key: cord-0010208-pjd6ovno
authors: Ríos, L. D.; Fuchs, E. J.; Hodel, D. R.; Cascante‐Marín, A.
title: Neither insects nor wind: ambophily in dioecious Chamaedorea palms (Arecaceae)
date: 2013-10-29
journal: Plant Biol (Stuttg)
DOI: 10.1111/plb.12119
sha: 1979ff91b55ef0dd1c0d85c7c6d20e168d13e020
doc_id: 10208
cord_uid: pjd6ovno

Pollination of Neotropical dioecious trees is commonly related to generalist insects. Similar data for non‐tree species with separated genders are inconclusive. Recent studies on pollination of dioecious Chamaedorea palms (Arecaceae) suggest that species are either insect‐ or wind‐pollinated. However, the wide variety of inflorescence and floral attributes within the genus suggests mixed pollination mode involving entomophily and anemophily. To evaluate this hypothesis, we studied the pollination of Chamaedorea costaricana,C. macrospadix,C. pinnatifrons and C. tepejilote in two montane forests in Costa Rica. A complementary morphological analysis of floral traits was carried out to distinguish species groups within the genus according to their most probable pollination mechanism. We conducted pollinator exclusion experiments, field observations on visitors to pistillate and staminate inflorescences, and trapped airborne pollen. A cluster analysis using 18 floral traits selected for their association with wind and insect pollination syndromes was carried out using 52 Chamaedorea species. Exclusion experiments showed that both wind and insects, mostly thrips (Thysanoptera), pollinated the studied species. Thrips used staminate inflorescences as brood sites and pollinated pistillate flowers by deception. Insects caught on pistillate inflorescences transported pollen, while traps proved that pollen is wind‐borne. Our empirical findings clearly suggest that pollination of dioecious Chamaedorea palms is likely to involve both insects and wind. A cluster analysis showed that the majority of studied species have a combination of floral traits that allow for both pollination modes. Our pollination experiments and morphological analysis both suggest that while some species may be completely entomophilous or anemophilous, ambophily might be a common condition within Chamaedorea. Our results propose a higher diversity of pollination mechanisms of Neotropical dioecious species than previously suggested.

Dioecious plants (i.e. with separated genders) comprise nearly 6% of angiosperm species and are distributed over 7% genera of flowering plants (Renner & Ricklefs 1995) . Pollination of dioecious plants is commonly attributed to generalist insects but approximately 30% of dioecious genera, mostly in temperate regions, are wind-pollinated (Bawa 1980; Renner & Ricklefs 1995) .

In tropical humid forests, palms (Arecaceae family) are conspicuous elements of the vegetation (Dransfield et al. 2008) and are predominantly composed of monoecious and hermaphroditic species; however, the dioecious genus Chamaedorea Willd. stands out as the largest Neotropical palm group, with 107 species (Govaerts et al. 2012) . Knowledge of the pollination systems in Chamaedorea has relied on observations of flower visitors and floral morphology that tentatively assigned species to an entomophilous (insect) or anemophilous (wind) system (see references in Henderson 1986) . Empirical studies on the pollination of Chamaedorea are relatively recent and only comprise around 6% of all the species. Results thus far have tended to categorise species as either entomophilus or anemophilous. Anemophily has been suggested for Chamaedorea pinnatifrons (Jacq.) Øerst. (Listabarth 1993) , Chamaedorea alternans H. Wendl. (Otero-Arnaiz & Oyama 2001) and Chamaedorea radicalis Mart. (Berry & Gorchov 2004) , while entomophily has been suggested for C. ernesti-augusti H. Wendl., C. oblongata Mart., C. neurochlamys Burret and Chamaedorea tepejilote Liebm. (Porter-Morgan 2007) .

Floral attributes and inflorescence architecture within Chamaedorea are associated with either wind or insect pollination. Species show bright or indistinctly coloured petals, dry-powdery or moist and sticky pollen, unscented or scented flowers with varying amounts of anatomical protection, and spicate or paniculate inflorescences with erect or pendulous rachillae (Hodel 1992a; Askgaard et al. 2008) . A combination of these floral traits within a single species is likely to allow successful pollination by both wind and insects.

The diversity in floral morphology within the genus suggests a combination of pollination modes, but previous experimental work classified Chamaedorea species as either entomophilous or anemophilous. Listabarth (1993) observed small thrips (Thysanoptera) and beetles (Coleoptera) visiting staminate inflorescences of C. pinnatifrons, but insects were rarely seen in pistillate flowers. Listabarth (1993) proposed that the insect's activity facilitated the release of pollen from flowers where wind could more easily transport it to pistillate flowers, in a system he named 'insect-induced wind pollination'. Otero-Arnaiz & Oyama (2001) also classified C. alternans as windpollinated, however reported activity of thrips on staminate inflorescences. Conversely, Porter-Morgan (2007) argued that thrips were the main pollinator of four Chamaedorea species and gave wind a secondary role. These authors, however, did not consider the possibility of a mixed pollination syndrome.

The simultaneous occurrence of insect and wind pollination, or ambophily, is documented in several angiosperm families, such as Ericaceae, Euphorbiaceae, Polemoniaceae, Ranunculaceae and Salicaceae (Faegri & van der Pijl 1979; Culley et al. 2002; Duan et al. 2009; Yamasaki & Sakai 2013) . Within Arecaceae, this system has been documented in seven species (Barfod et al. 2011) . In ambophilous plants, the role of pollination vectors varies, from species predominantly insect-pollinated to mostly wind-pollinated species, with most species showing intermediate frequencies (Culley et al. 2002) .

In this study, we aimed to understand the role of wind and insect pollination in four Chamaedorea species. The genus is a diverse group with a broad geographic distribution, from Central Mexico to Bolivia, and is more abundant and diverse in the understorey of middle elevation forests between 800 and 1500 m (Hodel 1992a) . Most species are restricted to the Mesoamerican region, where two centres of diversity are recognised: southern Mexico-Guatemala and the mountains of Costa Rica and Panama (Hodel 1992a; Henderson et al. 1995) .

Our specific research goals were to: (i) experimentally evaluate the contribution of insect and wind pollination to pistillate reproductive success of four Chamaedorea species in Costa Rica (ii) describe the pollination mechanism associated with entomophily; and (iii) identify potential entomophilous and anemophilous groups within the genus, as well as groups with a combination of the two pollination modes, inferred through a morphological analysis of floral traits on a sample of 52 Chamaedorea species.

The selected species, Chamaedorea costaricana Øerst., Chamaedorea macrospadix Øerst., C. pinnatifrons and C. tepejilote, have three to six (eight) pinnate leaves per stem; branched, racemose inflorescences with several rachillae; small (<5 mm), greenish to yellow, scented flowers; dry-powdery pollen; and black fruits at maturity (Hodel 1992a) . C. costaricana has multiple stems up to 8-m high and occurs from southern Mexico to western Panama, from 500 to 2350 m a.s.l.; C. tepejilote develops a single erect (rarely multiple) stem 5-m high and occurs from southern Mexico to Colombia, from 0 to 1600 m a.s.l.; C. macrospadix has an initially acaulescent habit but typically later forms a short stem and is restricted to Costa Rica and western Panama, from 100 to 1500 m a.s.l.; C. pinnatifrons develops single stems to 4-m high and has the widest geographic distribution, from southern Mexico to Bolivia in South America, from (100) 400 to 2600 m a.s.l. (Hodel 1992a; Grayum 2003) . Voucher specimens were deposited at the herbaria CR and USJ (Cascante 995, 996, 1166; R ıos 16, 21, 22, 23) .

Our study was conducted in two humid tropical montane forests of the Central Valley of Costa Rica. El Rodeo forest is located in the western part of the valley (9º54′25″N, 84º16′30″ W, 900-1000 m a.s.l.). It is a premontane wet forest (sensu Holdridge 1967) on hilly terrain where C. costaricana, C. macrospadix and C. tepejilote grow sympatrically. El Rodeo experiences a marked seasonality: a dry season from December to March, with mean annual precipitation of 2467 mm and mean annual temperature of 23.4°C (Instituto Meteorol ogico Nacional undated). All three Chamaedorea species were located within two 1-ha permanent plots established to monitor forest dynamics (Cascante-Mar ın et al. 2011). Our second site, La Carpintera forest (9°53′20″ N, 83°58′10″W) is a lower montane moist forest at 1750 m a.s.l. in Cartago Province. The forested area, mostly old secondary growth, covers nearly 2400 ha and is frequently covered with cloud (S anchez et al. 2008) . La Carpintera experiences a mild dry season from January to April, with mean annual precipitation of 1563 mm and mean annual temperature of 16.6°C (Instituto Meteorol ogico Nacional undated). C. pinnatifrons, which is relatively abundant and the only palm species in the understorey at La Carpintera, flowers from February to June.

We implemented conventional pollinator exclusion experiments on pistillate inflorescences and compared fruit set (i.e. ratio of fruits to flowers) per inflorescence to control inflorescences (Kearns & Inouye 1993) . We performed the experiments during two flowering seasons, 2005 -2006 and 2010 at El Rodeo and 2011 at La Carpintera. A total of 112 pistillate inflorescences of C. costaricana, 56 of C. macrospadix, 131 of C. pinnatifrons and 141 of C. tepejilote were distributed among the following pollination treatments: (i) insect exclusion using white mesh cloth bags with pores 150-200 lm in size and sprayed with domestic insecticide (active ingredient: pyrethroids) to deter insect visitation but that allowed windborne pollen to enter (pollen size 20-25 lm) (ii) control inflorescences without any manipulation and exposed to wind and insect pollination; and (iii) total exclusion using paper and plastic bags to test for apomixis. Bags covered the inflorescences for 6-8 weeks after anthesis to reduce fruit loss due to predation. Fruits were counted when they had reached nearly one-third of their final size.

Kruskal-Wallis and Mann-Whitney tests with a Bonferroni correction (K = 3, P = 0.0167) were used to compare fruit set among treatments per species. All analyses were performed using R 2.15.1 (R Core Development Team 2012).

We recorded insect visitors from inflorescences of both genders of all species. Visitation frequency of insects on pistillate inflorescences was surveyed twice a week in the flowering period of 2011-2012 in a sample of 164 pistillate inflorescences of C. costaricana, 82 of C. macrospadix, 140 of C. pinnatifrons and 159 of C. tepejilote. We collected insects by placing the inflorescence at anthesis in a plastic bag and gently shaking it to detach insects, which were preserved in 70% ethanol for later identification by a specialist.

Because preliminary observations indicated that thrips (Thysanoptera) were the most abundant flower visitors, we assessed pollen loads on thrips visiting pistillate flowers from five inflorescences of each species (one in C. macrospadix). All thrips from each pistillate inflorescence were placed in separated vials with 1 ml 70% ethanol. In the laboratory, the insects were mounted on glass slides and examined using a light microscope at the highest magnification (10009) to count pollen grains on their bodies. The ethanol solution was also checked for pollen grains that may have detached from the insects. Then, we estimated the mean number of grains per thrip by considering the total number of pollen grains and thrips per vial.

To evaluate the putative role of staminate inflorescences as brood sites (sensu Sakai 2002), we collected thrips from staminate inflorescences in anthesis and senescence (4-7 days after anthesis). Thrips were counted and classified according to their development stage (i.e.. adults or juveniles: instar I to pupae). We sampled 23 inflorescences (12 in anthesis and 11 in senescence) of C. tepejilote, 27 (14 and 13) of C. costaricana, and 15 (nine and six) of C. macrospadix during the 2010-2011 flowering season, and 16 (eight and eight) inflorescences of C. pinnatifrons during the 2011 season.

We also investigated if insects assisted the release of pollen from the anthers in species with staminate flowers with apically connate petals (sensu Listabarth 1993). We bagged five staminate inflorescences near anthesis of C. macrospadix and C. pinnatifrons with transparent plastic bags. These bags excluded thrips and were placed before insects visited inflorescences. The enclosed inflorescences were periodically checked during anthesis to determine if pollen was spontaneously released from the anthers and accumulated outside the flowers or on the inner bag surface.

Traps to collect wind-borne pollen were placed in four randomly selected stations in the understorey of both study sites in 2010-2011 (only 2011 for La Carpintera). Each pollen trap consisted of a microscope slide coated (ca. 2 cm 2 ) with melted fuchsin stained glycero-gelatin on one side (Kearns & Inouye 1993) . Each station contained three traps hanging down from vegetation at 1.5-1.8 m above the ground. Pollen traps were separated by at least 1 m from each other and were left in place for 3-4 days. New traps were set every 3-4 days at El Rodeo and weekly at La Carpintera. Pollen grains from Chamaedorea were identified and counted with a light microscope using a reference collection. Chamaedorea pollen is readily identifiable; however, because pollen morphology and size are comparable among all four studied species, it was not feasible to identify pollen to the species level.

To estimate the prevalent pollination syndromes within the genus, we surveyed 18 discrete floral traits grouped according to inflorescence morphology, which included attributes related to aerodynamic properties of wind-borne pollen (Niklas 1992; Friedman & Harder 2004) , flower attributes and pollen type traits from 52 Chamaedorea species. We tallied the presence/ absence of characters commonly associated with the wind or insect pollination, according to theoretical syndromes following Faegri & van der Pijl (1979) and Proctor et al. (1996) . Information on floral traits was obtained from Hodel (1992a) and additional sources: Douglas (1988) , Hodel (1992b Hodel ( , 1995 , Hodel et al. (1995) , Grayum (1998 Grayum ( , 2003 , Otero-Arnaiz & Oyama (2001), Berry & Gorchov (2004) , Porter-Morgan (2007), Askgaard et al. (2008) , and notes from herbaria specimens deposited at the National Herbarium of Costa Rica (CR), National Institute of Biodiversity (INB), Herbarium of the University of Costa Rica (USJ) and personal field observations.

Similarity among species was calculated using the simple matching index (Krebs 1999) for binary data. Then, a cluster analysis was carried out to group species according to their floral similarity, which putatively reflects a particular pollination syndrome, using the unweighted paired group method (UPGMA; Legendre & Legendre 2012) . In order to identify the floral traits responsible for the differences among groups generated by the cluster analysis, we carried out a similarities percentages (SIMPER) analysis using the Bray-Curtis index (Clarke & Warwick 1994) . These analyses were performed using PAST, version 2.16 (Hammer et al. 2001) .

Fruit set under the control treatment (i.e. inflorescences exposed to wind and insects) ranged from 27% in C. pinnatifrons and 34% in C. macrospadix, to 50-55% in C. costaricana and C. tepejilote (Fig. 1) . In all four studied species and reproductive years, the control treatment produced consistently more fruits than the insect exclusion treatment alone (all P-values <0.001 after Mann-Whitney test). Fruit set from insectexcluded inflorescences varied among species and flowering seasons, from nearly 7% in C. pinnatifrons to 22% in C. tepejilote (Fig. 1) . Within species, fruit set from wind pollination was significantly higher than fruit set from apomixis in most cases (Mann-Whitney test P < 0.001, except in C. tepejilote in 2005-2006 and C. macrospadix in 2011 P > 0.05). Pollen caught on pollen traps confirmed that wind transports Chamaedorea pollen (Fig. 2) . A mean of 86 AE 11 (AESD) pollen grains per trap was counted at El Rodeo, while at La Carpintera, we found 16 AE 4 grains per trap. Fruit development by putatively apomictic means was detected in all but one species (C. pinnatifrons), but varied among species and flowering seasons (Fig. 1) , and overall it was <11%. Seed viability tests on apomictic fruits were not performed.

Thrips were the most frequent and abundant flower visitors of all studied species, particularly on staminate inflorescences, at both sites (Fig. 3) . In a single staminate inflorescence, we counted more than 2000 individuals belonging to Brooksithrips chamaedoreae (Thripidae), a species described from our study site of El Rodeo (Retana-Salazar & Mound 2005) . Thrips were seen visiting 12% of the pistillate inflorescences of C. costaricana, 9% of C. macrospadix, 24% of C. pinnatifrons and 26% of C. tepejilote. Infrequent visitors to staminate inflorescences were the stingless bees Trigona fulviventris and Plebeia sp. at El Rodeo and Partamona orizabaensis (Apidae: Meliponini) at La Carpintera. These insects were not recorded visiting pistillate inflorescences. Other visitors on C. pinnatifrons staminate inflorescences were staphylinid, chrysomelid (Demotispa sp., Tetragonotes sp.) and cryptophagid beetles. Both Tetragonotes sp. and the unidentified cryptophagid mated in staminate inflorescences, but were rarely seen on pistillate flowers.

Thrips used staminate inflorescences of the four Chamaedorea species as brood sites, although we also observed juvenile thrips on pistillate inflorescences, but this was an uncommon event (n = 3). Adult thrips fed on pollen, mated and laid eggs on the adaxial face of the petals. It is noteworthy that in C. tepejilote, thrips not only fed on pollen but also preyed upon the tissue of the inflorescence rachis (Fig. 3) . After a few days, we observed juvenile thrips on senescent staminate inflorescences (Fig. 3, Table S1 ).

A total of 259 Chamaedorea pollen grains from 221 thrips were quantified on pistillate inflorescences of C. tepejilote, 212 grains from 200 thrips on C. costaricana and 400 grains from 224 thrips on C. pinnatifrons. Thus, on average, each thrips transported 1.0 AE 0.6 grains on C. tepejilote, 1.1 AE 0.7 on C. costaricana and 2.5 AE 2.9 on C. pinnatifrons; no pollen was found on the nine thrips caught from the C. macrospadix inflorescence.

Thrips are not required for pollen release from the anthers of flowers with apically connate petals of C. pinnatifrons and C. macrospadix. We observed pollen deposited outside the corollas and between flowers a few days after anthesis on inflorescences where insect visitation was prevented. 

Five groups were identified in the cluster analysis on floral traits, with an overall average dissimilarity of 49.7% (Fig. 4) . The inflorescence and floral traits that contributed more to separating groups included the length of the inflorescence peduncle, the number of rachillae, staminate flower showiness and anther position; these traits explained 53.1% of the dissimilarity. The traits with the lowest contribution to distinguishing groups included the texture of the rachis, pollen type, presence of scented staminate flowers and position of pistillate rachillae; these traits explained 10.5% of the differences. Noteworthy, none of the groups were entirely described by anemophilous or entomophilous traits alone; instead, most species possessed a combination of inflorescence and flower characteristics adapted for wind as well as insect pollination. Species in groups 1 and 3 (Fig. 4) have a prevalence of traits more conducive to wind and insect pollination, respectively. In group 1, species exhibited inflorescences with long peduncles (traits C and D), many rachillae (traits G and H), flowers not showy (traits K and L), exposed and divided stigmas (traits M and N) and exerted anthers (trait O); these traits are congruent or suggest anemophily. In contrast, species in group 3 exhibited traits more adapted to entomophily (Fig. 4) , e.g. inflorescences with short peduncles, few rachillae, showy flowers, entire and inserted stigmas and inserted anthers.

Species in the other groups (2, 4 and 5; Fig. 4 ) showed different ranges of combinations of entomophilous and anemophilous floral traits, either at the inflorescence or flower level. Three of our study species (C. costaricana, C. tepejilote and C. pinnatifrons) were placed among species with a combination of floral traits (group 2), while C. macrospadix was placed in group 1, an anemophilous group.

Our empirical findings from pollination experiments show that wind and insect pollination co-occur in our four studied species. Floral trait analyses suggest that pollination mechanisms within the dioecious palm genus Chamaedorea do not constitute a dichotomy between entomophilous and anemophilous species, since most species exhibit a combination of floral traits for both insect and wind pollination. While some Chamaedorea species may be completely entomophilous or anemophilous, most analysed species have floral traits that simultaneously allow for the transport of pollen by both vectors, a system termed 'ambophily' (reviewed in Culley et al. 2002) . These results suggest that based on morphological traits, a large number of species may be ambophilous. This type of mixed pollination should be explicitly considered in future experiments of pollination biology in Chamaedorea.

Our field experiments suggest that thrips (Order Thysanoptera) mainly carry out entomophilous pollination in the studied Chamaedorea spp. However, the role of other visiting insects cannot be discounted because pollination studies in the literature barely include 6% of all Chamaedorea spp. Also, other insect groups might pollinate species with wide geographic distribution or those found in diverse habitats (Herrera et al. 2006) . Thrips have been considered unreliable pollinators due to their limited flight ability and their apparently low specialisation for hosts (Kirk 1985a (Kirk , 1997 , but they appear to be effective pollinators in our study sites. The thrips pollination system in Chamaedorea fits the brood-site type of 'pollen parasites' described by Sakai (2002) and discussed by Porter-Morgan (2007) . Insects are probably attracted to the scented staminate flowers (Kirk 1985b ) even before anthesis, where they mate and feed on pollen and plant tissue and oviposit within the small flowers. Most of the 52 Chamaedorea species analysed (96%) had scented staminate flowers. Consumption of vegetal tissue was more evident on staminate inflorescences, particularly in C. tepejilote; this consumption might be related to a low amount of anatomical defences. The increasing number of juvenile thrips found on drying staminate inflorescences indicates that immature insects develop on senescent flowers within a few days. Later, juvenile insects probably fall to the soil to complete their development (Lewis 1973) . We found that thrips indeed transported pollen on their bodies, and volatile compounds (mainly sesquiterpenoids in C. tepejilote; Kaiser 1997) probably attract them to the pistillate flowers. Nonetheless, in our floral trait analysis only 31% of the species had scented pistillate flowers, which suggests that if thrips are the main pollinators of most of the analysed Chamaedorea, they also might locate pistillate flowers through visual cues. Pollen deposition onto the stigma of pistillate flowers of Chamaedorea is likely mediated by a 'mistake mechanism' (Dafni 1984) . Pistillate flowers are morphologically similar to their staminate counterparts and imitate the oviposition substrate. However, they do not provide a brood site inside the small pistillate flowers with reduced perianth parts and lack pollen as food. Also, and although not studied here, septal nectaries to provide nectar as reward have not been detected in flowers of Chamaedorea (Askgaard et al. 2008) . Pollination occurs when thrips arrive at pistillate flowers and do not stay for a long time, explaining the observed lower number of insects on pistillate inflorescence compared to staminate inflorescences. Before leaving, their activity on the inflorescence provides the opportunity for pollen to be deposited on the exposed stigmas and to develop the one-seeded fruit. Alternatively, pistillate inflorescences might inhibit thrips oviposition or larval development due to a possible larger anatomical defences (Uhl & Moore 1973) or active inhibition of larval development (Dufa€ y & Anstett 2004) .

Insects, particularly beetles, bees and flies, are generally assumed to pollinate palms. While thrips are occasionally mentioned as flower visitors, their potential role as palm pollinators has not been fully recognised (Kirk 1997; Barfod et al. 2011) . Some previous pollination studies in Chamaedorea may have overlooked the importance of thrips. For instance, the supposedly anemophilous C. alternans (Otero-Arnaiz & Oyama 2001) might also experience entomophily because the authors reported large amounts of thrips on staminate inflorescences. Listabarth (1993) suggested an indirect role of thrips in the pollination of C. pinnatifrons; however, our observations show that thrips have a direct role in pollen transport to pistillate flowers. The role of thrips as pollinators is documented in several plant groups with different taxonomic affinities, such as Zamiaceae (Mound & Terry 2001) , Annonaceae (Gottsberger 1994; Momose et al. 1998) , Dipterocarpaceae (Appanah & Chan 1981; Kondo et al. 2011) , Euphorbiaceae (Fiala et al. 2011) , Moraceae (Sakai 2001; Zerega et al. 2004 ) and Ericaceae (Hagerup & Hagerup 1953; Garc ıa-Fayos & Goldarazena 2008) .

In our study sites, anemophily appears to contribute significantly to the pollination biology of Chamaedorea. Initially, palms were considered anemophilous due to their floral morphology (i.e. small flowers, high pollen-ovule ratio and small pollen grains), but this view has changed to consider palms as mainly entomophilous (Henderson 1986; Barfod et al. 2011) ; however, several studies have documented the role of insects and wind in palm pollination (Anderson et al. 1988; Scariot et al. 1991; Consiglio & Bourne 2001; Melendez-Ramirez et al. 2004) . Our pollination experiments showed that fruit set in the studied Chamaedorea species is the result of a mixed pollination system. In addition to thrips, Chamaedorea pollen grains can also be wind-transported, as our pollen trapping experiment documented. Pollen release from the anthers in species with apically connate petals, such as C. pinnatifrons, might occur by spontaneous means and not necessarily need the intervention of thrips, as Listabarth (1993) suggested. Consequently, for species like C. tepejilote and C. costaricana with apically separated petals, pollen release into the wind is easily achieved without insect intervention. Pollen movement by wind in Chamaedorea can be a common phenomenon, even in the relatively low-wind forest understorey, because the majority (85%) of the species selected for our analysis of floral traits possessed dry and powdery, easily wind-transported pollen.

Combining wind and insect pollination might increase pollen transfer and is most advantageous when the reliability or effectiveness of insect pollinators varies in space and time (Culley et al. 2002; Duan et al. 2009; Yamasaki & Sakai 2013) . Having a dual mixed pollination system is expected to increase the reproductive success of plants (Costa & Machado 2012) and therefore may represent a selective advantage. This enhanced success is particularly important for species in conditions where insect pollinators are scarce or species compete for pollinators (references in Culley et al. 2002) . The effectiveness of both pollination systems in Chamaedorea is also likely to depend on floral traits that are adapted to both wind and insect pollination (Yamasaki & Sakai 2013 ) and very likely vary according to species, habitat and flowering season (Culley et al. 2002) . In our study species, wind pollination significantly contributed to overall fruit set. We are however likely to underestimate the role of wind because the bag itself probably reduced somewhat the inflow of wind-borne pollen (Kearns & Inouye 1993 ). In the ambophilous palm Attalea speciosa Mart. ex Spreng. (studied as Orbignya phalerata Mart.), anemophily was more important in open sites or pastures compared to forest conditions (Anderson et al. 1988) . In Cocos nucifera L., ambophily is recognised but the relative importance of wind and insect as pollen vectors might depend on wind conditions (Melendez-Ramirez et al. 2004) . In Chamaedorea, ambophily may also contribute to long-distance pollen dispersal because thrips dispersal might be limited (Kondo et al. 2011) .

Floral trait analysis showed that some Chamaedorea species might have flower traits that are congruent with wind pollination, while others have more insect pollination attributes. In our analyses, however, most species showed a combination of characters from both pollination mechanisms, which may favour ambophily. In fact, three of our study species are clustered in a group with species having both anemophilous and entomophilous floral traits. However, C. macrospadix was placed by the cluster analysis in a group with floral characters adapted for anemophily, while our pollination experiments showed that insect pollination played an important role. In this species both staminate and pistillate flowers are scented, which may promote insect visitation. In addition, the flowering period of C. macrospadix in our study population comprises the second peak of the rainy season, which may constraint the role of wind pollination (data not shown). This also indicates that the classification suggested by our morphological analysis needs to be confirmed empirically through pollination studies.

Several other ambophilous angiosperms show a similar phenomenon of intermediate floral morphology between wind and insect pollination, with some traits typical of wind pollination and other traits primarily of insect pollination (Karrenberg et al. 2002; Costa & Machado 2012; Yamasaki & Sakai 2013) . However, it remains to be determined whether ambophily in Chamaedorea represents a stable reproductive strategy or an intermediate step in the transition to full wind or insect pollination (Culley et al. 2002; Friedman & Barrett 2009 ). It is important to note that we only analysed 52 Chamaedorea species, and other pollination groups may arise if traits from all species are studied.

Chamaedorea palms grow in the understorey of undisturbed humid Neotropical forests and represent conspicuous elements of the vegetation at these sites. As dioecious species, Chamaedorea palms are more sensitive to the detrimental effects of habitat disturbance, deforestation and fragmentation on breeding dynamics (Somanathan & Borges 2000) . Information on the breeding system is necessary for the development of efficient conservation strategies, because they impact gene flow patterns, which in turn affect levels of genetic diversity and population structure (Loveless & Hamrick 1984) . This study aimed to increase knowledge about the pollination of this palm group, while trying to elucidate the existence of complex pollination mechanisms. Moreover, our results show a higher diversity in pollination mechanisms of Neotropical dioecious species than previously suggested by other authors (e.g. Bawa & Opler 1975; Renner & Ricklefs 1995) . We clearly show ambophily as the most likely pollination system for the four studied Chamaedorea species in Costa Rica. Future research directions on this topic might evaluate the role of both pollination mechanisms along a habitat or geographical gradient or in isolated populations in forest fragments. respectively; Paul Hanson (University of Costa Rica), Angel Sol ıs (National Institute of Biodiversity, INBIO) and Laurence A. Mound (Commonwealth Scientific and Industrial Research Organisation, CSIRO) of Australia for insect identification; field assistants for their valuable help and the comments and suggestions from two anonymous reviewers that increased the quality of the manuscript. This work was supported by Vicerrector ıa de Investigaci on at Universidad de Costa Rica [grant 111-BO-223 awarded to ACM and EJF].

Additional Supporting Information may be found in the online version of this article: Table S1 . Mean thrips abundance in staminate inflorescences on anthesis and senescence in the four studied Chamaedorea species.

Pollination ecology of a forest-dominant palm (Orbignya phalerata Mart.) in Northern Brazil

Thrips: the pollinators of some dipterocarps

Floral structure in the Neotropical palm genus Chamaedorea (Arecoideae, Arecaceae)

Twentyfive years of progress in understanding pollination mechanisms in palms (Arecaceae)

Evolution of dioecy in flowering plants

Dioecism in tropical forest trees

Reproductive biology of the dioecious understorey palm Chamaedorea radicalis in a Mexican cloud forest: pollination vector, flowering phenology and female fecundity

Tree turnover in a premontane Neotropical forest

Change in Marine Communities: an approach to statistical analysis and interpretation

Pollination and breeding system of a neotropical palm Astrocaryum vulgare in Guyana: a test of the predictability of syndromes

Flowering dynamics and pollination system of the sedge Rhynchospora ciliata (Vahl) K€ ukenth (Cyperaceae): does ambophily enhance its reproductive success?

The evolution of wind pollination in angiosperms

Mimicry and deception in pollination

Notes on pollinating the Eleutheropetalum group of Chamaedorea. Principes

Genera Palmarum, the evolution and classification of palms

Insect and wind pollination of an alpine biennial Aconitum gymnandrum (Ranunculaceae)

Cheating is not always punished: killer female plants and pollination by deceit in the dwarf palm Chamaerops humilis

The principles of pollination ecology

Pollination systems in pioneer trees of the genus Macaranga (Euphorbiaceae) in Malaysian rainforests

Wind of change: new insights on the ecology and evolution of pollination and mating in wind-pollinated plants

Inflorescence architecture and wind pollination in six grass species

The role of thrips in pollination of Arctostaphyllos uvaursi

As Anon aceas do cerrado e a sua polinizac ßão

World checklist of Arecaceae. The Board of Trustees of the Royal Botanic Gardens

Nomenclatural and taxonomic notes on Costa Rican palms (Arecaceae), with five new species

Manual de plantas de Costa Rica. Volumen II: Gimnospermas y Monocotiled oneas (Agavaceae-Musaceae). Monographs in Systematic Botany from the Missouri Botanical Garden

Thrip pollination of Erica tetralix

PAST: palaeontological statistics software package for education and data analysis

A review of pollination studies in the Palmae

Field guide to the palms of the Americas

Geographical context of floral evolution: towards an improved research programme in floral diversification

Chamaedorea palms: the species and their cultivation

Additions to Chamaedorea palms: new species from Mexico and Guatemala and miscellaneous notes

Three new species of Chamaedorea from Panama

Two new species of Chamaedorea from Honduras

Life zone ecology

New or uncommon volatile compounds in the most diverse natural scents. Rivista Italiana Eppos

Pollen vectors and inflorescence morphology in four species of Salix

Techniques for pollination biologists

Pollen-feeding and the host specificity and fecundity of flower thrips (Thysanoptera)

Effect of some floral scents on host finding by thrips (Insecta: Thysanoptera)

Feeding

Can tiny thrips provide sufficient pollination service during a general flowering period in tropical rainforests

Ecological methodology

Numerical ecology

Thrips: their biology, ecology and economic importance

Insect-induced wind pollination of the palm Chamaedorea pinnatifrons and pollination in the related Wendlandiella sp

Ecological determinants of genetic structure in plant populations

Mixed mating strategies and pollination by insects and wind in coconut palm (Cocos nucifera L. (Arecaceae)): importance in production and selection

Thrips cross-pollination of Popowia pisocarpa (Annonaceae) in a lowland dipterocarp forest in Sarawak. Biotropica

Thrips pollination of the Central Australian Cycad, Macrozamia macdonnellii (Cycadales)

Plant biomechanics: an engineering approach to plant form and function

Reproductive phenology seed-set and pollination in Chamaedorea alternans, an understorey dioecious palm in a rain forest in Mexico

Thrips as primary pollinators of sympatric species of Chamaedorea (Arecaceae) in

The natural history of pollination

Foundation for Statistical Computing, R Core Development Team

Dioecy and its correlates in the flowering plants

Character state variation in a new genus and species of Thripidae (Insecta: Thysanoptera) associated with Chamaedorea (Arecaceae) inflorescences in Central America

Thrip pollination of androdioecious Castilla elastica (Moraceae) in a seasonal tropical forest

A review of brood-site pollination mutualism: plants providing breeding sites for their pollinators

Diversidad de plantas, mam ıferos y mariposas en los cerros de La Carpintera

Reproductive biology of the palm Acrocomia aculeata in Central Brazil

Influence of exploitation on population structure, spatial distribution and reproductive success of dioecious species in a fragmented cloud forest in India

The protection of pollen and ovules in palms

Wind and insect pollination (ambophily) of Mallotus spp. (Euphorbiaceae) in tropical and temperate forests

Pollination in the New Guinea endemic Antiaropsis decipiens (Moraceae) is mediated by a new species of Thrip, Thrips antiaropsidis sp.nov. (Thysanoptera: Thripidae)

Staminate peduncle size: short (<50 cm; 1), long (>50 cm

many (>5; 0). E. Staminate rachis: white and soft (1), green and hard (0). F. Pistillate peduncle size: short (<50 cm; 1), long (>50 cm; 0). G. Pistillate rachillae position: erect (1), pendulous (0). H. Pistillate branching: spicate (1), branched (0). I. Number of pistillate rachillae: few (<5; 1), many (>5; 0). J. Pistillate rachis: white and soft (1), green and hard (0). Flower morphology K. Staminate colour: showy (yellow or orange; 1), unshowy (green; 0). L. Pistillate colour: showy (yellow or orange; 1), unshowy (green; 0)

Staminate: present (1), absent (0). Q. Pistillate: present (1), absent (0)

Type: moist and sticky (1), dry and powdery (0)

The authors would like to thank: staff of the University for Peace and Gu ıas y Scouts de Costa Rica for allowing us to conduct the field experiments at El Rodeo and La Carpintera,