The temporal distribution of the major extinctions over the past 250 million years has been investigated statistically using various forms of time series analysis. The analyzed record is based on variation in extinction intensity for fossil families of marine vertebrates, invertebrates, and protozoans and contains 12 extinction events. The 12 events show a statistically significant periodicity (P < 0.01) with a mean interval between events of 26 million years. Two of the events coincide with extinctions that have been previously linked to meteorite impacts (terminal Cretaceous and Late Eocene). Although the causes of the periodicity are unknown, it is possible that they are related to extraterrestrial forces (solar, solar system, or galactic).Virtually all species of animals and plants that have ever lived are now extinct, and the known fossil record documents some 200,000 such extinctions. It has been generally assumed that extinction is a continuous process in the sense that species are always at risk and that mass extinctions simply reflect relatively short-term increases in that risk. Following this view, the extinction process is often described mathematically as a time homogeneous process using standard birth-death models (1-3). There is increasing evidence, however, that many extinctions are actually short-lived events of special stress, separated by periods of much lower, or even negligible, risk. Fischer and Arthur (4) departed from convention by arguing that the major extinction events of the past 250 million years (ma) occurred periodically at nearly constant intervals of 32 ma (see also ref. 5). Their study used a limited data base, and no statistical testing was done. The purpose of this paper, therefore, is to test the proposition of periodicity in the record of marine extinctions over the past 250 ma (Late Permian to Recent) by using as rigorous a methodology as present data permit.
DATA BASEThe data for this study come from Sepkoski's compilation (6) of the temporal ranges of -3,500 families of marine animals (vertebrate, invertebrate, and protozoan), but the subset of data and method of expressing extinction intensity differ from our previous analyses of these data (7,8). For the present study, a culled subset of the total sample was used: all families with low-resolution ranges, not known to the level of the stratigraphic stage, were eliminated, as were families noted by Sepkoski (6) as having questionable taxonomic or stratigraphic designations. In addition, families still living today were ignored in order to avoid the damping effect of the "Pull of the Recent" (9). This culling process reduced the sample substantially (567 for the Late Permian to Recent) but in so doing removed much of the noise that characterizes data sets of this kind.The Late Permian to Recent interval analyzed in this study contains several well-documented mass extinctions, has a comparatively accurate time scale, and is divided into relatively short stratigraphic stages. The interval comprises 39 international s...