Sexual systems in the New Zealand angiosperm flora

ABSTRACT

We present data on sexual systems and associated traits in the New Zealand angiosperm flora and discuss reasons for the anomalously high levels of gender dimorphism in the flora, and the low levels of monoecy in woody species. Along with Hawai'i and New Caledonia, New Zealand has exceptionally high levels of gender dimorphism (19.5% of angiosperm species). The plant traits associated with gender dimorphism (woody growth, small, unspecialised white to yellow-green flowers, abiotic pollination, fleshy fruit) are the same as those in other regions and most gender dimorphic species belong to lineages that were already gender dimorphic on arrival in New Zealand. We attribute the high levels of gender dimorphism to several distinct factors. New Zealand’s evergreen forests have many small trees and understorey shrubs with fleshy fruit and small, open, inconspicuous flowers, a combination characterised by high levels of gender dimorphism elsewhere. Many of these species belong to lineages that migrated from the tropical north, a region with high levels of gender dimorphism. In comparison with many other regions, the New Zealand angiosperm flora has few annuals, and few plants with large, specialised flowers or pollinated by birds, traits elsewhere associated with exceedingly low levels of gender dimorphism. Finally, chance may have played a role through the association of gender dimorphism with rapidly radiating lineages. While the New Zealand angiosperm flora has similar levels of monoecy (14.2%) to other comparable regions, monoecy is exceptionally uncommon in the tree flora (3.4% for strictly monoecious species). However, the endemic Nothofagaceae and introduced woody monoecious species thrive in New Zealand. We suggest it is the lack of temperate sources for monoecious tree species, combined with the difficulty large-fruited monoecious tropical species have in crossing ocean gaps that may be ultimate reason for their failure to establish in greater numbers.

KEYWORDS:

• Dioecy
• cenozoic flora
• evergreen broadleaved temperate forest
• fleshy fruit
• flower colour and size
• gender dimorphism
• immigrant selection
• maximum height
• monoecy
• pollination mode

Introduction

The angiosperm floras of the Pacific archipelagos of New Zealand, New Caledonia, and Hawai'i have some of the highest global levels of gender dimorphism (Sakai et al. 1995b; Webb et al. 1999; Schlessman et al. 2014). In New Zealand, awareness of this striking feature of the flora and an abundance of research material has resulted in a sustained research effort into plant sexuality and gender dimorphism in particular (Godley 1957, 1964, 1979; Lloyd 1975, 1980a, 1980b; Delph 1990; Delph and Lloyd 1996; Webb et al. 1999). A survey of the sexual systems of the New Zealand seed plant flora (Webb et al. 1999) presented statistics at the genus level and showed gender dimorphism at the genus level to be an exceptional 23% but exhibiting the same correlations with plant traits as elsewhere. In the 20 years since the Webb et al. (1999) survey a large amount of revision at family, generic, and species level has been carried out, documentation of the flora – including high quality images of floral structures – has improved markedly, and molecular phylogenies now include most of the vascular plant genera present in the country. The time is therefore right for a re-examination of the situation.

Angiosperms rely on biotic (vertebrates, insects) and abiotic (wind, water, gravity) agents to transfer pollen from plant to plant. A diversity of floral structures has evolved to accommodate these different agents and environments. Importantly, the male (staminate) and female (pistillate) functions can be separated within flowers, between flowers on the same individuals, and between individuals in a population. The resulting proliferation of arrangements of staminate and pistillate functions ( = genders) fall into two major groups: monoclinous, where the flowers are all bisexual; and diclinous, where sexual functions are variously distributed among flowers and individuals. In Table 1 we list the most important sexual systems.

Table 1. Angiosperm sexual systems found in New Zealand angiosperms.

Flower level	Species level	Definition	Simplified category
Monoclinous, Perfect, Bisexual	Hermaphrodite	Each flower of each individual has stamens and pistils	Hermaphrodite (and Cosexual)
Diclinous, Imperfect, Unisexual	Gynomonoecious	Bisexual and pistillate flowers on same plant	Monoecious (and Cosexual)
	Andromonoecious	Bisexual and staminate flowers on same plant
	Monoecious	Unisexual flowers on the same plant.
	Dioecious	Each plant with only staminate or only pistillate flowers	Gender dimorphic
	Gynodioecious	Bisexual and pistillate flowers held on separate plants

Note: Complex variations of these systems occur when the idealised categories presented here blend into one another within a population. These are often referred to as polygamous systems or, in the case of dioecious plants where bisexual flowers occur sporadically, ‘leaky’.

The sexual system characteristic of a species is closely related to the order, family, or genus they belong to, their flower type, fruit dispersal mode, plant form, and environment (Renner and Ricklefs 1995; Renner 2014). Because taxonomic composition, plant traits and environments are spatially variable, substantial differences exist within and between regions in the proportions of diclinous species (monoecious, dioecious, gynodioecious and variants of these systems). Most remarked has been the prevalence of dioecy (and its variants) because dioecious species are not only diclinic, but gender dimorphic, that is, individual plants belong to separate sexes. Separation of sexes is important as it strongly influences outbreeding, sexual selection, and specialisation, and has ecological consequences with regard to pollinator and dispersal agents and competitive interactions with cosexual species.

We have compiled a database on plant traits, including sexual system, for the entire New Zealand angiosperm flora. Traits are often tightly linked within genera or families and this can confound analyses. Phylogenetic correction is often used to determine what evolutionary drivers are responsible for promoting which trait (Freckleton et al. 2015). However, most of the trait variance we are concerned with lies at the generic level, and most New Zealand genera are the product of independent colonisation and radiation. That is to say that the trait evolution we are concerned with happened elsewhere and the New Zealand situation is largely a consequence of biogeographic and ecological factors operating on predetermined trait bundles. Hence we have confined our analyses to documenting the distribution of traits at a species level and on trait bundles within genera as this provides sufficient phylogenetic information for our purposes.

We aim first to establish, in the light of new information, the distribution of sexual systems in the New Zealand flora at the genus and species level and to assess its association with morphological traits. We contrast and compare these results with those of similar studies carried out elsewhere. Our second aim is to address the role that these traits, the environment, biogeography, and history may have played in promoting the globally anomalous distribution of sexual systems in the New Zealand flora. To assist with this task, we draw on trait and distribution databases we have assembled and a comprehensive New Zealand-wide forest plot network. As nearly all indigenous New Zealand genera evolved elsewhere, and endemic species tend to have the same sexual system as their overseas siblings, we do not address the question of how the various sexual forms have evolved or phylogenetic issues.

Methods and materials

Trait definitions and data sources

Sexual systems: Definitions are given in Table 1. For our main analyses, the sexual system is simplified to: hermaphroditism – bisexual flowers only; monoecy – includes andromonoecy, gynomonoecy; gender dimorphism – includes dioecy, gynodioecy, androdioecy and polygamodioecy. In other analyses, we contrast cosexual species (hermaphrodite + monoecy) with gender dimorphism. We note that in international compilations of sexual system, this level of detail is often all that is given and ‘dioecy’ effectively refers to gender dimorphism. Sexual system was sourced from standard flora treatments in the Flora of New Zealand series (http://www.nzflora.info/) (Schönberger et al. 2018) supplemented by the compilation in Webb et al. (1999) and field guides (Mark 2012; Dawson et al. 2019).

Life forms: In Table 2 we define life forms for this paper. Life forms were sourced from standard flora treatments in the Flora of New Zealand series (http://www.nzflora.info/) (Schönberger et al. 2018) supplemented by field guides (Mark 2012; Dawson et al. 2019).

Table 2. Life form classification for the New Zealand angiosperm flora (n = 2080 taxa) with the number of species of each gender in each lifeform.

			Cosexual				Gender dimorphic
Life form	Definition	No. species	H	M	GM	AM	GD	D
Woody	Significant lignification of above ground parts
Trees^1	Self-supporting but ≥5 m tall	237	123	9	4	1	24	77
Shrubs	Self-supporting and <5 m tall	382	206	3	23	0	77	72
Lianas^2	Climbers reliant on other plants for support	31	12	0	1	0	1	17
Herbaceous	No or insignificant lignification of above ground parts
Herbs^3	Self-supporting; includes sub-shrubs	1415	1026	156	67	28	33	105
Vines^2	Climbers reliant on other plants for support	15	12	3	0	0	0	0
All species		2080	1379	171	95	29	135	271

Note: H = hermaphrodite; M = monoecious; GM = gynomonoecious; AM = andromonoecious; GD = gynodioecious; D = dioecious.

¹ Height definitions of trees can range from 2–6 m, and some ignore height as a factor, defining shrubs as having numerous stems

² Vine is often used interchangeably with liana and the term climber often used without qualification

³ Sub-shrubs have herbaceous canopies but woody root stocks and lower stems

Fruit and dispersal: Fruits are often broadly classified as ‘fleshy’ or ‘dry’, and this trait has a strong correlation with sexual system (Bawa 1980). Fleshy fruits in New Zealand are distributed by birds and a single bat species (Mystacina tuberculata). Dry-fruited species are mainly dispersed by wind or water, or simply fall from the plant canopy. Some dry-fruited species may have been dispersed by ingestion or attachment with barbs or spines (e.g. Acaena, Carex) but these are nearly all low-growing, and thus the distinction between ‘fleshy’ and ‘dry’ we make here is between dispersal by flighted animals and all other modes. Dispersal by flighted animals is inferred when the fruit is fleshy, or in the case of non-fleshy fruits, when the fruit is attractive to birds through the seed being coated with a brightly coloured fleshy aril or sarcotesta, or if the seed (usually glossy) is retained after the capsule opens and is displayed against a contrasting background of often brightly coloured valves. Dull coloured elaiosomes are assumed to be an adaptation to myrmecochory and not included. Other ‘dry’ dispersal types are inferred for dry capsular fruit with dull colours, fruits with wings (samaras) and winged seeds, hooked or spiny fruits, or no obvious dispersal adaptations. Dull coloured fruits with leathery skins, and fibrous, corky, or air-filled interiors are assumed to be water dispersed and included under the ‘dry’ category. Fruit traits and dispersal mode were sourced from standard flora treatments in the Flora of New Zealand series (http://www.nzflora.info/) (Schönberger et al. 2018), supplemented by Thorsen et al. (2009) and field guides (Mark 2012; Dawson et al. 2019).

Floral: Pollination mode is based for the most part on flower morphology and described either as abiotic (largely wind) or biotic (largely arthropods) as finer discrimination was not warranted by the small numbers pollinated by birds or other agents. Pollination mode was sourced from the Flora of New Zealand series (http://www.nzflora.info/) (Schönberger et al. 2018), supplemented by Newstrom and Robertson (2005). Flower size is the estimated maximum dimension of the corolla. Where not provided, we made an estimate based on twice the length of the petal. Flower size of the Asteraceae is not included as the compound inflorescence, with petals often confined to the outer ray florets, makes comparisons problematic. Flower colours were recorded for biotically pollinated species only and are from descriptions given in the floras or original publications. The most prominent colour was used to code the flower and, where there were no petals or sepals, the predominant colour of the calyx/stamens/style was used. The colour coding groups follow those in Godley (1979).

Height: The maximum height at maturity for individuals of a species is one of the fundamental traits that define plant strategy (Thomas 1996; Westoby et al. 2002). Heights (m) were taken from flora compilations and represent the upper bound of the range given, but not the extreme maximum (McGlone et al. 2010).

Range size and abundance in forest communities for woody species

We determined the presence of each woody species in each 1° latitude band and calculated species’ latitudinal range extent following McGlone et al. (2010). We used a network of 796 systematically located 20 m × 20 m (400 m²) plots across indigenous forests throughout New Zealand (Holdaway et al. 2017) to estimate the relative contributions to total basal area and stem count by hermaphrodite, monoecious, and gender dimorphic tree species.

Data analyses

We acknowledge the phylogenetic non-independence among species in our compilations but point out that this study is not concerned with the evolution of sexual systems but aims to provide a solid basis for understanding the ecological role of sexual systems in the flora and the role of history in their distribution. We, therefore, have focussed on the frequencies of different sexual systems and trait/sexual system combinations. We used Kolmogorov Smirnov statistics to test for differences in the distribution of species richness per genus values between cosexual and gender dimorphic genera within trees, all woody species, and herbs. Height profiles of woody floras were analysed by binning species into five height classes (<2 m, 2–4.9 m; 5–9.9 m; 10–19.9 m; ≥20 m) following (Wang et al. 2020b) and then plotting the frequency of hermaphrodite, monoecious, and gender dimorphic species within each height class. To examine the distribution of forest tree abundance across the three main sexual systems we summed basal area (m²) and the number of stems across all plots for hermaphrodite, monoecious, and gender dimorphic tree species (total area sampled was 796 × 0.04 ha = 31.84 ha). We used χ² tests to determine whether the distribution of basal area and stem counts were proportional to the number of species in each sexual system.

Results

We have compiled trait data for 2,080 angiosperm species in 358 genera and 113 families from the New Zealand region.

Sexual systems in the New Zealand angiosperm flora (Tables 2 and 3)

Hermaphroditism is the most common sexual system in New Zealand: 66.3% of the angiosperm flora and 74% of the genera have at least one hermaphroditic species. Dioecy is the next most common sexual system after hermaphroditism, with a level of 13% for the whole flora, similar to earlier estimates (12–13%; Godley 1979). Some 6.6% of genera, and 6.2% of species have been classified as gynodioecious. Gender dimorphism accounts for 19.5% of the flora, and 21.8% of the genera have at least one dimorphic member (Table 3). Monoecy (including gynomonoecy and andromonoecy) characterises 14.2% of species.

Table 3. Traits of New Zealand angiosperm genera with solely cosexual species or with at least one gender dimorphic species.

	Solely cosexual	Can be gender dimorphic
Number of genera	280	78
% all genera	78.2	21.8
% woody	22.8	62.8
% fleshy-fruited	17.2	50.0
% with biotic pollination	69.7	76.9

Note: Traits varied among species within genera and in this analysis, if any species in the genus was woody, fleshy-fruited, or had biotic pollination, the genus was considered to have that trait for this analysis.

Sexual systems in relation to plant traits

Habit

About one-third of the New Zealand angiosperm flora is woody (31.3%, 650 species; Tables 2 and 4 & S1) and these species have high levels of dioecy and gynodioecy (Tables 2 and 4) and thus gender dimorphism (Table 2, Tables 4 and 5). About 16% of woody species in the flora are gynodioecious and dominate the gynodioecious total (c. 76% of species and 50% of the 24 genera with gynodioecious species are woody). Monoecy is uncommon in woody plants (6.3%; Tables 2 and S2) and only 14 species of trees (5.9%) in seven genera are monoecious (i.e. Ascarina, Brachyglottis, Fuscospora, Leptospermum, Lophozonia, Ceodes ( = Pisonia), Raukaua) (Table S5).

Table 4. Distribution of plant sexual systems within traits and trait combinations in the New Zealand angiosperm flora.

Traits or trait combinations	Number of species	H	M	GM	AM	GD	D	Cosexual	Gender Dimorphic	χ^2	P
All species	2080	66.3	8.2	4.6	1.4	6.5	13.0	80.5	19.5
Woody + Fleshy + Abiotic Pollination	68	11.8	1.5	0.0	0.0	0.0	86.8	13.2	86.8	171.61	< 0.0001
Woody + Abiotic Pollination	77	11.7	7.8	0.0	0.0	0.0	80.5	19.5	80.5	159.06	< 0.0001
Fleshy-fruited + Abiotic Pollination	87	24.1	2.3	1.1	0.0	0.0	72.4	27.6	72.4	134.67	< 0.0001
Woody + Fleshy-fruited	278	26.6	2.5	0.0	0.0	23.4	47.5	29.1	70.9	336.96	< 0.0001
Fleshy-fruited	328	31.1	2.4	0.3	0.0	20.1	46.0	33.8	66.2	318.92	< 0.0001
Woody + Fleshy-fruited + Biotic Pollination	210	31.4	2.9	0.0	0.0	31.0	34.8	34.3	65.7	222.19	< 0.0001
Fleshy-fruited + Biotic Pollination	241	33.6	2.5	0.0	0.0	27.4	36.5	36.1	63.9	229.97	< 0.0001
Herbaceous + Fleshy-fruited + Biotic Pollination	31	48.4	0.0	0.0	0.0	3.2	48.4	48.4	51.6	17.72	< 0.0001
Woody	650	52.5	1.8	4.3	0.2	15.7	25.5	58.8	41.2	124.39	< 0.0001
Herbaceous + Fleshy-fruited	50	56.0	2.0	2.0	0.0	2.0	38.0	60.0	40.0	11.55	0.0007
Woody + Biotic Pollination	573	57.9	1.0	4.9	0.2	17.8	18.2	64.0	36.0	67.43	< 0.0001
Woody + Dry-fruited + Abiotic Pollination	9	11.1	55.6	0.0	0.0	0.0	33.3	66.7	33.3	0.39	0.5345
Biotic Pollination	1499	70.9	2.0	5.8	0.1	8.5	12.7	78.8	21.2	1.45	0.2289
Herbaceous + Fleshy-fruited + Abiotic Pollination	19	68.4	5.3	5.3	0.0	0.0	21.1	78.9	21.1	0.00	1.0000
Woody + Dry-fruited	372	71.8	1.3	7.5	0.3	9.9	9.1	80.9	19.1	0.02	0.9019
Woody + Dry-fruited + Biotic Pollination	363	73.3	0.0	7.7	0.3	10.2	8.5	81.3	18.7	0.08	0.7812
Abiotic Pollination	581	54.4	24.3	1.4	4.8	1.2	13.9	84.9	15.1	5.46	0.0195
Dry-fruited + Biotic Pollination	1258	78.1	1.9	6.9	0.1	4.9	8.1	87.0	13.0	22.81	< 0.0001
Herbaceous + Biotic Pollination	926	78.9	2.6	6.4	0.0	2.8	9.3	87.9	12.1	24.24	< 0.0001
Dry-fruited	1752	72.9	9.3	5.4	1.7	3.9	6.8	89.2	10.8	54.61	< 0.0001
Herbaceous + Dry-fruited + Biotic Pollination	895	80.0	2.7	6.6	0.0	2.8	7.9	89.3	10.7	33.87	< 0.0001
Herbaceous	1430	72.6	11.1	4.7	2.0	2.3	7.3	90.3	9.7	62.27	< 0.0001
Herbaceous + Dry-fruited	1380	73.2	11.4	4.8	2.0	2.3	6.2	91.4	8.6	76.81	< 0.0001
Herbaceous + Abiotic Pollination	504	60.9	26.8	1.6	5.6	1.4	3.8	94.8	5.2	59.06	< 0.0001
Dry-fruited + Abiotic Pollination	494	59.7	28.1	1.4	5.7	1.4	3.6	94.9	5.1	58.83	< 0.0001
Herbaceous + Dry-fruited + Abiotic Pollination	485	60.6	27.6	1.4	5.8	1.4	3.1	95.5	4.5	62.44	< 0.0001

Note: We present the number of species in the whole angiosperm flora (‘all species’) and in subsets of the whole flora according to one or more traits. We then give the percentage of those species in each of the six plant sexual systems (H = hermaphrodite; M = monoecious; GM = gynomonoecious; AM = andromonoecious; GD = gynodioecious; D = dioecious), followed by the sum of those percentage figures within cosexual and gender dimorphic species. The final two columns present results from χ² tests evaluating whether the distribution of cosexual and gender dimorphic species within a trait or trait combination group differs significantly from the overall distribution of cosexual and dimorphic species in the angiosperm flora (i.e. 80.5% cosexual and 19.5% gender dimorphic). Traits and trait combinations are sorted within the table in descending order of the prevalence of gender dimorphism.

Table 5. Gender dimorphism (dioecious + gynodioecious species) in the New Zealand angiosperm flora according to pairs of life history traits.

Level of comparison	Trait pair	Total	% gender dimorphism^1
Species	Herbaceous & abiotic pollination	504	5.2
	Herbaceous & biotic pollination	926	12.1
	Woody & abiotic pollination	77	80.5
	Woody & biotic pollination	573	36.0
Genus	Herbaceous & abiotic pollination	95	12.9
	Herbaceous & biotic pollination	160	11.2
	Woody & abiotic pollination	12	46.2
	Woody & biotic pollination	103	39.3

Note: Data are shown for species and genera. For genera comparisons, some genera are counted more than once because a genus was included in a trait pair if at least one species had that combination of woodiness and pollination mode.

¹ For species level comparisons, this is the % that are gender dimorphic; for genus level comparisons, this is the % of those genera that contain at least one gender dimorphic species

Over 70% of herbs are hermaphrodite (Tables 2 and 4) including all but one of the annual herbs (4% of the flora). Dioecy in herbaceous species is relatively common at 7.3% (Table 4; 8.5% for genera with a dioecious member). Of herbaceous species, 2.3% are gynodioecious (Table 4) and 4.3% of herb genera have a gynodioecious member. Herbaceous species are therefore 9.7% gender dimorphic, most of these dioecious (Tables 4 and S1) but just two 2 clades, Aciphylla / Anisotome / Gingidia (Apiaceae) and Astelia (Asteliaceae) provide c. 68% of the herbaceous gender dimorphic total. Monoecy (and its variants) is most common among herbaceous species (Tables 2 and S2) and 86% of monoecious species are herbs or vines (Table 2).

Fruit type

Fleshy-fruited species are 66.2% gender dimorphic whereas only 10.8% of dry-fruited species are (Tables 4 and S1, Figure 1). A substantial number of woody species are fleshy-fruited (43%) and 71% of these are gender dimorphic (Tables 4 and S1, Figure 1). In contrast, most herbaceous species are dry-fruited (96.5%) (Tables 4 and S1). Gender dimorphic herbs are 86% dry-fruited (mainly the Aciphylla/Anisotome/Gingidia clade), and of gender dimorphic herbs only Astelia and Gunnera have fleshy fruit. Monoecious herbs are nearly all (99%) dry-fruited.

Figure 1. Frequency of the three major plant sexual systems within species that share combinations of woodiness, pollination mode, and possession of fleshy or dry fruits.

Pollination syndrome

Dioecy is at near identical percentages across abiotic and biotically pollinated species (c. 13%; Table 4) and this even split persists at the genus level (14% of abiotically pollinated genera have dioecious members and 15.6% of biotically pollinated genera; Table 5). Twelve percent of New Zealand woody species are abiotically pollinated (Tables 4 and S1), and these species are 15.1% gender dimorphic, less than the incidence of gender dimorphism in the whole flora (19.5%; Table 4). However, if the analysis is restricted to trees (≥5 m), abiotically pollinated tree species (n = 35 species) are mostly either dioecious (74.3%) or monoecious (17.1%), and 60% of these woody abiotically pollinated species belong to a single genus, Coprosma. Thirty five percent of the herbaceous flora is abiotically pollinated (Tables 4 and S1), but only 5.2% of these species are gender dimorphic (Tables 4 and S1). Only three abiotically pollinated genera have gynodioecious species (i.e. Sarcocornia, Chionochloa, Austroderia) and there are no abiotically pollinated gynodioecious woody species (Table S1). Newstrom and Robertson (2005) list 24 New Zealand species as having specialist ornithophilous flowers just over 1% of the flora. All are woody aside from two Phormium species (Asphodelaceae), and all except for the gynodioecious Fuchsia species are hermaphroditic.

The association of abiotic pollination with monoecy (sensu stricto) is very strong at the species level (30.5% of abiotically pollinated species are monoecious versus 7.9% of biotically pollinated species; Tables 4 and S2, Figure 1). Although 64.4% of abiotically pollinated monoecious species are from just one genus, Carex, the association between abiotic pollination and monoecy remains strong at the genus level (15% of abiotically pollinated genera have a monoecious member versus 3.2% of biotically pollinated genera). Andromonoecious species are mostly grasses and sedges and therefore abiotically pollinated (Table S1), and gynomonoecious species are mostly asterads and therefore almost exclusively biotically pollinated.

Flower size and colour

Gender dimorphic flowers tend to be small (<4.9 mm) in sharp contrast with those hermaphrodite and monoecious species (Tables 6 and S3). More than 90% of the species in the speciose gender dimorphic Aciphylla / Anisotome / Gingidia clade have small flowers, and 35% of gynodioecious Veronica and Pimelea shrubs have small flowers. Very large flowers (≥15 mm) were most frequent among the hermaphrodites and were scarce among monoecious and gender dimorphic species (Table 6). More than one-third (37%) of the very large-flowered, gender dimorphic species were lianas from two genera, Clematis and Rubus.

Table 6. Flower size (corolla diameter) in New Zealand biotically pollinated angiosperms (excluding Asteraceae and excluding abiotically pollinated species).

Flower size group	All sexes	Hermaphrodite	Monoecious^1	Gender Dimorphic^2	χ^2	P
All biotically	1202	877 (73%)	17 (1.4%)	308 (25.6%)	–	–
pollinated species
Small (0–4.9 mm)	320	170 (53.1%)	5 (1.6%)	145 (45.3%)	47.4	<0.0001
Medium (5–9.9 mm)	346	235 (67.9%)	7 (2.0%)	104 (30.1%)	3.58	0.1671
Large (10–14.9 mm)	202	168 (83.2%)	2 (1.0%)	32 (15.8%)	9.48	0.0087
Very large (≥15 mm)	334	304 (91.0%)	3 (0.9%)	27 (8.1%)	35.4	<0.0001

Note: Data are counts of species with percentages within rows given in parentheses. The final two columns present results from χ² tests evaluating whether the distribution of sexual systems within each flower size group differs significantly from the overall distribution of sexual systems in the whole biotically pollinated flora excluding Asteraceae (n = 1202 species).

¹ including andromonoecious & gynomonoecious

² dioecious and gynodioecious

Almost 60% of the biotically pollinated flora has white flowers (Tables 7 and S4), and white flowers are the most frequent colour group in hermaphrodite and gender dimorphic species, but not monoecious species (Table S4). Blue flowers are the least common colour group, nearly all being hermaphroditic (Table 7), and there were no monoecious species with blue flowers. More than half of monoecious species (68 out of 118) and close to one-third of gender dimorphic species (102 out of 318) had green or yellow flowers (Table 7), while only about one-fifth of hermaphrodite species had green or yellow flowers (236 out of 1063 species; Tables 7 and S4). Red flowers are distributed across hermaphrodite, monoecious, and gender dimorphic species in proportion to the frequency of the sexual systems (Table 7). Of the 203 biotically pollinated tree species, 65 species (32%) had green or yellow flowers and more than half of those (57%) were gender dimorphic. In contrast, 97 species (48%) had white flowers and only 18% of those were gender dimorphic.

Table 7. Flower colour and sexual system in New Zealand biotically pollinated angiosperms (including Asteraceae and excluding abiotically pollinated species).

Flower size group	All sexes	Hermaphrodite	Monoecious^1	Gender Dimorphic^2	χ^2	P
All biotically	1499	1063 (70.9%)	118 (7.9%)	318 (21.2%)	–	–
pollinated species
Green	186	114 (61.3%)	18 (9.7%)	54 (29.0%)	7.44	0.0242
Yellow	220	122 (55.5%)	50 (22.7%)	48 (21.8%)	50.0	<0.0001
White	884	666 (75.3%)	41 (4.6%)	177 (20.0%)	10.60	0.0050
Blue	80	73 (91.2%)	0 (0%)	7 (8.8%)	16.4	0.0003
Red	129	88 (68.2%)	9 (7.0%)	32 (24.8%)	0.96	0.6197

Note: Data are counts of species with percentages within rows given in parentheses. The final two columns present results from χ² tests evaluating whether the distribution of sexual systems within each flower colour group differs significantly from the overall distribution of sexual systems in the whole biotically pollinated flora including Asteraceae (n = 1499 species).

¹ including andromonoecious & gynomonoecious.

² dioecious and gynodioecious.

Sexual system, height, and geographic range of woody plants

While only 9.7% of herbaceous species are gender dimorphic (Table 2), 41.2% of woody species are gender dimorphic (Table 2), and over 60% of woody genera have at least one gender dimorphic species (Table 3). Furthermore, as we have shown, gender dimorphism in herbaceous species is concentrated within just two clades but is more widespread among woody genera. Woody species, therefore, lend themselves to further analysis.

Height and sexual system in woody species

Hermaphrodites show strong dominance only in the tallest size class (>20 m; Figure 2). Gender dimorphic species are abundant throughout the height class range, falling to low levels (<15%) only in the tallest size class. Gender dimorphism is common among shrubs (<5 m) mostly because of the large number of gynodioecious Pimelea and Veronica species. Many small (5–10 m) to mid-sized (10–20 m) trees are dioecious and across these two height classes, hermaphroditic and gender dimorphic species have virtually identical proportions (47% and 46%, respectively; Figure 2). Monoecious species are uncommon in any height class.

Figure 2. Sexual system versus height in the woody angiosperm flora of New Zealand.

Canopy position, dominance, and sexual system in woody species

Emergent or tall canopy trees in New Zealand are far less likely to be gender dimorphic than small or understorey trees (Figure 2). Monoecious and hermaphrodite trees play an outsize role therefore, dominating angiosperm forest basal area and to a lesser extent stem density (Figure 3). Monoecious species made up 50% of the basal area sampled across 796 plots yet were represented by just 12 species. Hermaphrodite species made up 36% of basal area drawn from 82 species, and gender dimorphic species were just 14% of basal area in spite of being represented by 84 species (χ² = 166.91, d.f. = 2, P < 0.0001). Monoecious species made up 27% of stems, hermaphrodite species 42%, and gender dimorphic species 31%, again, statistically non-proportional to the number of species in each group (χ²=  43.46, d.f. = 2, P < 0.0001).

Figure 3. Total basal area (m²) and stem count for all tree species sampled across 796 forest plots throughout New Zealand. D = gender dimorphic (n = 84 species sampled); H = hermaphrodite (n = 82 species sampled); M = monoecious (n = 12 species sampled).

Range size, latitudinal richness, and sexual system in woody species

Gender dimorphic woody species show a strong trend towards being more widely distributed and dominate in the 8–13° latitudinal range category (Figure 4). Being more widely distributed, they also tend to be well-represented in terms of species richness in all latitudinal bands, this being most accentuated north of latitude 42^o among small trees (5–15 m tall; Figure 5).

Figure 4. Range sizes of New Zealand tree species, plotted by sexual system. Range size is the total number of 1° latitudinal bands where a species naturally occurs. Fitted linear models with 95% confidence intervals are shown for each sexual system.

Figure 5. Richness of New Zealand tree species in latitudinal bands by height class and sexual system.

Speciation and gender dimorphism

Considering genera that are significantly gender dimorphic (i.e. ≥33% of species are gender dimorphic) and comparing those genera to all other genera, significantly gender dimorphic genera have near identical mean numbers of species per genus to all other genera (6.0 vs 5.8, respectively). The distribution of species richness per genus was statistically similar for significantly gender dimorphic genera and all other genera (Kolmogorov Smirnov, D = 0.07, P = 0.93) but 18.5% of significantly gender dimorphic genera had ≥10 species, compared to 14.0% of all other genera.

Within genera that contain tree species, the distribution of species richness per genus was statistically similar for cosexual and gender dimorphic lineages (Kolmogorov Smirnov, D = 0.11, P = 0.99), and for woody plants as a whole (Kolmogorov Smirnov, D = 0.10, P = 0.92). In contrast, in genera containing herbs, those with gender dimorphic members had fewer species per genus than genera that were cosexual (Kolmogorov Smirnov, D = 0.32, P = 0.006).

Place of origin of gender dimorphism

As New Zealand has low (8–12%) levels of endemism at the generic level (Garnock-Jones 2014), we follow the conservative assumption that gender dimorphism originated outside New Zealand if it is reported for close relatives within the genus elsewhere or is characteristic for closely related genera or the whole family. Webb et al. (1999) suggested that 17 (20%) out of the 83 genera of seed plants with gender dimorphism had evolved gender dimorphism within New Zealand, but changing generic limits (Garnock-Jones 2014) and increased phylogenetic resolution means this conclusion has to be revised. We reviewed recent publications of molecular phylogenies and re-assessed the origins of gender dimorphism in angiosperm genera currently recognised in New Zealand. Of the 78 angiosperm genera currently recognised in New Zealand as including species with gender dimorphism, it seems likely that in 14 (18%) of them gender dimorphism evolved in New Zealand (Table 8). Our reasons for changes from Webb et al. (1999) are given below.

Table 8. Genera in which it is probable that gender dimorphism evolved autochthonously in New Zealand.

Taxon	Gender types	Dimorphic/cosexual	Fruit type	Pollination syndrome	Life forms
Bulbinella	D, H	2/4	Dry	Biotic	Herb
Callitriche	D, M	1/3	Dry	Abiotic	Herb
Chionochloa	GD, H	1/23	Dry	Abiotic	Herb
Gentianella	GD H	1/29	Dry	Biotic	Herb
Coprosma	D, H	54/1	Fleshy	Abiotic	Tree, shrub
Lepidium	D, H	2/18	Dry	Biotic	Shrub, herb
Leptinella	D, GD, M	10/15	Dry	Biotic	Herb
Leucopogon	D, H	1/4	Fleshy	Biotic	Tree, shrub
Mentha	GD	1/0	Dry	Biotic	Herb
Pachycladon	GD, H	4/10	Dry	Biotic	Herb
Ranunculus	GD, H	1/44	Dry	Biotic	Herb
Toronia	D	1/0	Fleshy	Biotic	Tree
Urtica	D, M	2/3	Dry	Abiotic	Shrub, herb, vine
Veronica	D, GD, H	34/126	Dry	Biotic	Tree, shrub, herb

Note: Traits (fruit type, pollination syndrome, life form) apply only to New Zealand representatives. Gender types: D = dioecious; GD = gynodioecious; M = Monoecious; H = Hermaphrodite.

Mentha (Lamiaceae) was considered only dubiously autochthonously gender dimorphic but M. cunninghamii is within a clade of hermaphrodite Australian species (Bunsawat et al. 2004) and thus very likely developed gender dimorphism in New Zealand.

Aristotelia (Elaeocarpaceae) was thought to have been autochthonously gender dimorphic within New Zealand as dioecious New Zealand species of Aristotelia form a clade along with several cosexual eastern Australian species. However, the Australasian clade is a sister to a dioecious southern South American species Aristotelia chilensis, a separation dating back to the Oligocene (Coode 1985; Crayn et al. 2006; Valdivia and Simonetti 2007). We, therefore, suggest gender dimorphism did not evolve in New Zealand and that the Australian species are secondarily hermaphroditic.

The genus Clematis (Ranunculaceae) has scattered occurrences of gender dimorphism elsewhere, but none were thought to be closely related to the New Zealand taxa. However,

Clematis in New Zealand has a close dioecious Australian relative C. gentianoides (Miikeda et al. 2006) and therefore gender dimorphism cannot conclusively be accepted as autochthonous in this case.

Passiflora tetandra (Passifloraceae) of New Zealand is a member of the dioecious subgenus Tetrapathea – the other two species are in eastern Australia and New Guinea (Krosnick et al. 2009) – and thus is unlikely to have evolved gender dimorphism in New Zealand.

Leptecophylla and Acrothamnus were once included in Cyathodes which Webb et al. (1999) stated evolved gender dimorphism autochthonously. However, both genera include gender dimorphic species in Australia.

Rubus (Rosaceae)in New Zealand is embedded within the Australasian subgenus Micranthobatus which is entirely dioecious (Bean 1995; Alice and Campbell 1999).

Coprosma (Rubiaceae) is autochthonously gender dimorphic. Recent phylogenies of Coprosma make it highly likely that this dioecious genus diverged from its cosexual sister genus Nertera around 25 million years ago while confined to New Zealand, and later dispersed widely across the Pacific (Cantley et al. 2014).

Melicytus (Violaceae) most likely evolved in New Zealand and spread to adjacent regions (Mitchell et al. 2009) but its nearest relative, Anchietea, a small lianescent genus of South America, is dioecious (Wahlert et al. 2014). We, therefore, suggest it derives from a dioecious lineage.

Toronia (Proteaceae) has only recently been recognised as dioecious (Gardner 2008), and this endemic New Zealand genus has a close Australian relative in the Persoonia Rufiflora clade which is hermaphroditic (Holmes et al. 2018).

We agree with Webb et al. (1999) that the origin of gender dimorphism in the Aciphylla / Anisotome / Gingidia / Lignocarpa / Scandia clade (Apiaceae) is unclear. Mitchell et al. (2009) showed the origin of this dioecious and gynodioecious Apiaceae clade could be either Australia or New Zealand. A combined phylogenetic and molecular study puts the Australian species within the main Aciphylla and Anisotome grouping (Radford et al. 2001) and thus increases the probability that this clade is of New Zealand origin. However, the common ancestor of the Apiales was dioecious (Schlessman 2010) and the ancestor to the Aciphylla radiation may have been as well.

Discussion

At 13%, the incidence of dioecy in the New Zealand angiosperm flora is approximately double the global mean for angiosperms (c. 6–8%) (Renner 2014; Wang et al. 2021). As most compilations do not take account of gynodioecy, for global comparisons we include it, along with other forms of dimorphism, under the general designation of gender dimorphism. With one-fifth of the flora being gender dimorphic, New Zealand levels are about three times the global average. While some tropical forested areas, such as Singapore and Xishuangbanna have similar or somewhat greater levels of gender dimorphism, on a botanical region basis, New Zealand levels are rivalled only by the two Pacific archipelagos of Hawai'i and New Caledonia (Table 9). While woody species provide most (62%) of the gender dimorphic total, herbs also have exceptionally high levels of 7.4% (8.5% for genera with a dioecious member) while the global level sits at 2.7% (Wang et al. 2021). Taiwan (Tseng et al. 2008) and the Cape flora of South Africa (Steiner 1988) have similar levels to New Zealand but this is unusual. In the Chinese flora, herbaceous species average 4% dioecious (range 1%–9.8% across the country, highest in the tropical south) (Wang et al. 2020a), and in the tropical islands of Sri Lanka, Hawai'i and New Caledonia, herbs range from 1.3–4.6% dioecious (Sakai et al. 1995b; Senarath 2008; Schlessman et al. 2014).

Table 9. Incidence of the three major plant sexual systems in a selection of global angiosperm floras.

Location	References	N species	% trees	% woody	% hermaphrodite	% monoecious^1	% gender dimorphic^2
Carolinas, USA	Conn et al. (1980)	3274	12.1	25.9	–	–	3.5
British Isles	McComb (1966), Kay and Stevens (1986)	1377	3.8	–	80.7	16.3	3.7
Wisconsin, USA	Givnish et al. (2020)	1547	3.7	13.7	73.9	18.9	4.7
Changbai Mountain, China	Xia et al. (2020)	1403	5.1	17.5	81.6	13.3	5.1
Puerto Rico & Virgin Islands	Flores and Schemske (1984)	2027	22.7	50.8	78.9	14.9	6.1
Sri Lanka	Senarath (2008)	3529	21.9	48.8	82.7	10.3	7.1
Venezuela, Central Plain	Ramírez (2005)	210	14.3	26.7	75.2	17.1	7.6
Canadian Arctic	Kevan and Godglick (2016)	355	1.1	8.7	–	–	7.9
Taiwan	Lin et al. (2020)	3484	14.9	37.5	73.1	10.7	8.1
Chamela, Mexico	Bullock (1985)	708	26.6	49.9	70.2	16.9	13
Restinga, Brazil	Matallana et al. (2005)	566	18.7	43.2	75	11	14
Hawai'i	Sakai et al. (1995a)	971	26.8	65.1	62.4	16.0	18.5
New Zealand	This work	2080	11.4	31.3	66.3	14.2	19.5
Singapore	Sodhi et al. (2008)	1882	42.5	64.2	69.4	10.3	20.3
New Caledonia	Schlessman et al. (2014)	3051	46.7	76.0	61.2	17.5	21.2
Xishuangbanna, China^3	Chen and Li (2008)	685	57.5	88.7	60.6	14.3	25.1

¹ including andromonoecious & gynomonoecious.

² dioecious and gynodioecious.

³ forest only.

While the overall level of monoecy in the New Zealand angiosperm flora is similar to that prevailing globally, this is mainly due to the large monoecious genus Carex (Figure 6) and several andromonoecious Asteraceae genera (e.g. Brachyglottis, Raoulia). New Zealand differs strikingly from many other botanical regions in the lack of monoecious sensu stricto trees and shrubs (1.7%); the tree flora has only 9 monoecious species (3.4%). The five tall monoecious Nothofagaceae species dominate upland forests throughout, but the other four monoecious tree species (Ascarina lucida, Raukaua edgerleyi, R. simplex, and Ceodes ( = Pisonia) brunoniana) are short and only locally abundant (Figure 6). In contrast, in tropical and northern temperate forests up to 33% of tree species are monoecious (Sakai et al. 1995b; Gross 2005; Chen and Li 2008; Senarath 2008). In China, c. 25% of trees are monoecious, particularly the taller species (Wang et al. 2020a).

Figure 6. Key examples of sexual systems in New Zealand angiosperms. A, Myrsine salicina (Myrsinaceae). Gender dimorphism is rich among fleshy-fruited, subcanopy tree species. B, Aciphylla colensoi (Apiaceae). Gender dimorphism in herbaceous species is strongly concentrated in two genera, namely Aciphylla and Astelia. C, Astelia fragrans (Asteliaceae). D, Fuscospora cliffortioides (Nothofagaceae). Monoecy is uncommon in New Zealand woody plants, but is found in the five species of Nothofagaceae that dominate forest structure (48.4% of basal area nationally). E, Ascarina lucida (Chloranthaceae) is a rare example of a small monoecious tree. F, Carex decurtata (Cyperaceae). Monoecy is found in just 296 species but 66.7% of those are species of Carex. G, Pterophylla racemosa (Cunoniaceae) and H, Beilschmiedia tawa (Lauraceae). Tall and dominant forest species tend to be cosexual, and these two species contribute 20% of basal area nationally.

While the typical global associations of traits with sexual systems are evident in New Zealand (Webb et al. 1999), the key question is how did biogeographic processes lead to the exceptional gender dimorphic total and the diminished level of monoecy. These biogeographic processes are in turn dependent on the major geographic and climatic alterations of the archipelago and therefore we briefly outline them first.

Zealandia, the now largely submerged continental fragment of which the New Zealand archipelago forms part (Mortimer et al. 2017), began separating from Gondwana 83–79 Ma with the opening up of the Tasman Sea. By 55 Ma sea floor spreading had ceased and the New Zealand archipelago was, and remains, surrounded by oceanic gaps of at least 1500 km separation from neighbouring continents (Sutherland et al. 2020). Intermittent island arc and marine ridge connections to the north have persisted down to the present. The New Zealand landmass itself endured a prolonged marine transgression (40–22 Ma) in the Oligocene which left just a scatter of islands at its peak. Few New Zealand lineages predate this transgression and the current flora is almost entirely the result of post-25 Ma immigration mostly either down the northern island chains or directly across from southern Australia (McCarthy et al. 2021). Subsequent tectonic activity and marine regression resulted in an expansion and coalescence of the dry land area and from about 12 Ma ago, mountain building commenced which culminated in the construction of the tall axial mountain ranges by the beginning of the Pleistocene (Trewick and Bland 2012). Climates were near tropical in the early Cenozoic and warm temperate for most of the Miocene-Pliocene (Reichgelt et al. 2015); cooling in the late Miocene to Pleistocene saw extinction of many tree lineages (Lee et al. 2001, 2016) and arrival and radiation of herbaceous and shrub lineages (Heenan and McGlone 2019). Phylogenetic analysis of the flora supported by fossil evidence shows that 90% of current lineages are younger than 15 million years (Heenan and McGlone 2019; McCarthy et al. 2021).

Drivers of high gender dimorphism in the New Zealand angiosperm flora

A number of explanations for the anomalous level of gender dimorphism in New Zealand have been proposed, summarised by Webb et al. (1999) as ‘ … the relatively unspecialised pollinating fauna, the high proportion of perennial and of woody species, the greater dispersibility of fleshy-fruited species in a vertebrate biota dominated by birds, and the source of the flora … ’. Below we discuss the four explanations suggested by Webb et al. (1999), and then species radiations and gynodioecy as additional contributing factors. In evaluating these explanations, three considerations must be borne in mind. First, that genera consist of assemblages of species which usually vary little in their fundamental traits such as sexual system, habit, fruit type, etc. Immigrant selection (Lloyd 1985) or species radiation focused on one trait can therefore passively increase the incidence of another. Second, the distribution of plant biomes is largely controlled by climate and, given the strong correlations of plant habit with sexual system, we should expect the prevailing climate of a region to have an important influence. Third, as we are dealing with the entire flora, biogeographic considerations are paramount.

Pollination syndromes

Around 10% of angiosperm species are wind pollinated (Friedman and Barrett 2009) and abiotic pollination in general is strongly associated with gender dimorphism. As about 28% of the New Zealand flora is abiotically pollinated (similar to many oceanic islands); (Barrett 1996), it would seem a possibility that this is a factor in the high rate of gender dimorphism. Thirty five percent of the herbaceous flora is abiotically pollinated, but these contribute only 6% to the dimorphic total. Twelve percent of New Zealand woody species are abiotically pollinated, contributing a roughly proportionate 15% to the woody dimorphic species total. Moreover, more than 60% of these woody and abiotically pollinated species belong to a single genus, Coprosma. Abiotic pollination thus is not a significant factor (Higham and McQuillan 2000).

There is a strong association between small flowers and gender dimorphism (Bawa and Opler 1975; Ibarra-Manriquez and Oyama 1992). Gender dimorphism (and dioecy in particular) is thought to be a response to the fact that these small, open, unspecialised flowers favour a wide range of small pollinators which results in less targeted movement of pollen. If these small flowers are numerous, plants are exposed to a higher risk of geitonogamy (pollination between flowers of the same plant) and selfing (pollination within a flower) (Thomson and Thomson 1992). Dioecy prevents geitonogamy through pistillate and staminate flowers being borne on different individuals. It also permits sexual specialisation which can help lessen the impact of inefficient pollinators through optimisation of sex allocation and differential attractiveness of staminate and pistillate flowers (Käfer et al. 2017).

The prevalence of small, pale, white, or green flowers in the New Zealand flora and the relative paucity of large, complex bright flowers has been attributed to the lack of specialised pollinators such as hover flies, tabanid flies, and long-tongued bees – the bee fauna being made up of short-tongued Colletidae and small-tongued Halictidae – and the low number of butterfly species (Dugdale 1989; Godley 1979). The purple to magenta colours favoured by social bees elsewhere are notably scarce in New Zealand (Campbell et al. 2010). In this context, Lloyd (1985) argued that a number of New Zealand flowers have ‘despecialised’ because of the absence of pollinators they co-evolved with (e.g. Melicytus and many Orchidaceae).

We doubt that the abundance of small, dull, simple flowers in the New Zealand flora is driven by absence of specialised pollinators (although the scarcity of large complex blossoms probably is). For example, SW China and New Caledonia have a much wider and more abundant range of pollinators than New Zealand, including specialised pollinators such as long-tongued social bees, but have a similar proportion of small, white, yellow or yellow-green generalist flowers (Chen and Li 2008; Schlessman et al. 2014). A similar result holds for the North American tree flora where small-flowered trees make up c. 60% of the total compared with 53% in New Zealand (our data). Therefore, while strongly linked to gender dimorphism, the incidence of unspecialised, small, dull flowers in New Zealand does not seem to differ significantly from that in other floras and by itself cannot be a determining factor.

A factor not previously discussed which has a bearing on the overall incidence of gender dimorphism, is the very low level of ornithophilous flowers in New Zealand (1%) compared to other continental areas such as Australia where 15% of the angiosperm flora is ornithophilous (Lloyd 1985). Specialist ornithophilous flowers are almost never gender dimorphic; of genera of angiosperms globally that have at least some dioecious or gynodioecious members, only five are known to have ornithophilous flowers (Higham and McQuillan 2000). Species with ornithophilous flowers thus dilute the proportion of gender dimorphism in a given flora and conversely, their scarcity in New Zealand must boost the overall proportion. As with the small, dull flower syndrome, a lack of pollinators seem unlikely to explain the low levels of ornithophilous flowers as 11 indigenous bird species are recorded as visiting flowers and visit an unusually wide range of blossom types (Newstrom and Robertson 2005).

We suggest that habitat rather than the absence of specialised pollinators is a more likely explanation for the prevalence of small, dull flowers and the scarcity of large, showy flowers. The green, low irradiance of a forest understory is a problem for many visual pollinators and, because showy, complex and energetically expensive blossoms are disfavoured, generalist pollination syndromes prevail (Givnish 2010) and these favour dioecy through lower floral energetic costs (Givnish 1982). On the other hand, New Zealand lacks the open, semi-arid shrublands, prairies, steppes, grasslands, and deserts common in continental settings and which have a wide functional variety of flowers, a tendency towards brighter colours, and a greater number of species with large diameter corollas and high levels of hermaphroditism (Machado and Lopes 2004; de Lima et al. 2020). Bird pollination favours large, colourful flowers and often is common in such settings where open, sunlit canopies favour production of the sugar-rich nectar that attracts them (Armstrong 1979; Ford et al. 1979). Lineages with specialist ornithophilous flowers thus may have either failed to colonise or have faced high rates of extinction because of a lack of suitable habitat, and this may apply more generally to other specialised flower types (Lloyd 1985).

Fleshy-fruited species and bird dispersal

At a whole flora level, New Zealand has about the same level of fleshy-fruited species as other temperate areas (Lord 1999; Burrows 1994) and much less than in tropical floras. Fleshy fruit incidence alone therefore cannot explain the high levels of gender dimorphism in New Zealand. Of more importance is that those trees and shrubs in New Zealand that are fleshy-fruited are strongly gender dimorphic (71%). Tall (20 m +) trees in New Zealand are far less likely to be gender dimorphic than shorter statured woody species (Figure 2), and this appears to be generally true elsewhere, for instance in China (Wang et al. 2020a). Over 70% of New Zealand trees are <15 m tall (McGlone et al. 2010) and gender dimorphism is concentrated in the lower canopy (Figure 2). Bird-dispersed and gender dimorphic genera in New Zealand such as Coprosma, Fuchsia, Melicytus, Pseudopanax, Pittosporum, and Myrsine consist for the most part of mid-canopy or understorey trees and shrubs but have also diversified into open shrublands, taking with them the bird dispersal and gender dimorphic traits (Figure 6). Understory species often have bird-dispersed fruit as fleshy fruit and bird dispersal is favoured by shady environments and gap phase dynamics (Bolmgren and Lönnberg 2005). These ecological factors favour more efficient fruit dispersal and dioecy (Vamosi et al. 2007).

High proportion of perennial and woody species

Sexual system is strongly correlated with habit: herbs are more likely to be hermaphroditic than woody plants; woody plants, especially trees, have higher levels of gender dimorphism than herbs (Vamosi et al. 2003; Caruso et al. 2016; Käfer et al. 2017). Annual plants have very low levels of gender dimorphism: for instance, in the native flora of California, 37% of the herbs are annuals but only 0.13% of these are dioecious (Freeman et al. 1979). Annuals are scarce in New Zealand (c. 4%) and of the 83 species, just one is gender dimorphic (Althenia bilocularis, Potamogetonaceae). For comparison, the native flora of Wisconsin has 14.1% annual species (Givnish et al. 2020), North and South Carolina 20.6% (Conn et al. 1980), and the north-western Iberian peninsula 32% (Buide et al. 1998). While it is not possible to estimate accurately the effect of this scarcity, if the number of annual species was increased so as to make up 20% of the New Zealand flora, the arithmetical effect would be to reduce the level of gender dimorphism in the flora by 3%.

For a temperate region, the New Zealand flora has a large proportion of trees (11.4%; Table 2), comparable with subtropical Taiwan (14.8%) (Tseng et al. 2008). In contrast, trees make up c. 4% of the flora in northern North America and c. 2% in Europe (McGlone et al. 2010) and this difference alone must be a major contributor to the much higher gender dimorphism in New Zealand relative to those regions. On the other hand, the flora has relatively low percentage of trees compared with those prevailing in the tropics, for example in Hawai'i (21%) (Sakai et al. 1995b), Sri Lanka (22%) (Senarath 2008), Singapore (42.5%) (Sodhi et al. 2008) or tropical regions in south China (57.5%) (Chen and Li 2008). Lianas are abundant in New Zealand lowland forests (Jimenez-Castillo et al. 2007) but there are relatively few species. While there is a general trend, in that floras with a low percentage of trees and lianas have low gender dimorphism, tropical regions with very high concentrations of trees and lianas generally do not have a higher gender dimorphic percentage than in the New Zealand flora (Figure 7).

Figure 7. Percent gender dimorphism in the flora versus percent tree species in the flora, for 16 floras. The three Pacific archipelagos with high gender dimorphism (New Zealand, New Caledonia, and Hawai'i) are shown. Fitted linear model with 95% confidence intervals excludes New Zealand because it is an apparent outlier (%) flora ∼ 2.91 + (0.384 × % trees); F­_1,13 = 57.1, P < 0.0001, adjusted R² = 0.84. Data and data sources are provided in Table 3.

Source of the flora

Aside from a very few ancient lineages, nearly every species in the New Zealand angiosperm flora has an ancestor that at some stage made an ocean crossing. The herbaceous flora is younger than the woody flora and has been sourced from across the globe and the genera often are late Miocene or Plio-Pleistocene immigrants from temperate Eurasia (McGlone et al. 2018) and are associated with low levels of dioecy. On the other hand, about two-thirds of New Zealand woody genera are shared with northern Australia, New Caledonia, New Guinea, and the Indo-Malaysian region (McGlone et al. 2016). The complex, evergreen forests of New Zealand structurally resemble tropical forests (McGlone et al. 2016). Tropical forests have high levels of animal dispersed fruits, and bird dispersal is most common in midstorey and understorey trees and shrubs (Muller-Landau and Hardesty 2005). These short-statured tree and shrub lineages with small, bird-dispersed fruit may have been able to colonise the New Zealand archipelago much more easily than tall, slower-maturing trees. Their source has often been the tropical north and thus from a region with naturally high levels of gender dimorphism. For example, of the 86 lowland woody genera in New Zealand, 60% have representatives in tropical or subtropical regions and are highly likely to have been sourced from there. These tropical source lineages are more likely to be gender dimorphic than those New Zealand genera without these linkages (44% versus 35%). Once established, their dispersal characteristics may have assisted them to persist as evidenced by their tendency to have wide ranges (Figure 4).

Species radiations and autochthonous gender dimorphism

A disproportionately large percentage of the angiosperm species In New Zealand belong to just a few species-rich genera (Fenner et al. 1997). These genera are young and owe their current species richness to radiations driven by the massive climatic and tectonic alteration of New Zealand since the late Miocene. As we have shown above, while possession of gender dimorphism certainly did not prevent radiation, there is no convincing evidence that it accelerated it. The high levels of gender dimorphism in the flora must therefore be attributed in part to chance association with a radiating genus, rather than any inherent advantage of the syndrome. The Aciphylla / Anisotome / Gingidia clade (Apiaceae) and Astelia (Asteliaceae) herbaceous clades are an example of this in that they supply c. 70% of the dioecious and 59% of the gender dimorphic herb total (Figure 6). Most Aciphylla are long-lived, tall, stout herbs – that is shrub-like – and Astelia species are fleshy-fruited suggesting these features, which are relatively uncommon in herbs, are involved. Excluding the Aciphylla / Anisotome / Gingidia and Astelia clades drops the New Zealand level of herbaceous gender dimorphism to less than 5%, and the total flora level to 16%. A high level of dioecy (6.2%) among the herbs of the South African the Cape Flora has been explained similarly by a concentration in the Restionaceae – a large group of over 300 species in South Africa, all of which are dioecious and many long-lived and stout; when excluded, the flora-wide incidence of herbaceous dioecy falls to a 3% level typical of global herbaceous floras (Steiner 1988).

Local evolution did not play a large part in the establishment of gender dimorphism as less than 20% of New Zealand angiosperm genera possessing gender dimorphism appear to have evolved it autochthonously (Table 8). Although collectively over 70% of species in these 14 genera are cosexual, they provide a proportionate contribution to the gender dimorphic species total (23%). They differ from genera that evolved gender dimorphism elsewhere in that most are dry-fruited and many herbaceous (Webb et al. 1999) but Coprosma, which contributes 46% of the autochthonous gender dimorphic total, is both fleshy fruited and woody.

Gynodioecy

Gynodioecy plays an outsized role in New Zealand being very much more prominent than in the global flora (Renner 2014). It is regarded as one of the most likely transitional states from hermaphroditism to dioecy (monoecy being the other) (Torices et al. 2011). Shifts can occur rapidly from gynodioecy to dioecy but also reverse frequently (Rivkin et al. 2016). Several large New Zealand genera (Veronica, Pittosporum, and Leptinella) have both gynodioecious and dioecious members and it is likely that such transitions have taken place.

Self-compatibility and associated inbreeding depression is thought to be one of the main drivers for gynodioecy as a male-sterile form can establish in a population by producing outbred progeny provided there is adequate cross-pollination (Rivkin et al. 2016). The New Zealand flora has moderate levels of self-incompatibility (Newstrom and Robertson 2005) and two gynodioecious woody species investigated (Veronica alpina and Fuchsia excorticata) are self-compatible and show severe inbreeding depression (Delph and Lloyd 1996; Robertson et al. 2011) as has been demonstrated also for the Hawai'ian gynodioecious shrub Lycium carolinianum (Miller et al. 2019). As self-compatibility is advantageous during colonisation, it is possible that the high levels of gynodioecy in New Zealand and Hawai'i arise from a founder effect or immigrant selection. In support of this, the two most speciose genera with high levels of gynodioecy are the relatively recently arrived Pimelea and Veronica. Pimelea is widespread and reached New Zealand during the Plio-Pleistocene (Foster et al. 2018) via Australia where many Pimelea species are gynodioecious. Gynodioecy developed autochthonously in Veronica after the recent immigration of a self-compatible herb (Wagstaff et al. 2002). Therefore, the most likely explanation for the anomalously high levels of gynodioecy in the New Zealand flora, despite the conspicuously low levels in the source areas, is the recent arrival of several self-compatible immigrants followed by their rapid and very recent radiation.

New Zealand monoecy in a global context

The low level of monoecy in New Zealand woody plants is about as unusual as the high frequency of dioecious species but has not been commented on as far as we are aware. The abundance of Nothofagaceae (Figure 3), which occupy an ecological niche similar to that of monoecious Quercus and Fagus (both Fagaceae) in the Northern Hemisphere, and the success in New Zealand of the invasive, monoecious Acer pseudoplatanus (Sapindaceae) show there is no compelling environmental argument for the scarcity of monoecious tree species, and in particular tall monoecious trees in New Zealand. We instead suggest regional biogeography is responsible.

The monoecious tree genera that characterise the cool temperate forests of the Northern Hemisphere have never had a presence in the southwest Pacific. Monoecy in the tropical forests to the north of New Zealand is concentrated in just a few families (Arecaceae, Euphorbiaceae, Moraceae, Phyllanthaceae, and Sapindaceae) all poorly represented in New Zealand. These families provide clues as to why tree monoecy is rare in New Zealand. While a strong correlation between abiotic dispersal and monoecy in tropical tree floras has been claimed (Flores and Schemske 1984; Gross 2005), modelling suggests that regardless of dispersal mode, monoecious species are likely to have heavy, expensive fruit (de Jong et al. 2008) and this appears to be often the case in the Arecaceae, Moraceae, Euphorbiaceae, and Sapindaceae, many of which have large, vertebrate dispersed fruit. The small, frugivorous birds in New Zealand cannot match the rich array of mammalian and bird dispersers in the tropics and this may explain why New Zealand trees have seeds with a low mass compared with the tropics (geometric mean of 7 mg versus 100 mg, Moles et al. 2007). Figs (Moraceae), another largely tropical group with high levels of monoecy, faced yet another barrier in that they needed specialist wasp pollinators and only with the recent human-assisted arrival of these wasps in New Zealand have figs been able to spread naturally (Newstrom and Robertson 2005).

We, therefore, suggest that many potential monoecious tree colonists were disadvantaged by the low likelihood that their large fruit could cross ocean barriers and by lack of suitable vertebrate dispersers (or in the case of figs, specialist pollinators) on arrival.

Conclusions

The New Zealand angiosperm flora has one of the highest levels of dioecious species and genera in the world, and the highest incidence of gynodioecy. Consequently, gender dimorphism (total flora, c. 20%; woody species, c. 41%) is equal to or greater than that prevailing in nearly all botanical regions. It also has one of the lowest incidences of monoecy among trees (3.4%). The plant traits associated with these sexual systems in New Zealand show no significant deviation from those exhibited elsewhere, aside from the apparently unusual association of gynodioecy with woodiness. As less than 20% of the angiosperm genera in New Zealand evolved their sexual system locally, explanations for the high level of gender dimorphism and the lack of monoecy in trees, must centre not on the evolution of these sexual systems, but on biogeographic factors concerning environment, source, and immigration.

Previous discussion of gender dimorphism has mainly focussed on the fact that New Zealand lacks specialised insect pollinators, thus favouring small, unspecialised non-showy flowers, and transoceanic migration of fleshy-fruited, bird-dispersed trees and shrubs (Webb and Kelly 1993; Webb et al. 1999). Fleshy fruit and unspecialised flowers are both strongly linked with gender dimorphism. However, although New Zealand has few large, specialised flowers, in particular, ornithophilous flowers, it has about the same proportion of small, unspecialised flowers as many other regions, and has a relatively low proportion of fleshy-fruited trees and shrubs. Therefore, the incidence of these traits in New Zealand cannot by themselves explain the exceptionally high levels of gender dimorphism. We suggest that there are three major contributing factors to the high level of gender dimorphism in New Zealand.

First, a moist, oceanic climate has resulted in an extensive vegetation cover of dense, evergreen forest hosting numerous shrub and short-statured tree species. This environment, resembling in some respects that of the tropics, has proved ideal for a subset of fleshy-fruited trees and shrubs with exceptionally high levels of gender dimorphism. Many of these lineages were migrants from the tropical regions to the north of the New Zealand archipelago, their transit favoured by their small, bird-dispersed fruit.

Second, the moist oceanic climate also excluded steppe, savanna, and desert communities. The annual plants and the large, specialised, often ornithophilous, flowers typical of such open, dry environments are usually hermaphroditic. A New Zealand with such environments would have inevitably had have a lower proportion of gender dimorphic species in the flora as a whole.

Finally, we suggest that species radiations in a few gender dimorphic genera, in particular the Aciphylla / Anisotome / Gingidia lineage, has contributed to the high percentage of gender dimorphic species. Gynodioecy in late-arriving but rapidly speciating Pimelea and Veronica has also been a significant factor. As there is no evidence that gender dimorphism per se has promoted speciation in the flora, its chance association with radiating genera may also explain its high levels, especially as regards the herbaceous element.

The low level of monoecy in woody species appears to be directly linked to the absence of a handful of families. Those characteristic of the northern hemisphere (e.g. Fagales) are too geographically remote to have colonised, while tropical monoecy-rich families (e.g. Arecaceae, Moraceae, Phyllanthaceae) have traits (mammalian dispersal, heavy fruits, and specialised pollinators) which largely prevented them making the transoceanic crossing or establishing.

While the evolutionary factors promoting different sexual systems have now been well canvassed, and biogeographic and environmental factors addressed, the ecology of sexual system in the New Zealand angiosperm flora has been, aside from pollination studies, largely ignored. Ecological studies rarely even mention sexual system despite many involving forests and shrublands where most of the woody species are gender dimorphic. As possession of separate sexes must have a profound influence on fruit production, fruit dispersal, and regeneration, this lack of interest is both surprising and unfortunate. It is a field ripe for exploration.

Acknowledgements

Eric Godley was a pioneer in studies of sexual system in the New Zealand flora and David Lloyd advanced theoretical understanding of sexual system in a sequence of ground-breaking papers. We wish to acknowledge their contributions here, along with those of their students and colleagues Colin Webb, Lynda Delph-Lively, Linda Newstrom-Lloyd, and Alastair Robertson. We thank Angela Brandt for a detailed review of a draft of this manuscript. We acknowledge the use of data drawn from the Natural Forest plot data collected between 2009 and 2014 by the LUCAS programme for the Ministry for the Environment and the New Zealand Department of Conservation. Data were acquired using the National Vegetation Survey (NVS) databank (https://nvs.landcareresearch.co.nz/). Work was funded by Strategic Science Investment Funding for Crown Research Institutes from the Ministry of Business, Innovation and Employment. Photos in Figure 6 that are not by the authors: C, J. Barkla, https://inaturalist.nz/observations/107088651; E, A. Fergus, https://inaturalist.nz/observations/16682611; (f) https://inaturalist.nz/observations/111145836; H, T. Park, https://inaturalist.nz/observations/66972497.

Data availability

A full copy of the data are available from the Manaaki Whenua – Landcare Research Datastore website: https://doi.org/10.7931/qzf7-9q70.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Ministry for Business Innovation and Employment [Strategic Science Investment Fund to Manaaki Whenua].