If these two species were excluded, then the distribution of birds showed a significant nestedness. The observed number of holes (547) was 2.51 standard deviations (p = 0.006) smaller than the simulated mean number (654.8 ? 43.0 SD). Thus, the test of nested pattern seems to be very sensitive to the occurrence of a few rare species on small bogs.