We examined the representation of shrew species within assemblages at 197 sites in the southern Appalachian Mountains, USA. Assemblages were classified according to representation of functional groups, including fossorial, small epigeal, and large epigeal. Average (±SD) species richness was 2.9±1.0 and assemblages averaged 0.8 species in the fossorial and large epigeal groups and 1.3 species in the small epigeal group. Compliance with Fox's assembly rule (Fox 1987, Evol. Ecol. 1:201–213) was evaluated by dividing sites into those likely under the rule (favored) and those unlikely under the rule. The number of favored sites was compared to expected numbers of favored sites under three null models of community development, which variously incorporated observed characteristics of species in the regional pool. Number of favored sites (175 [89%]) exceeded the number found in each of 25,000 sets of assemblages simulated (P<0.00004) using algorithms that incorporated the observed species richness of each site and the observed incidences of each species. However, the number of favored sites did not differ (P=0.1054) from the number expected when patterns of allopatry and sympatry were preserved in the null model. Thus, the tendency for the equitable representation of functional groups in shrew assemblages was primarily due to the pattern of allopatry among similar species, and we believe that the striking compliance of these sites with Fox's assembly rule is largely due to structure within the regional species pool, rather than extant competitive interactions.