Previous studies suggest that a network of nature reserves with maximum efficiency (obtained by selecting the minimum area such that each species is represented once) is likely to be insufficient to maintain species in the network over time. Here, we test the performance of three selection strategies which require presence–absence data, two of them previously proposed (multiple representations and selecting an increasing percentage of each species' range) and a novel one based on selecting the site where each species has exhibited a higher permanence rate in the past. Multiple representations appear to be a safer strategy than selecting a percentage of range because the former gives priority to rarer species while the latter favours the most widespread. The most effective strategy was the one based on the permanence rate, indicating that the robustness of reserve networks can be improved by adopting reserve selection procedures that integrate information about the relative value of sites. This strategy was also very efficient, suggesting that the investment made in the monitoring schemes may be compensated for by a lower cost in reserve acquisition.