We tested the applicability of the “passive sampling” hypothesis and theory of island biogeography (TIB) for explaining the diversity of forest-dwelling carabid assemblages (Carabidae: Coleoptera) on 30 forested islands (0.2–980.7 ha) in Lac la Ronge and the adjacent mainland in Saskatchewan, Canada. Species richness per unit area increased with distance to mainland with diversity being highest on the most isolated islands. We detected neither a positive species-area relationship, nor significant differences in species richness among island size classes, or between islands and the mainland. Nonetheless, carabid assemblages distinctly differed on islands <1 ha in area and gradually approached the structure of mainland assemblages as island area increased. Small islands were characterised by abundant populations of small-bodied, winged species and few if any large-bodied, flightless species like Carabus taedatus Fabricius. Our findings suggest that neither the “passive sampling” hypothesis nor the theory of island biogeography adequately explain carabid beetle diversity patterns observed among islands in Lac la Ronge. Instead, we hypothesise that population processes such as higher extinction rates of large-bodied, flightless species and the associated release of smaller-bodied, flying species from intra-guild predation on small islands contribute to observed differences in the structure of carabid assemblages between islands.