It is conventional to divide extinctions into two distinct kinds: background and mass extinction. The term "mass extinction" is most commonly reserved for the so-called "Big Five" events: short intervals in which 75-95% of existing species were eliminated (Table 1). The K-T event, mentioned earlier, is one of the Big Five, but not the largest. Although the Big Five were important events, their combined species kill amounted to only about 4% of all extinctions in the past 600 Myr (11). The mass/background dichotomy is unfortunate because it implies two modes of extinction, yet there is no evidence for a discontinuity between them. Fig. 2 shows variation in percent species kill in 1-Myr intervals for the past 600 Myr. The events called mass extinctions are concentrated in the right-hand tail, but there is no break between this tail and the main distribution. The data appear to produce a single, highly skewed distribution. Thus, segregating mass extinction from background has no more meaning than distinguishing hurricanes from other tropical cyclonic storms on the basis of some arbitrary wind speed (64-knot sustained surface winds). Continued use of the mass/background dichotomy serves only to hide interesting structure in extinction data.
Fig. 3 shows a cumulative distribution of extinction frequency, the so-called "kill curve" for species of the past 600 Myr. The format is one used commonly to analyze severe storms, floods, earthquakes, and other natural phenomena where the larger the event, the rarer it is. The kill curve gives the average time interval (mean waiting time) between an extinction event and the next one of equal or greater magnitude. An "event" is defined as the species kill occurring in an arbitrarily short interval. Thus, 10% (or more) of the standing crop of species goes extinct, on average, every 1 Myr, 301% every 10 Myr, and 65% every 100 Myr (9). The 100-Myr events include the Big Five mass extinctions.
Without further analysis, one could assert that the kill curve is a natural result ofchance coincidence ofindependent events. That is, pure chance might produce an episode of nearly simultaneous extinctions if we wait long enough. This is emphatically not the case. For a random model that assumes that all species extinctions are independent of one other, the probability of a 10%o extinction every 1 Myr (on average) is vanishingly small (9). In fact, a kill curve based on this model could not be plotted at the scale of Fig. 3: the curve would be indistinguishable from the horizontal axis. The only available conclusion is that extinctions are nonrandomly clustered in time, and this implies strongly that the K-T extinctions, for example, had a common cause.
Some of the clustering of extinction may be due to the removal of one or a few species that are crucial to the existence of other species. Or clusters may be due to destruction of one important ecosystem or habitat. However, for the larger events, at least, the extinctions are far more pervasive. At the end of the Cretaceous, high levels of species extinction (>50%) are found in all geographic areas and involve organisms as different as burrowing mollusks, planktonic microorganisms, land plants, and dinosaurs. This suggests, among other things, that the big mass extinctions cannot be explained by Darwin's species interactions unless one is willing to postulate an incredible degree of connectedness in the biosphere.
A striking effect of the typical mass extinction is its aftermath. For as long as 5-10 Myr, fossil faunas and floras are impoverished and are often dominated by only one or two species. The longest such interval followed the late Permian extinction (the largest of the Big Five): many major phyla and classes, known to have survived from later occurrences, are absent from the early Triassic assemblages. And about a third of the Triassic is characterized by what has been called the ".coal gap," an interval where no coal deposits have been found-either of temperate or of tropical origin (A. M. Ziegler, personal communication).
When full diversity does return, it often has a strikingly different character. A classic example is the history of marine reefs. Reef communities have been wiped out several times in the past 600 Myr, coinciding in four cases with Big Five events. Each time reefs reappear, the principal framework organisms have changed, switching back and forth between calcareous algae, sponges, bryozoans, rudist mollusks, and various corals (13, 14). The contemporary term "coral reef' describes only the current occupants of that adaptive zone.