Résumé
In angiosperms, dioecious clades tend to have fewer species than their nondioecious sister clades. This departure from the expected equal species richness in the standard sister clade test has been interpreted as implying that dioecious clades diversify less and has initiated a series of studies suggesting that dioecy might be an 'evolutionary dead end'. However, two of us recently showed that the 'equal species richness' null hypothesis is not valid in the case of derived char acters, such as dioecy, and proposed a new test for sister clade comparisons; preliminary results, using a data set available in the litterature, indicated that dioecious clades migth diversify more than expected. However, it is crucial for this new test to distinguish between ancestral and derived cases of dioecy, a criterion that was not taken into account in the available data set. Here, we present a new data set that was obtained by searching the phylogenetic literature on more than 600 completely dioecious angiosperm genera and identifying 115 sister clade pairs for which dioecy is likely to be derived (including > 50% of the dioecious species). Applying the new sister clade test to this new dataset, we confirm the preliminary result that dioecy is associated with an increased diversification rate, a result that does not support the idea that dioecy is an evolutionary dead end in angiosperms. The traits usually associated with dioecy, that is, an arborescent growth form, abiotic pollination, fleshy fruits or a tropical distribution, do not influence the diversification rate. Rather than a low diversification rate, the observed species richness patterns of dioecious clades seem to be better explained by a low transition rate to dioecy and frequent losses.