Does extinction risk show any patterns that might help us to understand the processes affecting biodiversity loss? The “field of bullets” scenario, in which extinction strikes completely at random, is a widely used null model for extinction (Raup et al., 1973; Van Valen, 1976). The metaphor comes from trench warfare, where soldiers’ survival may have depended more on luck than skill. This scenario predicts that threatened species should constitute a stochastically constant fraction of any sample. Mammalian extinction risk is not a simple field of bullets but shows both geographic and phylogenetic patterns (Russell et al., 1998; Mace and Balmford, 2000; Baillie et al., 2004). The prevalence of risk is higher in the Old World than in the New World and higher on islands than on continents (Mace and Balmford, 2000; Baillie et al., 2004). It varies among clades too, being higher than average in primates and perissodactyls and lower than average in rodents (Baillie et al., 2004). Species with few close relatives are also more likely to be at risk (Russell et al., 1998; Purvis et al., 2000a).
These patterns reject the original field-of-bullets model, but the model lacks a geographic dimension because there may have been fields that were near to the battle but that nevertheless had no bullets. Likewise, human pressures have changed some places beyond recognition but left others almost untouched. Because closely related species often live in the same region, geographic heterogeneity in threat intensity could, by itself, cause taxonomic selectivity in extinction risk (Russell et al., 1998). Alternatively, the selectivity could arise because biological differences among clades affect species’ abilities to withstand threats (McKinney, 1997). How important for species’ survival is staying out of the firing line, and how important is being bulletproof?
Human population density predicts proportions of threatened mammal species among continental countries (McKinney, 2001), supporting the “firing-line” model. However, an analysis of extinction risk within terrestrial World Wildlife Fund (WWF) ecoregions shows that phylogenetic nonrandomness is common within single ecoregions, where pressures may be more even than across the globe. Within each ecoregion, we generated phylogenetically independent contrasts (essentially, differences between sister clades) (Harvey and Pagel, 1991) in extinction risk (0 for LC species, 1 for species having a higher status; data-deficient and unevaluated species were excluded) on the phylogeny (with polytomies resolved arbitrarily and branch lengths set to unity) and compared the sum of the absolute values of standardized contrasts with the sums obtained from 1,000 randomizations of the risk data among the ecoregion’s species. If high-risk species are strongly clumped in the phylogeny, the observed sum will be lower than in 95% of the simulations. Of 691 ecoregions with at least three higher-risk species and some variance in extinction risk, 386 (54%) showed significant clumping. Interestingly, the strength (rather