Extinctions are not biologically random: certain taxa or functional/ecological groups are more extinction-prone than others. Analysis of molluscan survivorship patterns for the end-Cretaceous mass extinctions suggests that some traits that tend to confer extinction resistance during times of normal ('background') levels of extinction are ineffectual during mass extinction. For genera, high species-richness and possession of widespread individual species imparted extinction-resistance during background times but not during the mass extinction, when overall distribution of the genus was an important factor. Reanalysis of Hoffman's (1986) data (Neues Jb. Geol. Palaont. Abh. 172, 219) on European bivalves, and preliminary analysis of a new northern European data set, reveals a similar change in survivorship rules, as do data scattered among other taxa and extinction events. Thus taxa and adaptations can be lost not because they were poorly adapted by the standards of the background processes that constitute the bulk of geological time, but because they lacked - or were not linked to - the organismic, species-level or clade-level traits favoured under mass-extinction conditions. Mass extinctions can break the hegemony of species-rich, well-adapted clades and thereby permit radiation of taxa that had previously been minor faunal elements; no net increase in the adaptation of the biota need ensue. Although some large-scale evolutionary trends transcend mass extinctions, post-extinction evolutionary pathways are often channelled in directions not predictable from evolutionary patterns during background times.