The five major mass extinctions identified in 1982 by Raup and Sepkoski have expanded to six, with the suggestion that the Permian-Triassic extinction was a double event. Is there a general explanation for great mass extinctions, or can they result from different triggers, or even from internal system instabilities? The two most-discussed candidates for a general extinction mechanism are impacts and flood-basalt eruptions. A compilation of evidence for impact at the times of mass extinctions shows that this cause is abundantly confirmed in the case of the Cretaceous-Tertiary extinction and the late Eocene, which is a time of minor and gradual extinction, but little or no evidence connects other major extinctions to impact. On the other hand, there is a remarkable time correlation between flood basalts and four major extinctions, but no other evidence that flood basalts cause mass extinctions. The evidence for an impact-extinction linkage is strikingly different from that for a connection between flood basalts and extinctions. Flood basalts cover larger areas than craters and their associated thick ejecta blankets, which are thus less likely to be found. Impacts distribute proxies globally at instantaneous time horizons, whereas flood-basalt events are extended in time, and no remote proxies have been recognized. Many global killing mechanisms have been proposed in the case of impacts, but few have been suggested for flood basalts. It is possible that flood basalts are triggered by impact, but it is not obvious how impacts could result from anything other than chance. The hypothesis that impacts are the general cause of mass extinctions has not received supporting evidence, but has not been falsified. The hypothesis that flood basalts are the general cause of mass extinctions is supported by evidence from timing, but is not susceptible to falsification. Other candidates for general extinction causes, especially sea-level changes and system instabilities, would require separate treatment. The question is still very much open.