Oecologia DOI 10.1007/s00442-011-2043-8

PLANT- INTERACTIONS - ORIGINAL PAPER

Plant–pollinator interactions and floral convergence in two of from the Caribbean Islands

Silvana Marte´n-Rodrı´guez • W. John Kress • Ethan J. Temeles • Elvia Mele´ndez-Ackerman

Received: 4 April 2010 / Accepted: 26 May 2011 Ó Springer-Verlag 2011

Abstract Variation in interspecific interactions across nectar chambers and long corollas, whereas on Hispaniola, geographic space is a potential driver of diversification and H. bihai flowers resembled those of H. caribaea with local adaptation. This study quantitatively examined vari- longer nectar chambers and shorter corolla tubes. Mor- ation in floral phenotypes and pollinator service of Heli- phological variation in floral traits corresponded with conia bihai and H. caribaea across three Antillean islands. geographic differences or similarities in the major pollin- The prediction was that floral characters would correspond ators on each island. The Hispaniolan mango, Anthraco- to the major pollinators of these species on each island. thorax dominicus, is the principal pollinator of both Analysis of floral phenotypes revealed convergence among H. bihai and H. caribaea on Hispaniola; thus, the similarity species and populations of Heliconia from the Greater of floral phenotypes between Heliconia species suggests Antilles. All populations of H. caribaea were similar, parallel selective regimes imposed by the principal polli- characterized by long nectar chambers and short corolla nator. Likewise, divergence between H. bihai populations tubes. In contrast, H. bihai populations were strongly from Dominica and Hispaniola corresponded with differ- divergent: on Dominica, H. bihai had flowers with short ences in the pollinators visiting this species on the two islands. The study highlights the putative importance of pollinator-mediated selection as driving floral convergence Communicated by Steven Johnson. and the evolution of locally-adapted plant variants across a geographic mosaic of pollinator species. Electronic supplementary material The online version of this article (doi:10.1007/s00442-011-2043-8) contains supplementary material, which is available to authorized users. Keywords Convergent evolution Á Heliconia Á Á Islands Á Pollination S. Marte´n-Rodrı´guez Á W. John Kress Department of Botany, National Museum of Natural History, MRC-166, Smithsonian Institution, Washington DC 20013-7012, USA Introduction

Present Address: Variation in selective regimes imposed by pollinators S. Marte´n-Rodrı´guez (&) Departamento de Biologı´a Evolutiva, Instituto de Ecologı´a, across plant populations is thought to be a key element AC, Ap. postal 63, 91070 Xalapa, VER, Me´xico driving floral diversification (Johnson 1997, 2006; Boyd e-mail: [email protected]; [email protected] 2002; Herrera et al. 2006; Nattero and Cocucci 2007). As with local adaptation to the abiotic environment (e.g., E. J. Temeles Department of Biology, Amherst College, Amherst, different soil ecotypes; Wright et al. 2006), selection on Massachusetts 01002-5000, USA floral traits that enhance reproductive performance under particular pollination environments can drive the evolution E. Mele´ndez-Ackerman of locally-adapted floral variants or pollination ecotypes Institute for Tropical Ecosystem Studies, University of at Rio Piedras, P.O. Box 70377, San Juan, (Johnson 2006; Harder and Johnson 2009). If populations PR 00936-8377, USA are geographically isolated, limited gene flow between 123 Oecologia ecotypes could ultimately lead to species divergence pollinator assemblages on three islands that have different associated with shifts in pollination systems (Grant and pollinator communities: Dominica (), His- Grant 1965; Grant 1992; Johnson 2006). Local adaptation paniola and Puerto Rico (Greater Antilles). of species using the same pollinators could also drive Extending the study of Heliconia floral variation to the convergence of floral traits in sympatric species, although Greater Antilles, where the purple-throated carib is absent, this phenomenon has seldom been documented (but see allows us to start assessing the hypothesis that floral traits Anderson and Johnson 2009). Therefore, placing the study in Caribbean Heliconia have diversified under selective of plant–pollinator interactions within a geographic context regimes imposed by local pollinator faunas. Accordingly, is important to gain insights into the conditions that have we expect to find an association between Heliconia floral promoted floral diversification and convergence. trait variation and variation in bill characteristics of the Oceanic archipelagos offer ideal settings to assess the major pollinators of Heliconia on each island. This cor- role of geographic variation and local adaptation on plant relative approach is the first step to investigate whether and animal diversification. For instance, adaptive radia- local adaptation to pollinator assemblages is potentially tions in archipelagoes reflect repeated opportunities for driving floral diversification of on the Caribbean speciation in response to environmental selective mosaics, Islands. To evaluate our predictions we: (1) quantified resulting in wide phenotypic diversification and conver- patterns of pollinator visitation for H. bihai and H. cari- gence (e.g., Grant 1986; Losos 1992; Givnish et al. 2009). baea on Hispaniola and Puerto Rico, (2) characterized However, knowledge of spatial variation in plant–pollina- variation of floral traits important for pollination in Heli- tor interactions across islands of oceanic archipelagoes is conia populations from Hispaniola, Puerto Rico (Greater limited. This paucity of information is particularly true for Antilles) and Dominica (Lesser Antilles), and (3) examined the Caribbean Islands where strict characterization of the association between pollinator assemblages and floral plant–pollinator interactions over a range of island popu- variation across the three islands. lations has only been attempted for the genus Heliconia (e.g., Temeles and Kress 2003; Gowda 2009). Accordingly, we examined floral and pollination system variation in Materials and methods Heliconia populations from three Antillean islands. Caribbean Heliconia offer an ideal study system for vari- Study system ous reasons. First, only two species of Heliconia are native to the Antilles, but these species display great variation in The floral traits across islands (Berry and Kress 1991). Second, the Caribbean Islands are moderately isolated from each Heliconia bihai (L.) Griggs and H. caribaea Lamarck are other; therefore, limited gene flow among islands might two closely related species that comprise the only native favor the evolution of stable, locally-adapted variants. representatives of the genus Heliconia in the Antilles. Third, comprehensive studies of plant–pollinator interac- Heliconia caribaea is distributed across the Antilles, from tions, floral ecology, and hummingbird behavior are Eastern Cuba to Saint Vincent, with populations present on available for various islands of the Lesser Antilles, pro- all mountainous islands (Anderson 1981). Heliconia bihai viding baseline information for comparative studies is distributed from northern South America through the (Temeles et al. 2000, 2005, 2009; Temeles and Kress 2003; Lesser Antilles, and is also present on Hispaniola (Greater Gowda 2009. Antilles). Heliconia bihai has also been reported from Earlier studies on the islands of St Lucia and Dominica Puerto Rico (Acevedo and Strong 2010); however, it is showed a strong association between Heliconia floral known only from one collection (Acevedo, personal com- phenotypes and the bills and energy requirements of their munication), and we were not able to find it in our survey sexually-dimorphic hummingbird pollinator, the purple- across the island. For this reason, we report data for both throated carib (Eulampis jugularis) (Temeles et al. 2000; Heliconia species from Hispaniola and Dominica, but only Temeles and Kress 2003). Heliconia bihai has long, curved for H. caribaea from Puerto Rico. flowers that are pollinated by female purple-throats, Heliconia plants are large perennial herbs mostly whereas H. caribaea has shorter and straighter flowers that occurring in disturbed habitats, along roads, trails, rivers, are pollinated primarily by male purple-throats (Temeles and in forest gaps. They have rhizomatous growth, a mu- and Kress 2003; Fig. 1). The geographic range of the soid growth habit, and produce multiple inflorescences purple-throated carib is restricted to the Lesser Antilles each of which can last from 1 to 3 months. Flowering (Raffaele et al. 1998), but the native ranges of H. bihai and seasons for both species of Heliconia on Hispaniola and for H. caribaea extend to the Greater Antilles (Berry and Kress H. caribaea on Puerto Rico ranged from February through 1991). Here, we assess floral variation in relation to July, with peak flowering in April–May. Inflorescences are 123 Oecologia

Table 1 Floral visitors of Heliconia caribaea and H. bihai in Hispaniola and Puerto Rico Species Sex Bill length Body Island Common name Behavior Category (mm) mass (g)

Anthracothorax F 24.5 ± 0.3 5.4 ± 0.5 Hispaniola Hispaniolan Pollinator LB dominicus M 22.6 ± 0.2 6.2 ± 0.5 mango F 25.1 ± 0.2 6.2 ± 0.3 Puerto Rico Green mango Pollinator LB viridis M 23.1 ± 0.2 6.6 ± 0.3 Eulampis F 23.3 ± 0.3 – Puerto Rico Green-throated Pollinator LB holocericeus M 20.1 ± 0.3 5.6 ± 0.4 carib F 17.6 ± 0.4 3.7 Hispaniola Hispaniolan Nectar robber/occasional SB swainsonii M 16.6 ± 0.3 3.2 emerald pollinator Chlorostilbon F 13.8 ± 0.2 2.9 ± 0.2 Puerto Rico Puerto Rican Nectar robber/occasional SB maugeus M 12.5 ± 0.1 2.9 ± 0.2 emerald pollinator Coereba flaveola M and 10.5 9.7 ± 0.1 Hispaniola, Puerto Bananaquit Nectar robber/pollinator SB F Rico Eulampis jugularis F 26.6 ± 0.12 7.9 ± 0.09 Dominica Purple-throated Pollinator LB M 19.8 ± 0.36 9.9 ± 0.1 carib All species listed for Hispaniola visited both Heliconia species. Categories refer to pollinator functional groups based on bill characteristics: long-billed (LB), short-billed (SB). Measurements of bill length and body mass are drawn from Kodric-Brown et al. (1984), Wunderle (1995), Fumero-Caba´n and Melendez-Ackerman (2007), and Temeles, unpublished data composed of large showy bracts that, on Puerto Rico and H. bihai, which has long, curved flowers (Temeles and Hispaniola, are bright yellow in H. caribaea, and red- Kress 2003, 2010; Gowda 2009). Although three additional orange with yellow edges in H. bihai. Each bract holds an species of are present on Dominica, they average of 10–24 flowers over the season, but no more than visit the native heliconias at much lower frequencies than one flower is produced daily within a bract; anthesis lasts the purple-throated carib. Three hummingbird species are 1 day and there is no temporal separation of sexual phases native to Hispaniola (Raffaele et al. 1998), two of which, (Temeles et al. 2005; Gowda 2009). The flowers are the long-billed Hispaniolan mango (Anthracothorax dom- bisexual, zygomorphic, tubular, and greenish-white in inicus) and the short-billed (Chloro- color. Both Heliconia species have one ovule per carpel, stilbon swainsonii), we observed at heliconias. Six and they usually produce from one to three seeds per fruit. hummingbird species are native to Puerto Rico (Raffaele Breeding systems of Caribbean Heliconia vary across et al. 1998), three of which, the long-billed green mango islands; populations of both species on Dominica are par- (A. viridis) and green-throated carib (E. holosericeus) and tially self-incompatible (Gowda 2009), whereas hand-pol- the short-billed Puerto Rican emerald (C. maugeus), we lination crosses conducted in Puerto Rico revealed H. recorded at heliconias. On both islands, we also observed caribaea on Puerto Rico is fully self-compatible (Marte´n- bananaquits (Coereba flaveola) at heliconias. None of the Rodrı´guez, unpublished data). We lack data for breeding hummingbirds on Hispaniola or Puerto Rico exhibit the systems of Heliconia from Hispaniola. extreme in bill length or shape of the purple-throated carib (Schuchmann 1999; Temeles, The unpublished data). For analyses, we classified floral visitors into pollinator functional groups, sensu Fenster et al. Previous work has demonstrated that the purple-throated (2004), according to their size, bill length and feeding carib is the primary pollinator of both H. bihai and H. behavior (whether they were legitimate visitors or nectar caribaea on Dominica, accounting for 90% of the visits to robbers; see Results: ‘‘Pollinator visitation’’). Thus, we H. caribaea and 99.9% of visits to H. bihai in our 9 years grouped bird visitors into three categories: long-billed of studies on that island (Temeles and Kress 2003, 2010; hummingbirds, short-billed hummingbirds, and banana- Gowda 2009). This hummingbird species exhibits extreme quits (see Table 1 for bill dimensions). sexual dimorphism in bill length and curvature associated with sexual partitioning of these heliconias: males with Study sites short, straight bills visit H. caribaea, which has short, straight flowers, whereas females with long, curved bills On Puerto Rico, H. caribaea was studied in El Yunque occasionally visit H. caribaea but are the primary visitor of National Forest (various sites along roads 191 and 966, and

123 Oecologia

Fig. 1 Floral phenotypes and major hummingbird visitors of Heliconia caribaea and H. bihai studied on Hispaniola, Dominica, and Puerto Rico. Note the shorter nectar chambers and longer corollas that characterize flowers of H. bihai from Dominica. Scale bars 20 mm. Hummingbird drawings originally published in Raffaele et al. (1998), and authorized to be used in this publication by Herbert Raffaele

at El Verde Field Station), in Bosque Estatal Carite, and flowering season. We recorded floral visitors through direct along road 105 to Maricao; elevation ranged between 350 observation and with the help of video cameras (Sony and 700 m a.s.l. With the exception of the Maricao popu- Handycam DCR-SR85), and from both types of observa- lation, all sites were within the boundaries of protected tion noted the species and sex (when dimorphic only) of the forest. On Hispaniola, H. caribaea and H. bihai were visitor, the number of inflorescences visited, and the studied in various sites across Cordillera Septentrional, number of flowers probed. Honeybees (Apis mellifera) Cordillera Central and Sierra de Bahoruco, all located in were observed foraging in two patches of H. caribaea on the Dominican Republic side of the island, between 200 Hispaniola; honeybees collected pollen from anthers and and 600 m a.s.l. These study sites consisted of patches of spent long periods of time in the flower moving mostly secondary growth vegetation, nested within a matrix of between flowers of the same inflorescence or plant. Over- agricultural land. Specific locations and coordinates are all, they were rare visitors to Heliconia flowers, although listed in Online Resource 1. they were common at one site. Honeybees are not native to the Antilles and it is unlikely they have had time to influ- Pollinator visitation ence the floral evolution of long-lived Heliconia since their introduction to these islands less than 400 years ago (Cox We conducted pollinator observations during March, April, 1994). Because the goal of this study was to evaluate the and June 2009 in Puerto Rico, and March and May 2009 in role of geographic variation in pollinator communities on Hispaniola. Pollinator visitation data for H. bihai and the evolution of floral variation in Heliconia, we did not H. caribaea from Dominica were available from past studies quantify visits by non-native species. (Temeles and Kress 2003; Gowda 2009). We observed the We also recorded the total number of plants, inflores- plants during clear, cloudy or light rainy weather; no cences, and open flowers on each patch. on tall observations were conducted during heavy rain. In Puerto stems were counted using a mirror attached to a pole. Rico, we observed 13 patches of H. caribaea for 12 h each, Pollinator visitation frequencies were calculated as the in sets of 2-h periods; we covered all periods between 0630 number of visits per flower per patch per hour. The number and 1830 hours over the course of several days. Patches of visits was then multiplied by 12 (approximate number of consisted of one to five distinct clumps of stems. In His- daylight hours at these latitudes) to obtain a metric that paniola, we observed six patches of H. caribaea and six of reflected the number of visits per flower per patch per day. H. bihai using the same methodology as in Puerto Rico. On Since flowers last only 1 day, a minimum of one visit per both islands, observations were conducted during the peak day is necessary for pollen deposition and removal.

123 Oecologia

Floral trait variation corolla height (an indirect measure of curvature; i.e., greater heights indicate greater curvatures). We measured six floral characters (nectar chamber length, nectar chamber width, corolla length, corolla width, corolla arc length, and corolla curvature height, in mm; Online Results Resource 2) on fresh flowers collected from two to nine individuals in two to four populations of each Heliconia Pollinator visitation species on each island (Hispaniola, Puerto Rico and Dominica). For each individual plant, we measured one to Visitation rates differed significantly among pollinator four flowers and averaged these values to obtain a single functional groups and between islands (Table 2). On His- value per trait per individual. The traits we measured were paniola, H. bihai and H. caribaea were pollinated primarily expected to be under selection by pollinators based on our by the long-billed Hispaniolan mango (males and females), previous studies of floral trait variation (Temeles et al. accounting for 90% of the visits; other visitors included 2000; Temeles and Kress 2003; Gowda 2009). Curvature bananaquits and the short-billed Hispaniolan emerald height provides an indirect measure of curvature; the (Fig. 2). On Puerto Rico, H. caribaea plants received visits greater values of height indicate greater curvature. by two species of long-billed hummingbirds (the green mango and the green-throated carib, 41% of all visits), Statistical analyses bananaquits (40% of all visits), and the short-billed Puerto Rican emerald (19% of all visits; Fig. 2). Short-billed Differences in pollinator visitation were tested with hummingbirds were primarily nectar thieves, failing to ANOVA using PROC GLIMMIX in SAS v.9.2 (SAS contact the reproductive organs of the flower (Marte´n- Institute 2008). The model included pollinator functional Rodrı´guez, unpublished data). Bananquits acted both as group (long-billed hummingbird, short-billed humming- nectar robbers, by piercing a hole at the level of the nectar bird, and bananaquit), island (Puerto Rico or Hispaniola), chamber, and as pollinators, by directly contacting repro- Heliconia species (H. bihai and H. caribaea), and all ductive organs and picking-up pollen on different parts of possible interactions as predictor variables. Neither the head and chest while taking nectar accumulated in the island 9 species interaction, nor the three-way interaction corolla pouch. While bananaquits are known to effectively were possible because native populations of H. bihai are transfer pollen among Heliconia flowers (Mele´ndez-Ack- absent from Puerto Rico. Covariation in visitation of the erman, unpublished data), the large variation in feeding same patches by each functional group of pollinators was behaviors makes bananaquits unlikely agents of directional accounted for in a random residual statement (patch was or stabilizing selection on floral traits. Visitation occurred specified as subject and option unordered was used for the throughout the day in both Heliconia species. The only covariance matrix). A Poisson distribution was specified in temporal pattern observed was that long-billed humming- the model statement with a link = log option. Back- birds were primary visitors during early morning hours transformation of means was obtained by the ilink option (0600–0800 hours), whereas bananaquits and short-billed under the lsmeans statement. A priori contrasts were used hummingbirds became frequent visitors after 0900 hours, to compare visitation rates among Heliconia populations coinciding with the time when nectar accumulation reached and among pollinators. the top of the nectar chamber in unvisited flowers. To assess variation of flowers as integrated phenotypes across islands, we performed a canonical discriminant Floral trait variation analysis using the CANDISC procedure in SAS v.9.2 (SAS Institute 2008). This analysis generates linear combinations On both Hispaniola and Puerto Rico, corollas of H. cari- of variables (e.g., floral traits) that have the highest mul- baea were relative short with long nectar chambers, com- tiple correlations with the classes (e.g., populations of parable to the flowers of that species on Dominica Heliconia from different islands). It also estimates the (Table 3). In contrast, floral phenotypes of H. bihai dif- contribution of individual traits to the generated canonical fered among islands. Corollas of H. bihai on Dominica variables as standardized canonical coefficients. Two were approximately 10 mm longer than on Hispaniola, canonical axes were selected for plotting floral phenotypes whereas nectar chambers of H. bihai on Dominica were because the third axis of variation did not contribute to any approximately 10 mm shorter than on Hispaniola distinct separation among clusters. The traits that contrib- (Table 3). Significant canonical correlations were found uted most to variation in CAN1 were nectar chamber among floral traits underlying the clustering of floral phe- length and corolla length, whereas the traits that contrib- notypes (Wilks k, F(24, 405) = 48.8, P \ 0.0001). Floral uted most to variation in CAN2 were corolla width and phenotypes of Heliconia segregated in two clusters: one 123 Oecologia

Table 2 ANOVA statistics of pollinator visitation and island for the phenotypic similarities and differences observed among H. bihai and H. caribaea from Hispaniola and Puerto Rico in 2009 flowers of Caribbean Heliconia.

Effect F(num df, den df) P value Variation in pollinator assemblages across islands was associated with patterns of floral variation, which was Pollinator 16.40(2, 75) \0.0001 particularly evident in the comparison between Dominica Island 1.39(1, 75) 0.2511 and Hispaniola, islands where both Heliconia species are Species 0.69(1, 75) 0.4016 present. On Dominica, the presence of the purple-throated Pollinator 9 island 9.59(2, 75) 0.0009 carib hummingbird, with its marked bill dimorphism, has Pollinator 9 species 0.14(2, 75) 0.8719 provided opportunities for Heliconia to evolve in response to selection by either male or female hummingbird poll- inators, explaining the large floral differences among Heliconia species on this island (Temeles and Kress 2003, and unpublished data). In contrast, both H. bihai and H. caribaea on Hispaniola are subject to selection by a species of hummingbird that lacks the marked sexual dimorphism in bill size and shape of the purple-throated carib (Table 2). The striking similarities among flowers of H. bihai and H. caribaea on Hispaniola may thus be the result of floral convergence driven by selection mediated by a common pollinator species. Floral convergence of closely related sympatric species due to sharing of the same pollinator is a rarely docu- mented phenomenon; however, there are a few well-stud- ied cases (Schemske 1981; Anderson and Johnson 2009). Fig. 2 Visitation rates to H. bihai and H. caribaea from Hispaniola In one example, the floral tube length of fly-pollinated and Puerto Rico, March–June 2009, by three pollinator groups. Means plants in South Africa reflected convergence between with identical letters are not significantly different from each other sympatric species and divergence between allopatric (experiment-wide a = 0.05, Tukey adjusted) populations of the same species, a pattern of floral variation that was associated with variation in the proboscis length of that contained H. bihai from Dominica, and the other that a single fly pollinator species (Anderson and Johnson included all other populations of Heliconia (Fig. 3). Clus- 2009). The case of Caribbean Heliconias provides a similar tering of floral phenotypes along canonical axis 1 (CAN1) example, with the difference that not a single pollinator was associated with variation in nectar chamber length, visited all Heliconia populations in all islands. corolla length, and corolla width (Table 3). Along canon- In a second well-documented example of floral con- ical axis 2 (CAN2), populations of Heliconia were more vergence, similarities in nectar production, phenology, and dispersed; thus, floral trait variation associated with this floral morphology of two co-occurring Costus species from axis did not contribute to cluster separation (Fig. 3). Barro Colorado Island were attributed to selection to enhance visitation by a single bee pollinator species. This occurred in an environment where low flower density Discussion caused by weevil predation resulted in insufficient polli- nator visitation (Schemske 1981). In contrast with the In the Neotropics, the genus Heliconia has flowers adapted Costus system, facilitation is unlikely to play a role in to bird pollination; therefore, variation in floral traits driving floral convergence of Heliconias on Hispaniola among species is not associated with major transitions in because Heliconia patches are usually monospecific and pollination syndromes (e.g., birds vs. ). Instead, floral the two species rarely share the same habitats. Heliconia traits reflect variation in the bill characteristics of the bihai is found almost exclusively on steep rocky walls of particular hummingbird pollinators of each Heliconia ravines, whereas H. caribaea occurs on varied terrain along species (Stiles 1975; Temeles and Kress 2003; Temeles stream edges, wet forest edges, and gaps. Thus, although et al. 2010). Geographic differences in hummingbird the two Heliconia species share the same pollinator, dif- assemblages thus have the potential to influence the evo- ferent habitat preferences make co-occurrences rare within lution of locally-adapted floral phenotypes in Heliconia. the island. The results of this study support the idea that geographic The use of different local habitats by H. caribaea and variation in pollinator composition and service underlies H. bihai on Hispaniola is perhaps a reason why some forms 123 Oecologia

Table 3 Trait measurements (mean ± SEM), one-way ANOVA statistics, and standardized coefficients from Canonical Discriminant Analysis of floral variation in H. caribaea and H. bihai from Hispaniola, Puerto Rico and Dominica Floral trait Greater Antilles Lesser Antilles ANOVA Standardized canonical coefficientsa

H. caribaea H. caribaea H. bihai H. caribaea H. bihai F(4, 120) P CAN1 CAN2 (n = 32) (n = 14) (n = 18) (n = 31) (n = 31) Puerto Rico Hispaniola Hispaniola Dominica Dominica

Nectar chamber 20.8 ± 0.24 a 21.9 ± 0.41 a 21.4 ± 0.52 a 17.8 ± 0.33 c 10.6 ± 0.21 b 227.2 \0.0001 2.44 1.02 length (mm) Nectar chamber 5.6 ± 0.07 a 5.1 ± 0.11 b 5.7 ± 0.13 a 5.0 ± 0.07 b 6.8 ± 0.12 c 53.1 \0.0001 -0.11 0.37 width (mm) Corolla length 37.3 ± 0.35 a 35.7 ± 0.23 a,b 37.5 ± 0.51 a 35.8 ± 0.22 b 47.8 ± 0.43 c 206.1 \0.0001 -1.44 -0.47 (mm) Corolla width 5.9 ± 0.07 a 5.4 ± 0.10 a 6.7 ± 0.14 b 5.6 ± 0.09 a 9.0 ± 0.13 c 202.1 \0.0001 -1.07 1.62 (mm) Corolla arc 42.8 ± 0.39 a 42.3 ± 0.25 a,b 43.1 ± 0.63 a 40.5 ± 0.22 b 54.2 ± 0.51 c 186.9 \0.0001 -0.40 0.92 length (mm) Corolla height 8.7 ± 0.12 a 9.3 ± 0.18 a 7.8 ± 0.15 b 9.1 ± 0.11 a 11. ± 0.17 C 69.7 \0.0001 -0.17 -1.59 (mm) Sample sizes (n) refer to the number of individual Heliconia plants from which flowers were measured. Means followed by identical letters are not significantly different from each other (experiment-wide a = 0.05, Tukey adjusted) a Standardized canonical coefficients represent the contribution of each variable to the canonical axes while holding other variables in the model constant

These two forms of floral isolation may also act in concert (Grant 1992), and in conjunction with other forms of reproductive isolation (Kay and Sargent 2009). In Carib- bean Heliconia, floral isolation is observed on the island of Dominica, where both structural differences in flower morphology and pollinator preferences determine effective pollination by either male or female purple-throated caribs (Temeles et al. 2009). However, this situation does not occur on Hispaniola, where there are no morphological differences between Heliconia species that would promote pollen placement on different parts of the pollinator’s body. Fig. 3 Canonical discrimant analysis of floral traits for two species of These observations suggest that strong selection to enhance Heliconia with populations on three Caribbean islands. On axis Can1, pollen transfer by the single pollinator species, coupled populations on the left have flowers with long corollas and short nectar chambers, and populations on the right have flowers with short with reproductive isolation promoted by different habitat corollas and long nectar chambers preferences, may be driving floral convergence and pre- venting character displacement. Other explanations are also possible to explain the of reproductive isolation involving floral divergence observed patterns of floral similarity among species of between species are apparently lacking. In theory, when Heliconia on Hispaniola. For instance, correspondence close relatives occur in sympatry and share the same of morphological traits could reflect common expression of pollinators, floral traits should diverge to promote repro- plastic phenotypes in a shared environment (Pigliucci ductive isolation in order to reduce reproductive losses 2001). Phenotypic plasticity is not a very plausible incurred by hybrid crosses (reinforcement) or to reduce hypothesis, however, because the individuals measured in competition for pollinators (character displacement; Grant this study were sampled from populations at different 1972; Johnson 2006). Floral isolation may occur via elevations and different regions of Hispaniola, encom- changes in flower structure that direct pollen placement passing wide environmental variations in temperature, onto particular parts of the pollinator’s body (mechanical precipitation, light and soils. Therefore, the common isolation; Grant 1949), or via pollinator’s preferences for expression of floral traits cannot be explained by a common particular plant species (ethological isolation; Grant 1949). response to a set of particular abiotic conditions. 123 Oecologia

A second explanation for the shared floral similarities of Anderson B, Johnson SD (2009) Geographical covariation and local H. bihai and H. caribaea on Hispaniola is that convergent convergence of flower depth in a guild of fly-pollinated plants. New Phytol 182:533–540 floral phenotypes evolved in allopatric ancestral popula- Berry F, Kress WJ (1991) Heliconia: an identification guide. tions in other islands and were later assembled in sympatry Smithsonian Institution Press, Washington, DC by chance dispersal. Because flowers most closely resem- Boyd A (2002) Morphological analysis of sky island populations of bling H. bihai from Hispaniola correspond to populations Macromeria viridiflora (Boraginaceae). Syst Bot 27:116–126 Cox B (1994) AHB in Puerto Rico. Am Bee J 134:668–669 of this species from islands near the coast of South Fenster CB, Armbruster WS, Thomson JD, Wilson P, Dudash MR America (; Kress and Temeles, (2004) Pollination syndromes and floral specialization. Annu unpublished data), long-distance dispersal would have to Rev Ecol Evol Syst 35:375–403 be invoked to support this hypothesis. This explanation, Fumero-Caba´n JJ, Melendez-Ackerman EJ (2007) Relative pollina- tion effectiveness of floral visitors of Pitcairnia angustifolia albeit unlikely, cannot be rejected until the phylogeography (Bromeliaceae). Am J Bot 94:419–424 of Caribbean Heliconia is better understood. Nevertheless, Givnish TJ, Millam KC, Mast AR, Paterson TB, Theim TJ, Hipp AL, even if heliconias from Hispaniola did not evolve their Henss JM, Smith JF, Wood KR, Sytsma KJ (2009) Origin, actual floral phenotypes in situ, pollinator-mediated selec- adaptive radiation and diversification of the Hawaiian lobeliads (Asterales: Campanulaceae). Proc Biol Sci 276:407–416 tion has probably played an important role in preventing Gowda V (2009) Pollination biology and inter-island geographical the divergence of floral traits that would be expected under variation in the mutualistic Heliconia (Heliconiaceae)-humming- drift or reinforcement. bird (Trochilidae) interaction of the eastern Caribbean Islands. In summary, both phenotypic divergence and conver- PhD dissertation, George Washington University, Washington, DC gence appear to have played an important role in the Grant PR (1972) Convergent and divergent character displacement. diversification of floral traits in Antillean Heliconia. Pre- Biol J Linn Soc 4:39–68 vious studies have documented large floral differences Grant PR (1986) Ecology and evolution of Darwin’s finches. between H. bihai and H. caribaea on Dominica (Temeles Princeton University Press, Princeton Grant V (1949) Pollination systems as isolating mechanisms in and Kress 2003). Our study reports a case of divergence angiosperms. Evolution 3:82–97 between allopatric populations of the same species, H. bi- Grant V (1992) Floral isolation between ornithophilous and sphing- hai on Hispaniola and Dominica, and convergence between ophilous species of Ipomopsis and Aquilegia. Proc Natl Acad Sci two species of Heliconia, H. bihai and H. caribaea,on USA 89:11828–11831 Grant V, Grant KA (1965) pollination in the Phlox family. Hispaniola. The associations between floral characteristics Columbia University Press, New York and bills of the principal hummingbird pollinators on each Harder LD, Johnson SD (2009) Darwin’s beautiful contrivances: island suggest that adaptation to local pollinator faunas is a evolutionary and functional evidence for floral adaptation. New major evolutionary force driving floral diversification of Phytol 183:530–545 Herrera C, Castellanos MC, Medrano M (2006) Geographical context Caribbean Heliconia. of floral evolution: towards an improved research programme in floral diversification. In: Harder L, Barrett SCH (eds) Ecology Acknowledgments The authors thank T. Clase, A. Cherenfant, and evolution of flowers. Oxford University Press, Oxford, J. Fumero, N. Marte´n, B. Peguero, R. Perez, R. Rodrı´guez, N. Ruiz pp 278–294 and M. Vindas for assistance conducting field work. The authors Johnson SD (1997) Pollination ecotypes of Satyrium hallackii thank P. Acevedo for discussion of ideas and providing useful (Orchidaceae) in South Africa. Bot J Linn Soc 123:225–235 botanical information, and I. Lopez for generating the distribution Johnson SD (2006) Pollinator-driven speciation in plants. In: Harder map. V. Gowda, S. Johnson, and three anonymous reviewers provided L, Barrett SCH (eds) Ecology and evolution of flowers. Oxford insightful comments on earlier versions of this manuscript. Logistic University Press, Oxford, pp 278–294 support for fieldwork was provided by Jardı´n Bota´nico Nacional de Kay KM, Sargent RD (2009) The role of animal pollination in plant Santo Domingo in Dominican Republic, and El Verde Field Station in speciation: integrating ecology, geography, and genetics. Annu Puerto Rico. Research permits were awarded by Depto de Recursos Rev Ecol Evol Syst 40:637–656 Naturales y Ambientales in Puerto Rico, and by the Secretarı´ade Kodric-Brown A, Brown JH, Byers GS, Gori DF (1984) Organization Estado de Medio Ambiente y Recursos Naturales in the Dominican of a tropical island community of hummingbirds and flowers. Republic, and experiments complied with the current laws of those Ecology 65:1358–1368 countries. Funding was provided by Smithsonian Institution grant to Losos JB (1992) The evolution of convergent structure in Caribbean SMR, NSF Grant DEB 0614218 to E. Temeles and W.J. Kress. Anolis communities. Syst Biol 41:403–420 Nattero J, Cocucci AA (2007) Geographical variation in floral traits of the tree tobacco in relation to its hummingbird pollinator fauna. Biol J Linn Soc 90:657–667 Pigliucci M (2001) Phenotypic plasticity: beyond nature and nurture. References Johns Hopkins University Press, London Raffaele H, Wiley J, Garrido O, Keith A, Raffaele J (1998) A guide to Acevedo P, Strong M (2010) Catalogue of the seed plants of the West the birds of the West Indies. Princeton University Press, Princeton Indies. Smtihsonian Institution. http://botany.si.edu/Antilles/ SAS Institute (2008) SAS for Windows, version 9.2. SAS Institute, WestIndies Cary Anderson L (1981) Revision of Heliconia sect. Heliconia (Musaceae). Schemske DW (1981) Floral convergence and pollinator sharing in Nordic J Bot 1:759–784 two bee-pollinated tropical herbs. Ecology 62:946–954

123 Oecologia

Schuchmann KL (1999) Family Trochilidae (hummingbirds). In: del Temeles EJ, Goldman RS, Kudla AU (2005) Foraging and territory Hoyo J, Elliott A, Sargatal J (eds) Handbook of the birds of the economics of sexually dimorphic purple-throated Caribs (Eu- world. Barn owls to hummingbirds, vol 5. Lynx, Barcelona, lampis jugularis) on three Heliconia morphs. Auk 122:187–204 pp 468–680 Temeles EJ, Koulouris CR, Sander SE, Kress WJ (2009) Effect of Stiles FG (1975) Ecology, flowering phenology, and hummingbird flower shape and size on foraging performance and trade-offs in pollination of some Costa Rican Heliconia species. Ecology a tropical hummingbird. Ecology 90:1147–1161 56:285–301 Temeles EJ, Miller JS, Rifkin JL (2010) Evolution of sexual Temeles EJ, Kress WJ (2003) Adaptation in a plant–hummingbird dimorphism in bill size and shape of hermit hummingbirds association. Science 300:630–633 (Phaethornithinae): a role for ecological causation. Philos Trans Temeles EJ, Kress WJ (2010) Mate choice and mate competition by a R Soc Lond B 365:1053–1063 tropical hummingbird at a floral resource. Proc Biol Sci Wright JW, Stanton ML, Sherson R (2006) Local adaptation to 277:1607–1613 serpentine and non-serpentine soils in Collinsia sparsiflora. Evol Temeles EJ, Pan IL, Brennan JL, Horwitt JN (2000) Evidence for Ecol Res 8:1–21 ecological causation of sexual dimorphism in a hummingbird. Wunderle JM (1995) Responses of bird populations in a Puerto-Rican Science 289:441–443 forest to hurricane Hugo—the first 18 months. Condor 97:879–896

123