Mass extinctions as statistical phenomena: an examination of the evidence using chi-square tests and bootstrapping


Although much natural historical evidence has been adduced in support of the occurrence of several mass extinctions during the Phanerozoic, unambiguous statistical confirmation of the mass extinction phenomenon has remained elusive. Using bootstrapping techniques that have not previously been applied to the study of mass extinction, we have amassed strong or very strong statistical evidence for mass extinctions (see text for definitions) during the Late Ordovician, Late Permian, and Late Cretaceous. Bootstrapping therefore verifies three of the mass extinction events that were proposed by Raup and Sepkoski (1982). A small amount of bootstrapping evidence is also presented for mass extinctions in the Induan (Triassic) and Coniacean (Cretaceous) Stages, but high overall turnover rates (including high origination) in the Induan and uncertain estimates of the temporal duration of the Coniacean force us to conclude that the evidence is not compelling.

We also present the results of more liberal X2 tests of the differences between expected and observed numbers of familial extinctions for stratigraphic stages. In addition to verifying the mass extinctions identified using bootstrapping, these analyses suggest that several stages that could not be verified as mass extinction stages using bootstrapping (including the last three in the Devonian, and the Norian Stage of the Triassic) should still be regarded as candidates for mass extinction. Further analysis will be required to test these stages in more detail.

Hubbard, Alan E.
Gilinsky, Norman L.
Publication date: 
January 1, 1992
Publication type: 
Journal Article