Key innovations have often been invoked to explain the exceptional diversification of particular groups. However, there are few convincing examples of traits that are repeatedly and consistently associated with increased diversification. The paucity of such cases may reflect the contingent nature of the diversifying effect of key traits. These contingencies can be viewed as statistical interactions between the trait and at least three kinds of factors: (1) other taxa, (2) other traits of the group itself, and (3) the physical environment. I describe tentative examples in each of these categories: (1) a dampening of the diversification of clades with image-forming eyes by groups that earlier evolved such eyes, (2) an effect of growth form (woody or herbaceous) on the diversifying effect of biotic seed dispersal in angiosperms, and (3) an effect of atmospheric CO2 level on the diversifying effect of C4 photosynthesis in monocots. These examples suggest the need for more complex analyses of the relationship between possible key traits and diversification. They also suggest that radiations may be predictable given certain circumstances, thus supporting a view of evolution as both predictable and contingent. Ironically, a certain degree of predictability may be critical to arguments for evolutionary contingency.