(p. 337), Darwin expressed confidence that if ancient species could be re-created today and put in competition with their modern counterparts, the old species would be "beaten and exterminated." This is definitely not the current view, and major research programs are now being devoted to determining the extent, if any, of selectivity in past extinctions. A common (though by no means proven) view is that the victims of extinction are in no way different from the survivors, except for the fact of their extinction. Simpson (1944) was clearly moving in this direction when he suggested in Tempo and Mode that the mammals were the lucky recipients of space vacated by the dinosaurs.
Taxonomic Selectivity. Much of current extinction research attempts to identify taxonomic selectivity. Do some taxonomic groups suffer significantly more species extinction in an extinction episode than other groups? These studies can take advantage of the availability of taxonomic data bases, such as those compiled by Sepkoski for marine genera (Sepkoski, 1989) and families (Sepkoski, 1992), and thus have the benefit of large samples. The approach carries the tacit assumption that genealogical relatedness implies similarity of physiology, ecology, or other attributes that determine susceptibility to extinction.
Taxonomic selectivity has been documented, but the effect is generally quite small and requires massive samples for confirmation. For example, when extinction rates for several taxonomic groups are compared with the mean for all groups, about 10% differ from the mean at a 0.05 significance level, whereas 5% would be expected by chance. Similarly, about 2% of the tests are significant at the 0.01 level. Thus, taxonomic selectivity is present but minor.
Occasionally, pronounced taxonomic selectivity has been found. The dinosaur extinction is such a case. In the latest Cretaceous of western North America, Clemens (1986) tabulated 117 genera of fossil mammals, amphibians, reptiles, and fish; 50 of these (43%) died out at or near the end of the Cretaceous, including all 22 dinosaur genera. The null hypothesis that all genera shared the same probability of extinction (0.43) can be rejected easily, and this demonstrates a clear bias against dinosaur survival. But such cases are relatively rare. In the same data set, for example, 8 of 24 mammalian genera died out (33%), but because of the small sample size, it is impossible to demonstrate that this extinction rate is significantly lower than the mean for all groups.
Small sample sizes have plagued many studies of selectivity, giving rise to generalizations that are widely accepted but not supportable statistically. For example, it is often claimed that the amphibians survived the K-T event with little difficulty. In the Clemens data set, only one-third of the amphibian genera went extinct (equal to the