Community assembly rules specify patterns of species co-occurrence and morphology dictated by interspecific competition. We collected data on the occurrence of ground-foraging ant species in 22 ombrotrophic bogs and adjacent forest plots of New England to test two general assembly rules: reduced co-occurrence of species among communities, and even spacing of body sizes of species within communities. We used null models to generate random communities unstructured by competition and evaluated patterns at regional and local spatial scales. At the regional scale, species co-occurrence in forests, but not bogs, was less than expected by chance, whereas, at the local scale, co-occurrence in both habitats was not different from random. At the regional scale, spacing of body size distributions was random (in bogs) or aggregated (in forests). At the local scale, body size patterns were weakly segregated in bogs, but random or weakly aggregated in forests. In bogs, size ratio constancy was accompanied by greater generic diversity than expected. Although assembly rules were originally developed for vertebrate communities, they successfully explained some patterns in New England ant assemblages. However, the patterns were contingent on spatial scale, and were distinctly different for bog and forest communities, despite their close proximity and the presence of many shared species in both assemblages. The harsh physical conditions of bogs may act as a habitat filter that alters community assembly rules.