Extinctions of megafauna species during the Late Quaternary dramatically reduced the global diversity of mammals. There is intense debate over the causes of these extinctions, especially regarding the extent to which humans were involved. Most previous analyses of this question have focused on chronologies of extinction and on the archaeological evidence for human-megafauna interaction. Here, I take an alternative approach: comparison of the biological traits of extinct species with those of survivors. I use this to demonstrate two general features of the selectivity of Late Quaternary mammal extinctions in Australia, Eurasia, the Americas and Madagascar. First, large size was not directly related to risk of extinction; rather, species with slow reproductive rates were at high risk regardless of their body size. This finding rejects the 'blitzkrieg' model of overkill, in which extinctions were completed during brief intervals of selective hunting of large-bodied prey. Second, species that survived despite having low reproductive rates typically occurred in closed habitats and many were arboreal or nocturnal. Such traits would have reduced their exposure to direct interaction with people. Therefore, although this analysis rejects blitzkrieg as a general scenario for the mammal megafauna extinctions, it is consistent with extinctions being due to interaction with human populations.