Familiar quantitative reserve-selection techniques are tailored to simple decision problems, where the representation of species is sought at minimum cost. However, conservationists have begun to ask whether representing species in reserve networks is sufficient to avoid local extinctions within selected areas. An attractive, but previously untested idea is to model current species' probabilities of occurrence as an estimate of local persistence in the near future. Using distribution data for passerine birds in Great Britain, we show that (i) species' probabilities of occurrence are negatively related to local probabilities of extinction, at least when a particular 20-year period is considered, and (ii) local extinctions can be reduced if areas are selected to maximize current species' probabilities of occurrence We suggest that more extinctions could be avoided if even a simple treatment of persistence were to be incorporated within reserve selection methods.