Volume 101, Issue 10 p. 1588-1596
AJB Centennial Review
Free Access

The relative and absolute frequencies of angiosperm sexual systems: Dioecy, monoecy, gynodioecy, and an updated online database

Susanne S. Renner

Corresponding Author

Susanne S. Renner

Systematic Botany and Mycology, University of Munich, 80638 Munich, Germany

E-mail for correspondence: [email protected]Search for more papers by this author
First published: 01 October 2014
Citations: 450

Freely available online through the AJB open access option.

For information on specific groups, I thank A. Antonelli, T.-L. Ashman, S. Barrett, V. Bittrich, D. Charlesworth, L. Delph, K. Burt-Utley, M. Gottschling, M. Hughes, S. Kubota, C. Persson, and D. Thomas.


Premise of the study: Separating sexual function between different individuals carries risks, especially for sedentary organisms. Nevertheless, many land plants have unisexual gametophytes or sporophytes. This study brings together data and theoretical insights from research over the past 20 yr on the occurrence and frequency of plant sexual systems, focusing on the flowering plants.

Methods: A list of genera with dioecious species, along with other information, is made available (http://www.umsl.edu/∼renners/). Frequencies of other sexual systems are tabulated, and data on the genetic regulation, ecological context, and theoretical benefits of dioecy reviewed.

Key results: There are 15600 dioecious angiosperms in 987 genera and 175 families, or 5–6% of the total species (7% of genera, 43% of families), with somewhere between 871 to 5000 independent origins of dioecy. Some 43% of all dioecious angiosperms are in just 34 entirely dioecious clades, arguing against a consistent negative influence of dioecy on diversification. About 31.6% of the dioecious species are wind-pollinated, compared with 5.5–6.4% of nondioecious angiosperms. Also, 1.4% of all angiosperm genera contain dioecious and monoecious species, while 0.4% contain dioecious and gynodioecious species. All remaining angiosperm sexual systems are rare. Chromosomal sex determination is known from 40 species; environmentally modulated sex allocation is common. Few phylogenetic studies have focused on the evolution of dioecy.

Conclusions: The current focus is on the genetic mechanisms underlying unisexual flowers and individuals. Mixed strategies of sexual and vegetative dispersal, together with plants’ sedentary life style, may often favor polygamous systems in which sexually inconstant individuals can persist. Nevertheless, there are huge entirely dioecious clades of tropical woody plants.

“Dioecious plants, however fertilised, have a great advantage over other plants in their cross-fertilisation being assured. But this advantage is gained … with some risk … of their fertilisation occasionally failing. Half the individuals, moreover, namely, the males, produce no seed, and this might possibly be a disadvantage… dioecious plants cannot spread so easily as monoecious and hermaphrodite species, for a single individual, which happened to reach some new site, could not propagate its kind … Monoecious plants also can hardly fail to be to a large extent dioecious in function, owing to the lightness of their pollen and to the wind blowing laterally, with the great additional advantage of occasionally or often producing some self-fertilised seeds. When they are also dichogamous, they are necessarily dioecious in function. Lastly, hermaphrodite plants can generally produce at least some self-fertilised seeds … When their structure absolutely prevents self-fertilisation, they are in the same relative position to one another as monoecious and dioecious plants [except] that every flower is capable of yielding seeds.”

These lines from Darwin (1876, p. 414) express why the evolution of dioecy in flowering plants is so difficult to understand. As he points out, sexual specialization entails genetic, demographic, and ecological costs, and its benefits in terms of outcrossing can instead be achieved by alternative physical, temporal, and genetic mechanisms that help prevent self-pollination, self-fertilization or both (Lloyd, 1972, 1974, 1980, 1982, 1984; Charlesworth and Charlesworth, 1978; Thomson and Barrett, 1981; Ross, 1982; Charlesworth, 1985).

The evolution of dioecy has long intrigued evolutionary biologists. Over the past 40 yr, David Lloyd's (1937–2006) empirical and conceptual contributions (Lloyd, 1972, 1974, 1975a, b, c, 1980, 1981, 1982) and Brian and Deborah Charlesworth's (Charlesworth and Charlesworth, 1978) model for the evolution of dioecy (cited 621 times; Web of Science, 28 April 2014) have largely shaped the field. Lloyd's approach stressed the ecological context of plant mating strategies (Barrett and Charlesworth, 2007). Among his legacies is the categorization of seed plant sexual systems into two functional classes, depending on whether all individuals of a population are monomorphic in sexual behavior or instead dimorphic. This grouping by “function” formally recognized the interplay between a plant's own sex allocation (its investment in male or female functions) and the sex allocation of other individuals in the population (Lloyd, 1980; Pannell, 2005). The Charlesworths’ model assumes two mutations affecting male and female fertility in a closely linked region, followed by a recombination suppressor (possibly an inversion) that would take populations of hermaphrodites to chromosomally determined dioecy via a gynodioecious intermediate phase. This model has proven extremely fruitful for the field of plant sex chromosome evolution (Ming et al., 2011; Charlesworth, 2013).

In the prephylogenetic 1980s, many possible correlations between dioecy and various eco-morphological traits were studied to infer conditions that would favor dioecy (Renner and Ricklefs, 1995 and work cited therein). These broad-scale efforts drew to a close around 2003 (Vamosi et al., 2003) with the increasing availability of molecular phylogenies and sophisticated methods to study the evolution of correlated traits (D. R. Maddison, 1994; W. P. Maddison, 1995; Pagel, 1994, 1999; Pagel et al., 2004). The hope arose that having a phylogenetic framework would clarify transitions to dioecy, and the conditions favoring or delaying them, because in theory a comparative approach can inform us about the sequence in which traits have evolved. However, traits with frequent transitions, such as the sexual systems in many plant clades, can only be inferred on densely sampled, well-resolved phylogenies with reliable trait scoring (Salisbury and Kim, 2001), something rarely possible because of scarcity of both suitable plant material and field observations on sexual systems in natural populations (of many species in a clade). While it is a truism that sister species are not each other's ancestors, the misinterpretation of phylogenies as supporting “pathways” from one tip state to another is still extremely common (Losos, 2011).

For the present study, I revised and updated a list of angiosperm genera with dioecious species and added information on their “presence” in the NCBI GenBank, which allowed me to find phylogenetic work focusing on the evolution of dioecy. This review contains the first modern data (since Yampolsky and Yampolsky, 1922) on the frequencies of both gynodioecy and monoecy in genera with dioecy as well as the relative, or in some cases absolute, frequencies of other sexual strategies. I briefly review work on genetic sex determination and insights from theoretical models, and I start with separate sexedness in non-angiosperm land plants because knowing the sexual systems of the surviving outgroups to the flowering plants helps understand what is special about flowering plant mating strategies.


Literature review and GenBank data

Over the past years, I used a combination of web-based literature searches and personal communications to update a previous compilation of dioecious genera (Renner and Ricklefs, 1995). The family assignments of all genera were updated in March 2014 using GenBank (http://www.ncbi.nlm.nih.gov/nuccore/), which was also used to check for phylogenetic studies on the evolution of dioecy. Studies were downloaded from the World Wide Web or obtained from their authors and checked for information about mating systems and changes in generic circumscriptions. Experts on particular families and genera were consulted by e-mail (see acknowledgments).


Sexual specialization in land plant gametophytes and sporophytes

Figure 1 provides an overview of the frequency of combined or separate male and female function across land plants (embryophytes). All embryophytes cycle between a haploid and a diploid life stage, and there are four basic kinds of sexual systems: (1) Monoicy, in which archegonia and antheridia are produced on each gametophyte. This is traditionally considered the derived state (Wyatt and Anderson, 1984), although the data are ambiguous (Villarreal and Renner, 2013). (2) Dioicy, in which archegonia and antheridia are produced on separate gametophytes. (3) Monoecy, in which archegonia/embryo sacs and antheridia/microsporangia are produced on each sporophyte. (4) Dioecy, in which archegonia/embryo sacs and antheridia/microsporangia are produced on separate sporophytes (Wyatt and Anderson, 1984; Bateman and DiMichele, 1994). In dioicous hornworts, liverworts, or mosses, sex determination is always chromosomal, with the diploid sporophyte always heterozygous at the sex-determining locus and producing male and female spores (future gametophytes) in a 50:50 ratio. There is no opportunity for recombination of sex chromosomes in haploid-dominant plants. Bryophyte sex chromosomes are therefore fundamentally different from those of vascular plants (reviewed by Bachtrog et al., 2011). Ferns and lycophytes with their free-living haploid and diploid generations have few sexually specialized species (Wyatt and Anderson, 1984; Tanurdzic and Banks, 2004; Fig. 1). In the gymnosperms, however, dioecious sporophytes are the predominant system (Fig. 1), and sex chromosomes have evolved in several genera, such as Podocarpus (Hair and Beuzenberg, 1958; Hizume et al., 1988).

Details are in the caption following the image

Distribution of land plant sexual systems (modified from Villarreal and Renner, 2013). The terms monoicy and dioicy are used for gametophytes (Wyatt and Anderson, 1984), monoecy and dioecy for sporophytes. Triangle width is proportional to species number and absolute species numbers are shown at the tips. The angiosperm species number is from Stevens (2001 onward).

Within flowering plants, it is still unclear whether bisexual or unisexual flowers are the ancestral state and whether there was sexual specialization of sporophytes. The fossil record does not provide an answer since bisexual and unisexual flowers are both found in the oldest floras and since there are no population samples revealing the occurrence of specialized male and female sporophytes (Friis et al., 2011).

Dioecy, self-incompatibility, and polyploidy

Whether flowering plant self-incompatibility systems, i.e., genetic self-recognition involving pollen and stigma, style, or ovule tissues, influenced the evolution of dioecy is unclear. If dioecy were mainly an outcrossing mechanism, one would expect a negative correlation between self-incompatibility and dioecy. In the only review of the topic, Charlesworth (1985) unfortunately had to conclude that the data on the occurrence of the two systems were insufficient.

The chromosome numbers of few dioecious species that have also been studied in terms of their sex determination mechanisms are known, and the suggestion by Westergaard (1958) that nuclear-determined dioecy might break down in polyploids because of dosage imbalances therefore remains untested. There are polyploid dioecious species in Bryonia, Fragaria, Mercurialis, and Rumex (Obbard et al., 2006; Volz and Renner, 2008; Ashman et al., 2013; Njuguna et al., 2013), all of which also have diploid monoecious or gynodioecious species, and genomic work (some ongoing) should shed light on how polyploidy affects genetic sex determination.

Relative and absolute frequencies of angiosperm sexual systems

Table 1 shows the frequencies of all sexual systems of flowering plants and their known or proposed relationships to dioecy, discussed further in the next sections. Appendix S1 (see Supplemental Data with the online version of this article) lists the 987 genera known to include some or only dioecious species. The list implies a minimum bound of 871 independent origins of dioecy because 126 of the 987 genera belong to 10 families that are entirely dioecious (987–116 = 871). An upper bound can also be extrapolated: The species in the 987 genera that are dioecious add to 15600 (Table 1). Some 6650 of these species belong to just 34 clades that are entirely dioecious (Table 1). The remaining 8950 species are in families or genera with dioecious and nondioecious species, indicating one or more origins of dioecy within them. The number of origins of dioecy may therefore lie somewhere between 871 and 5000, greatly exceeding a previous estimate of 100 independent origins of dioecy in the flowering plants (Charlesworth and Guttman, 1999; Charlesworth, 2002).

Table 1. Sexual systems found in flowering plants, their frequency, suggested relationship to dioecy, and selected references, preference being given to reviews. Sexual system here refers to features of flowers, individuals, and populations that influence the potential for crossing between individuals. The realized extent of successful crossing will depend on pollinator behavior, self-incompatibility alleles, and consanguinity of the population.
System Level at which visible Frequency Linked to dioecy? Key references
Temporal separation of gamete export/import
    Dichogamy Each flower or inflorescence with temporal separation of male and female function Common (the default condition in bisexual flowers?) No Lloyd and Webb, 1986; Bertin and Newman, 1993
    Ontogenetic sex change (plasticity sensu Charnov and Bull, 1977) Young individuals male, older plants female (reversals are possible) Rare, c. 250 species of Gurania/Psiguria (Cucurbitaceae); Catasetinae (Orchidaceae); Arisaema (Araceae), Elaeis (palms) No Schlessman, 1988; Condon and Gilbert, 1988
    Duodichogamy Each individual with sequential batches of flowers, usually in the sequence male, female, male, the individuals being out of synch with each other Rare, few species of in 5 genera in 5 families (Acer, Bridelia, Castanea, Cladium, Dipteronia) No Lloyd and Webb, 1986; Luo et al., 2007 (reviews all cases)
    Heterodichogamy (incl. flexistyly) Populations with two types of genetically different individuals, ones that are first male, then female, others vice versa (can involve bisexual or unisexual flowers) Rare, c. 50 species in 20 genera in 12 families Four of the genera (Acer, Grayia, Spinacia, Thymelaea) contain a few dioecious species, but evolutionary directions are unclear Renner, 2001; Teichert et al., 2011; X. Wang et al., 2012b; Fukuhara and Tokumaru, 2014
Spatial separation of gamete export/import
    Monocliny Each flower bisexual Most angiosperms No (by definition) Darwin, 1876, 1877
    Dicliny Each flower unisexual All monoecious and dioecious angiosperms Yes (by definition) This study
    Heterostyly (distyly, tristyly) Populations with two or three types of genetically different individuals, ones with short styles, and long stamens, plus one or two other types (all bisexual flowers) A few hundred species in 28 families No Barrett and Shore, 2008
    Enantiostyly Each inflorescence or each individual with two types of flowers, some left-handed, others right-handed (monomorphic or dimorphic enantiostyly) Rare; monomorphic: 25 genera in 10 families; dimorphic: 5 genera in 3 families No Jesson and Barrett, 2003
    Monoecy Each individual with unisexual flowers Probably slightly higher frequency than dioecy in terms of species, genera, and families Statistically strongly positively linked Renner and Ricklefs, 1995; this study
    Andromonoecy Each individual with bisexual and male flowers Rare Unknown Torices et al., 2011
    Gynomonoecy Each individual with bisexual and female flowers Common, esp. in Asteraceae No (de Jong and Geritz, 2001) Torices et al., 2011
    Androdioecy Populations with bisexual individuals and male individuals A handful of species: Datisca glomerata, Fraxinus ornus, Mercurialis annua, Schizopepon bryoniifolius; Sagittaria spp.; Phillyrea angustifolia Positively associated Pannell, 2002
    Gynodioecy Populations with bisexual individuals and female individuals 275 angiosperm genera contain at least 1 gynodioecious species Spigler and Ashman (2012: p. 531) concluded “we still have very little evidence for the second step, i.e. the transition from gynodioecy to dioecy.” Spigler and Ashman, 2012; Dufay et al., 2014
59 (0.4% of 14559) genera have dioecious + gynodioecious species
    Dioecy Populations with male and female individuals 987 genera contain at least 1 dioecious species N/A This study
5–6% of all angiosperm species are dioecious, resulting from 871 to 5000 independent origins of dioecy
210 (1.4% of 14559) genera have dioecious + monoecious or polygamous species
    Trioecy Populations with bisexual individuals, male individuals, and female individuals Atriplex canescens; Carica papaya; Fraxinus excelsior; Pachycereus pringlei Positively associated McArthur et al., 1992; Fleming et al., 1994; this study

The 987 genera represent 6.8% of the 14559 genera accepted in The Plant List (2013). They belong to 175 of the 405–449 (38–43%) families of angiosperms currently accepted in The Plant List or the Angiosperm Phylogeny website (Stevens, 2001 onward) and represent 6% of the 261750 total species accepted in the Angiosperm Phylogeny website or 5% of the 304419 species accepted in The Plant List (which contains many synonymous names). While such fractions will continue to change, I was surprised by how little has changed since a previous estimate (Renner and Ricklefs, 1995: 38% families, 7% genera, 6% species).

Correlates of dioecy: Wind pollination

At least 16700 angiosperm species are wind-pollinated, or 5.5 to 6.4% of the 304419 to 261750 estimated species of angiosperms. Of the 15.600 dioecious species, at least 4935 (Appendix S1) are wind-pollinated (31.6%), supporting earlier analyses (Freeman et al., 1979; Renner and Ricklefs, 1995; Vamosi et al., 2003). At the genus level, of 806 genera that have data for pollination type, 240 (30%) are wind- or water-pollinated (Appendix S1). The proposed explanation is that bisexual flowers are difficult to optimize for abiotic pollen export and import, without stigmas becoming clogged with self-pollen (Darwin, 1876; Lloyd and Webb, 1986).

Whether unisexual flowers facilitate the evolution of wind pollination or the other way around is still unclear, and the two traits may feed back on each other and evolve in concert. In Thalictrum (Ranunculaceae), with 196 species worldwide, some insect-pollinated, some wind-pollinated, some monoecious, some dioecious, wind pollination was inferred to have evolved early, followed by multiple losses and regains, and generally preceding the origin of unisexual flowers (Soza et al., 2012). By contrast, there was no evidence for unisexual flowers preceding wind pollination.

Correlates of dioecy: Longevity

The other clear correlation of dioecy is with longevity (the proxy being woody growth; Renner and Ricklefs, 1995; Vamosi et al., 2003), with the classic explanation being that long-lived species are better able to survive a season of reproductive failure (Darwin, 1876; Freeman et al., 1979). On the other hand, sister clade comparisons among groups with and without dioecious species (Heilbuth, 2000) and simulations of the dynamics of population size in competing cosexual and dioecious plants (Heilbuth et al., 2001) demonstrate the negative demographic effects of dioecy and point to lower speciation or higher extinction rates of dioecious species compared with monomorphic (cosexual) relatives, whether monoecious or perfect-flowered. Whether dioecy is strongly correlated with lower diversification rates, however, remains a controversial topic (Vamosi and Vamosi, 2004 vs. Käfer and Mousset, 2014; Käfer et al., 2014). That 43% of all dioecious angiosperms are in just 34 entirely dioecious clades (Appendix S1) argues against a consistent negative influence of dioecy on diversification.

Correlates of dioecy: Monoecy and gynodioecy

About 275 (1.8%) of the 14559 genera of angiosperms have one or more gynodioecious species, and 59 genera have gynodioecious as well as dioecious species (Dufay et al., 2014). By contrast, about 210 (21%) of the genera with dioecious species also contain monoecious species (see column 6 in Appendix S1). In other words, 1.4% of genera contain dioecious and monoecious species, while 0.4% contain dioecious and gynodioecious species. However, from herbarium material, these systems cannot be reliably distinguished, and this is true even with some living populations (Spigler and Ashman, 2012). Herbarium-based species descriptions therefore often resort to Linnaeus's term polygamy, which refers to the presence of unisexual and bisexual flowers on some or all individuals. Polygamy is used dozens of times by Darwin (1877) and 75 times by Yampolsky and Yampolsky (1922). The relatively few reports of gynodioecy, compared with monoecy, may be an artifact of scarce fieldwork on tropical plants.

Phylogenetic studies on the gain and loss of dioecy

A literature search, combined with a GenBank search for sequences of relevant taxa, turned up few studies focusing on the evolution of dioecy. One of the largest is that on Thalictrum by Soza and colleagues (2012), who sampled 63 of the 196 species and found that dioecy, andromonoecy, and gynomonoecy evolved at least twice from hermaphroditism. Lloyd (1972, 1975a, b, 1980) in his studies of the Asteraceae genus Leptinella may have been the first to infer returns from dioecy to monoecy, inferences since supported by molecular phylogenetic data (Himmelreich et al., 2012). A study of the African Cucurbitaceae genus Momordica, with 58 of its 59 species sampled, also found repeated returns from dioecy to monoecy (Schaefer and Renner, 2010), and similar returns have also been inferred in the small Cucurbitaceae genus Bryonia (Volz and Renner, 2008) and the medium-sized family Caricaceae (Carvalho and Renner, 2012). Another well-sampled phylogeny, for the Caryophyllaceae Schiedea, with monomorphic species as well as 10 gynodioecious, subdioecious, or dioecious species (Willyard et al., 2011), unfortunately lacked resolution to infer the evolutionary sequence of the implied transitions, a common issue in phylogenetic work on rapidly evolving young clades.

Well-studied genera with labile sex allocation, but little strict dioecy, are Acer (Renner et al., 2007) and Fragaria (Njuguna et al., 2013), the latter discussed further in the next section. Additional phylogenetic studies of the origin of dioecy are listed in Appendix S1 (e.g., Ecballium: Costich and Meagher, 1992; Luffa: Filipowicz et al., 2014; Coccinia, a group of 35 dioecious species, one with the largest Y chromosome known; Sousa et al., 2013).

In sum, phylogenetic studies show that dioecy is not an evolutionary dead end, but instead can be lost and regained repeatedly. Since most have focused on small clades, these studies might give the impression that dioecious clades tend to be small. However, 43% of all dioecious angiosperms are in just 34 species-rich entirely dioecious clades (Appendix S1) that simply have not been studied.

The genetic underpinnings of dioecy in flowering plants

The suppression of male or female function in flowering plants has evolved perhaps 871 to 5000 times (see earlier section Relative and absolute frequencies of angiosperm sexual systems), implying the repeated establishment of recombination-suppressed chromosome regions (Charlesworth and Charlesworth, 1978) or of gene groups on different chromosomes corresponding to environmental cues (Yin and Quinn, 1992, 1995a, b; Chuck, 2010; Golenberg and West, 2013). Developing a single framework for sex determination in flowering plants is difficult because of many different combinations of sex expression at the flower, individual, and population level, which can involve spatial separation of sexual function, temporal separation, or both, as in heterodichogamy (Table 1 for all known angiosperm sexual strategies). Moreover, the different forms are not hierarchically grouped in clades, but can coexist in single species or among close relatives, for example, in the Cucurbitaceae, which have environmental sex determination (Condon and Gilbert, 1988) as well as chromosomal sex determination (Sousa et al., 2013).

Visibly distinct (heteromorphic) sex chromosomes currently are known from 19 species in four families, namely Cannabis and Humulus in Cannabaceae, Silene in Caryophyllaceae, Coccinia in Cucurbitaceae, and Rumex in Pologynaceae (reviewed by Ming et al., 2011). So far, no relationship between age and degree of heteromorphism is apparent (Sousa et al., 2013). The largest Y chromosome is that of the cucurbit Coccinia grandis, with an age of about 3 Myr (Sousa et al., 2013), followed by that of Silene latifolia, estimated to be 5–10 Myr old (Bergero et al., 2007). How exactly these Y chromosomes became so large is no yet clear, although the accumulation of transposons and other types of repetitive DNA has been implicated. Surprisingly many of the Y-linked genes of S. latifolia are still present and transcribed (Bergero and Charlesworth, 2011; Chibalina and Filatov, 2011). An interesting discovery in this context is that the closely related S. diclinis has a neo-sex chromosome system (XY1Y2) that must have evolved from an ancestral XY system of the type still present in S. latifolia (Weingartner and Delph, 2014). Such XY1Y2 sex chromosome systems are also found in several species of Rumex that in addition can have XY chromosomes. This is the case in R. hastatulus, in which high throughput transcriptome sequencing revealed ongoing degeneration of Y-linked genes (Hough et al., 2014).

Visibly homomorphic sex chromosomes, identified by classic genetic crossing or by molecular methods, such as FISH cytogenetics, are known from another 20 species in 13 families (Ming et al., 2011). Among the better-studied homomorphic sex chromosomes are those of strawberries, papaya, and date palms. Genetic mapping in dioecious and gynodioecious species of Fragaria uncovered proto-sex chromosomes and sex chromosome turnover among sibling species (Goldberg et al., 2010; Spigler et al., 2008, 2010). Proto-sex chromosomes also exist in Carica papaya, with the male-specific region predicted to be approximately 8–9 Mb and larger than its X homologue, mostly due to retrotransposon insertions, organelle DNA-derived sequences, and movement of genes onto the Y (Liu et al., 2004; J. P. Wang et al., 2012a; VanBuren and Ming, 2014). Dioecy is the ancestral condition in the papaya family (35 species; Carvalho and Renner, 2012), and wild papaya is strictly dioecious (Chávez-Pesqueira et al., 2014) while cultivated papaya is trioecious, with pure male plants, pure female plants, and plants with bisexual flowers that have a nonfunctioning Y chromosome, implying a breakdown of dioecy. In the date palm, Phoenix dactylifera, the sex-segregating region has been localized to 5–13 Mb on chromosome 12, and since the sex-linked markers group by sex, not by species, recombination suppression may have begun before the separation of the ca. 14 species (Cherif et al., 2013; Mathew et al., 2014).

Plasticity in sex expression and theoretical models on plasticity

Plasticity in sex expression means that sexual function changes adaptively during each individual's lifetime (Charnov and Bull, 1977; Korpelainen, 1998). According to models contrasting the evolution of plasticity in sex allocation in animals and plants, habitat-dependent allocation in plants evolves differently from that in plants because of the way male and female gametes are dispersed (Guillon et al., 2006). An extreme form of sexual plasticity is environmental sex determination or sex choice (Freeman et al., 1979; McArthur et al., 1992; Delph and Wolf, 2005; Table 1), which occurs in the Araceae Arisaema, the oil palm Elaeis, the orchids Catasetum and Cycnoches, and the Cucurbitaceae Gurania/Psiguria (Table 1). In other cases, there is a more or less regular cycling of functionally male and female phase over the flowering season, as in duodichogamy, which involves asynchronous switches between male and female flower production in the individuals of a population. A genus with well-documented temporal plasticity is Acer; monoecious maples can exhibit duodichogamy and sometimes heterodichogamy (Renner et al., 2007; Shang et al., 2012: Acer pictum subsp. mono; Table 1). Besides in Acer, duodichogamy occurs in Castanea (Fagaceae; Stout, 1928, who also coined the term); Dipteronia, the sister group of Acer; the Cyperaceae Cladium jamaicense (Snyder and Richards, 2005), and the Phyllanthaceae Bridelia tomentosa, which in addition is androdioecious (Luo et al., 2007).

Spatially organized plasticity in allocation to male and female function is much better documented than the just discussed temporal sexual strategies, probably because the latter can only be detected with prolonged monitoring of living individuals. Spatial plasticity was reviewed by Delph and Wolf (2005) and has been the focus of much recent work (e.g., Stehlik et al., 2008; Dorken and Mitchard, 2008; Yakimowski and Barrett, 2014 and studies cited therein). An example is Sagittaria latifolia, which at its northern range limits has gynodioecious to subdioecious populations, possible because of male inconstancy in which males also produce a few fertile female flowers (Yakimowski and Barrett, 2014).

An important question is whether temporal or spatial plasticity in sexual strategy facilitates or delays the evolution of dioecy. Few models have tried to address this question (Delph and Wolf, 2005; Pannell, 2005; Ehlers and Bataillon, 2007; Crossman and Charlesworth, 2014). Some find that depending on the genetics of sex determination, pure dioecy, stable subdioecy or trioecy (as in Carica papaya), or coexistence of pure males, inconstant males, and pure females (as in S. latifolia) can be stable (Ehlers and Bataillon, 2007; Crossman and Charlesworth, 2014). These findings bear not only on the evolution of dioecy, but also its breakdown. Crossman and Charlesworth (2014), extending the models of Ehlers and Bataillon (2007), further showed that dioecy is stable to invasion by modified males with cosexual phenotypes in a large region of parameter space. A model presented by Pannell (2005), on the other hand, shows that irrespective of the genetic or developmental basis of sex determination, frequency-dependent selection will bring the phenotypic frequencies of males and hermaphrodites to rest at a predictable equilibrium.

Conclusions and future directions

Important open questions regarding dioecy concern (1) the causes and evolutionary role of sexual plasticity in local ecological contexts, including the breakdown of dioecy at the population level (a research direction also suggested by Barrett and Hough, 2013); (2) the genetic mechanisms by which sexual specialization is brought about (Chuck, 2010; Ming et al., 2011; Golenberg and West, 2013; Hough et al., 2014); and (3) how the determination of floral sex and that of entire individuals hang together. There is at least one case in which sex chromosomes carry MADS-box genes and where a transposon inserted into such a gene can interrupt the encoding of a functional protein, thereby interrupting carpel suppression, which then results in bisexual flowers in papaya males with this chromosome (H. Matsumura, Shinshu University, Ueda, Japan, personal communication, July 2014). In Cucumis melo, by contrast, pure females (instead of the normal monoecious plants) result from the repression of a transcription factor promoter due to retrotransposon-mediated DNA methylation (Martin et al., 2009). More such studies of the actual mechanisms underlying plant sexual systems may shed light on the well-established statistical correlations between monoecy, polygamy (both polygamodioecy and polygamomonoecy), and dioecy.