Managers often face the dilemma of planning reserve networks with limited data on species' distributions; “umbrella species,” as surrogates for other co-occurring taxa, were thus proposed. Here, the relative efficiencies of “target species” representation in reserves selected using “single-species umbrellas” and “umbrella species groups” are compared, both relative to each other and to target species representation in randomly selected reserve areas. Distribution data for vertebrates and plants on islands of six Great Lakes basin archipelagos were analyzed. Reserves selected using “umbrella groups” contained more species than did those selected using “single-species umbrellas.” Random selection constrained to the same total area occupied by umbrellas typically performed as well as umbrellas of any type. Reserve systems selected at random but constrained to the same number of islands occupied by umbrellas, however, contained lower proportions of target species than did reserve systems selected using umbrellas. Where data are limited, managers may be consoled by the result that random reserve selection appears to perform at least as well as any of the traditional applications of “umbrella species.”