Dioecy, Monoecy, and Their Ecological Correlates in the Littoral Forest of Madagascar Laura B. Vary 1,6 , Daniel L. Gillen 2 , Miramasonadro Randrianjanahary 3 , Porter P. Lowry II 4,5 , Ann K. Sakai 1 , and Stephen G. Weller 1 1 Department of Ecology and Evolutionary Biology, University of California-Irvine, Irvine, California 92697, U.S.A. 2 Department of Statistics, University of California-Irvine, Irvine, California 92697, U.S.A. 3 Missouri Botanical Garden—Madagascar Office, BP 3391, Antananarivo 101, Madagascar 4 Missouri Botanical Garden, PO Box 299, St. Louis, Missouri 63166-0299, U.S.A. 5 Muse ´ um National d’Histoire Naturelle, De ´ partement Syste ´ matique et Evolution (UMR 7205), C.P. 39, 57 rue Cuvier, 75231 Paris Cedex 05, France ABSTRACT Dioecy is a rare breeding system in flowering plants, but one that has evolved multiple times in different plant lineages. Dioecious species are commonly associated with several ecological traits, including woody habit, fleshy fruit, and small, inconspicuous flowers, although the significance of these correlations has been debated ex- tensively. Monoecy is a breeding system that may lead to the evolution of dioecy, but ecological correlates of monoecious species have rarely been analyzed. We determined the diversity of breeding systems in the littoral forests of Madagascar and used multivariate methods to estimate which ecological traits have the strongest association with dioecy and are the best predictors of breeding systems. The Malagasy littoral forest flora is a well-documented subset of the Malagasy flora, with 13 percent of total species diversity in an area o 1 percent of the island’s total. We found high levels of dioecy (18.4%) and monoecy (9.2%), similar to incidences in other tropical floras. Using multinomial logistic regression, dioecy has the strongest association with woody habit and fleshy fruit. Monoecious species have a strong association with small flowers, although this association does not hold at higher taxonomic levels. Using classification and regression tree (CART) methods, the best predictors of dioecy are woody habit and fleshy fruit; monoecy is equally predicted by fleshy and dry fruit. For the Malagasy littoral forest, both methods provide further support for the importance of woody habit and fleshy fruit in the evolution of dioecy. Foreign language abstract is available in the online version of this article. Key words: breeding systems; classification and regression trees (CART); fleshy fruits; multinomial logistic regression; small flowers; woody habit. DIOECY AND MONOECY ARE RARE IN FLOWERING PLANTS ( 6% dioe- cious,  4% monoecious angiosperms worldwide, Renner & Rick- lefs 1995), although they have evolved multiple times in different plant lineages (Yampolsky & Yampolsky 1922, Renner & Ricklefs 1995). Over the past 30 yr, surveys of floras around the world have shown significant associations of dioecy with many ecological traits, including woody habit, fleshy fruit, animal-dispersed fruit, unspe- cialized pollination syndromes, small, inconspicuous flowers, climbing growth habit, dry habitats, tropical climates, and island habitats (reviewed in Sakai & Weller 1999). In comparison, few floristic studies have examined monoecy and its associated ecolog- ical traits, with dry fruit as its most common association (Flores & Schemske 1984, Gross 2005, Chen & Li 2008). Although these traits are statistically associated with dioecy or monoecy, many of the explanations for observed associations have been developed post- hoc and are difficult to test (Sakai & Weller 1999). In some cases, the ecological traits themselves may be correlated, and ecological traits may only be indirectly associated with dioecy because of cor- relations among other traits (e.g., fleshy fruits are much more likely to be found on woody than herbaceous species; Bawa 1980, Givnish 1980, Muenchow 1987). Despite the difficulties in using statistical associations to un- derstand the relationship between dioecy and ecological traits, the distribution of dioecy in floras has been used to develop hypotheses for the association of ecological traits with dioecy. Dioecy is distrib- uted unevenly around the world, with low levels in temperate con- tinental floras (California: 2.8%; Alaska: 3.9%, Fox 1985; Chile: 6.1%, Arroyo & Squeo 1990) and relatively high levels in tropical, island floras (New Zealand: 13%, Godley 1979; Hawaii: 14.7%, Sakai et al. 1995a; New Caledonia: 16%, Bawa 1992). Dioecious species may be favored in insular, island habitats due to the genetic advantages of obligate outcrossing but also the ecological advantages of associations with unspecialized pollination syndromes and peren- nial habit (Carlquist 1966, Baker 1967, Bawa 1980, Thomson & Barrett 1981). In addition, dioecious species may be more common in tropical forest habitats given the association of dioecy with fleshy, bird-dispersed fruit, a potential advantage for understory trees where wind-dispersal is not effective (Carlquist 1974, Bawa 1980, Givnish 1980, Baker & Cox 1984, Muenchow 1987). The geographic distribution and ecological correlates of monoecy are not as well described as those for dioecy. Like dioecy, monoecy is also more common in tropical floras, but only a few studies have suggested hypotheses for its high prevalence in these areas (Table 1; Flores & Schemske 1984, Sakai et al. 1995b, Gross 2005, Chen & Li 2008). In the Puerto Rican, Australian, and Chi- nese floras, monoecious species are associated with dry fruit (Flores Received 29 December 2009; revision accepted 24 October 2010. 6 Corresponding author; e-mail: lvary@uci.edu BIOTROPICA 43(5): 582–590 2011 10.1111/j.1744-7429.2010.00742.x 582 r 2011 The Author(s) Journal compilation r 2011 by The Association for Tropical Biology and Conservation