Studies of extinction in the fossil record commonly involve comparisons of taxonomic extinction rates, often expressed as the percentage of taxa (e.g., families or genera) going extinct in a time interval. Such extinction rates may be influenced by factors that do not reflect the intrinsic severity of an extinction trigger. Two identical triggering events (e.g., bolide impacts, sea level changes, volcanic eruptions) could lead to different taxonomic extinction rates depending on factors specific to the time interval in which they occur, such as the susceptibility of the fauna or flora to extinction, the stability of food webs, the positions of the continents, and so on. Thus, it is possible for an extinction event with a higher taxonomic extinction rate to be caused by an intrinsically less severe trigger, compared to an event with a lower taxonomic extinction rate.
Here, we isolate the effects of taxonomic susceptibility on extinction rates. Specifically, we quantify the extent to which the taxonomic extinction rate in a substage is elevated or depressed by the vulnerability to extinction of classes extant in that substage. Using a logistic regression model, we estimate that the taxonomic susceptibility of marine fauna to extinction has generally declined through the Phanerozoic, and we adjust the observed extinction rate in each substage to estimate the intrinsic extinction severity more accurately. We find that mass extinctions do not generally occur during intervals of unusually high susceptibility, although susceptibility sometimes increases in post-extinction recovery intervals. Furthermore, the susceptibility of specific animal classes to extinction is generally similar in times of background and mass extinction, providing no evidence for differing regimes of extinction selectivity. Finally, we find an inverse correlation between extinction rate within substages and the evenness of diversity of major taxonomic groups, but further analyses indicate that low evenness itself does not cause high rates of extinction.