The Bahama Woodstar (Calliphlox evelynae), a hummingbird endemic to the Bahama Archipelago, comprises two currently recognized subspecies: Calliphlox e. evelynae, found throughout the Bahamas and in the Turks and Caicos Islands, except on Great and Little Inagua; and C. e. lyrura, named for its unique, lyre-shaped outer tail feathers and found only on the islands of Great and Little Inagua. The two were originally described as separate species, partly on the basis of their divergent tail morphology, but were subsequently lumped by Peters (1945). These taxa are members of the North American “bee” hummingbird clade, which produce mechanical sounds with their tails during courtship displays. Changes in tail shape may produce significant acoustic divergence. To determine the extent of differentiation between lyrura and evelynae, we collected field recordings of calls, songs, and courtship displays from New Providence and Great Inagua islands and surveyed morphological variation across the archipelago. We sequenced 4 nuclear loci and 2 mitochondrial genes from 9 individuals of evelynae and 6 individuals of lyrura. Both sexes of lyrura and evelynae can be diagnosed by vocal calls, and males can be diagnosed by morphology, song, and courtship display. Phylogenetic reconstructions based on the genetic data indicate that the 2 populations are reciprocally monophyletic and that they diverged ∼0.69 mya. Our data indicate that lyrura is a unique evolutionary lineage that warrants species status under both the phylogenetic and the biological species concept.
Island archipelagos play an important role in avian speciation because isolation can restrict gene flow among different island populations. The West Indies, in particular, have served as a focal point for studies of diversification in the wood warblers (Lovette et al. 1998, Klein et al. 2004, Markland and Lovette 2005) and other avian taxa (Bellemain et al. 2008, Garrido et al. 2009). Within the West Indies, the Bahama Archipelago, which includes the islands of the Bahamas and the Turks and Caicos (Figure 1), has been the focus of evolutionary investigation. Studies examining variation within Yellow-throated Warbler (Setophaga dominica; McKay et al. 2010), Brown-headed Nuthatch (Sitta pusilla; Hayes et al. 2004), Cuban Parrot (Amazona leucocephala; Reynolds and Hayes 2009, Russello et al. 2010), and Greater Antillean Oriole (Icterus dominicensis; Price and Hayes 2009) have all found Bahamian island populations to be distinct from those on the mainland or on other islands in the West Indies.
Furthermore, the Bahamas harbor distinct subspecies among different islands within the archipelago. Examples include the endemic Bahama Woodstar (Calliphlox evelynae; Gill and Donsker 2013) and subspecies of broader-ranging Caribbean taxa such as Cuban Parrot, Thick-billed Vireo (Vireo crassirostris; Buden 1985), and Greater Antillean Bullfinch (Loxigilla violacea; Buden 1987). Bahamian populations all exhibit a similar pattern of geographic variation: One or more subspecies are restricted to a southern island, and an additional subspecies ranges across one or more northern islands.
This distributional pattern is likely due, in part, to the complex geological history of the Bahama Archipelago. The majority of the northern islands are situated on either the Great or Little Bahama Bank, whereas most of the southern islands are each situated on their own separate banks (Figure 1; Carew and Mylroie 1995). Over the past 2.5 Ma, fluctuations in sea level driven by Pleistocene glacial cycling have resulted in periods during which the Great and Little Bahama Banks were exposed, uniting the majority of northern islands into 2 large land masses (Carew and Mylroie 1995). By contrast, southern islands during these periods of sea-level minima remained isolated on their separate banks (Buden 1987, Carew and Mylroie 1995). This complex history of island size fluctuation may have reproductively isolated bird populations on different Bahamian islands.
The Bahama Woodstar (Calliphlox evelynae), a member of the North American branch of the “bee” hummingbird clade, is endemic to the islands of the Bahamas and the Turks and Caicos (Cory 1880, Ridgway 1911, Bond 1936, McGuire et al. 2007, 2014). There are currently 2 recognized subspecies: C. e. evelynae (Bourcier 1847) and C. e. lyrura (Gould 1869). Calliphlox e. lyrura is restricted to the southern islands of Great and Little Inagua, whereas evelynae is found throughout the remaining islands in the archipelago (Figure 1). Calliphlox e. lyrura was originally described as a distinct species from evelynae on the basis of its diagnostic morphology (Bourcier 1847, Gould 1869, Cory 1880, 1918, Ridgway 1911, Bond 1936). These 2 taxa were treated as distinct species in all accounts until 1945, when they were lumped in the Check-list of Birds of the World by Peters (1945). A third form, C. e. salita—described by Greenway (1936) as a subspecies of C. evelynae from the Turks and Caicos—is no longer recognized.
The 2 primary characters used to diagnose lyrura from evelynae were the presence of iridescent feathers on the forehead and the shape of the elongated, outwardly curving, outer tail feathers (Figure 1; Bourcier 1847, Gould 1869, Cory 1880). Taxonomists have placed heavy weight on both tail morphology and iridescent gorget colors in diagnosing discrete hummingbird lineages (Mulsant and Verraux 1866, Gould 1869, Stiles 1972, 1983).
Iridescent gorget feathering and tail-feather shape both appear to play a role in sexual selection and mate choice. Iridescent feathers are erected and oriented toward females during displays, whereas tail feathers produce species-specific sounds (Clark and Feo 2008, 2010, Feo and Clark 2010, Clark 2011, Clark et al. 2011, 2013c). During a display dive, airflow causes the feather to vibrate and produce tonal sound. Pitch is set in part by the feather's shape, but the effect of shape change on sound is not easy to predict on the basis of morphology alone. Small changes in feather shape can result in changes in a feather's mode of flutter that produce changes in sound (Clark and Feo 2010, Clark et al. 2011, 2012, 2013a, 2013b).
The diversity in tail morphology among bee hummingbirds is likely the result of sexual selection on acoustic elements of male courtship displays (Clark 2010). As with many other species of bee hummingbird, differences in tail-feather morphology and iridescent plumage between evelynae and lyrura may indicate reproductive isolation. We investigated the degree to which populations of evelynae and lyrura have diverged in morphology, courtship displays, vocalizations, and genetics. We use these data to reevaluate the species status of lyrura.
We observed evelynae on New Providence island, Bahamas, during December 4–10, 2009, and March 3–8, 2012 (Figure 1). We studied individuals in a variety of habitats across the island, including Caribbean pine (Pinus caribaea) woodland south of the airport (Figure 1: PW; 25.016°N, 77.451°W) and residential neighborhoods in Coral Harbor on the southwest coast (CH; 24.982°N, 77.461°W), Fort Winton on the northeast coast (FW; 25.043°N, 77.266°W), and West Bay Street on the northwest coast (WB; 25.055803°N, 77.50016°W). We observed and recorded lyrura on Great Inagua in and just east of Matthew Town (Figure 1: MT; 20.950°N, 73.675°W), during February 24–March 2 and October 29–November 2, 2012. Specimens from both populations were collected in 2012 and deposited at the Yale Peabody Museum (Appendix Table 4).
We measured the length of rectrix 1 (R1), rectrix 5 (R5), exposed culmen, and folded wing chord on individuals captured in the field and on study skins from the Yale Peabody Museum (YPM), Museum of Comparative Zoology (MCZ), Field Museum of Natural History (FMNH), and American Museum of Natural History (AMNH). We performed a discriminant function analysis (with prior probabilities computed from group size) and t-tests to test for significant morphological differences between adult male, immature male, and female evelynae and lyrura. Mass was measured from individuals that we captured in the field, and hovering wingbeat frequency was measured from high-speed video.
Courtship Displays and Vocalizations
We recorded video and audio of courtship displays that were incidental or solicited by placing a female in a cage on a male's territory (Clark et al. 2012). Audio recordings were made with a 24-bit recorder (Sound Devices 702). Recordings were made with shotgun microphone (Sennheiser MHK70) sampling at 48 kHz in 2009, or with a Sennheiser MKH 20 microphone in a Telinga Pro Universal Parabola sampled at 96 kHz in 2012. Sounds were analyzed in Raven version 1.4 (see Acknowledgments) and converted into spectrograms using a 512-sample Hann window function and 50% overlap for recordings sampled at 48 kHz, and a 1,024-sample Hann window for recordings sampled at 96 kHz. High-speed videos of courtship displays were obtained with a monochrome and color MIRO EX4 high-speed cameras (Vision Research, Wayne, New Jersey, USA) recording at 500 and 1,265 frames s−1 (800 × 600 pixel resolution).
Outer tail feathers (R5) from adult male evelynae and lyrura were tested in a wind tunnel to determine their capacity to produce sound (sensu Clark et al. 2013a). We mounted feathers in the airflow and recorded sound produced over a range of airspeeds from 10 to 20 ms−1, which bracket the likely speeds traveled during a dive. The feathers were filmed at 23,121 frames s−1 to reveal which part fluttered to produce sound.
DNA Extraction, PCR, and Sequencing
Total genomic DNA was isolated from 6 lyrura, 7 evelynae, and 1 each of Lucifer Hummingbird (Calothorax lucifer) and Slender Sheartail (Doricha enicura) using a QIAGEN DNeasy extraction kit (Appendix Table 4). We collected new DNA sequence data for 4 nuclear loci and 2 mitochondrial genes (Appendix Table 5): fibrinogen beta chain intron 7 (FGB I7), adenylate kinase 1 intron 5 (AK1 I5), ornithine decarboxylase 1 introns 6 and 7 with intervening exon (ODC1), Z-linked muscle, skeletal, receptor tyrosine kinase intron 3 (MUSK I3), mitochondrially encoded NADH dehydrogenase 2 (MT-ND2), and mitochondrially encoded NADH dehydrogenase 4 and flanking leucine tRNA (MT-ND4). These regions were amplified using standard polymerase chain reaction (PCR) protocols and sequenced at the DNA Analysis Facility on Science Hill at Yale University. All sequence data are deposited in NCBI GenBank (accession nos. KP136320–KP136422).
We calculated several measures of genetic diversity within populations of evelynae from New Providence and lyrura from Great Inagua using the program DnaSP (Librado and Rozas 2009), including haplotype diversity (Hd; Nei 1987), nucleotide diversity (π; Nei 1987), Watterson's estimate of population mutation rate per site (θw; Watterson 1975, Nei 1987), and average number of nucleotide differences among sequences (Tajima 1983). These measures were calculated for each locus and population. We also investigated genetic divergence between the 2 populations by calculating the number of fixed differences (Hey 1991) and average number of nucleotide substitutions per site (Nei 1987) between populations for all loci. For these and subsequent genetic analyses, we included genetic data from McGuire et al. (2014) for 2 individuals of evelynae from New Providence (Appendix Table 6). Finally, because estimation of several measures required knowing individual alleles, we reconstructed haplotypes with PHASE (Stephens et al. 2001, Stephens and Donnelly 2003), implemented in DnaSP using default settings.
Selection and recent demographic changes can influence estimates of genetic divergence and estimates of time since divergence. We calculated Fu's Fs (Fu 1997) and Tajima's D (Tajima 1989) statistics for all loci to test for departure from neutral evolution. To test whether the values we calculated for each locus differed significantly from neutral expectations, we generated a distribution of expected Fu's Fs and Tajima's D values using the “coalescent simulations” option in DnaSP. For each locus and population, 1,000 simulations were conducted, assuming a neutral infinite-sites model and large constant population size, to generate the expected neutral distribution of Fu's Fs or Tajima's D values. Actual Fu's Fs or Tajima's D values were deemed statistically significant if they fell outside the 95% interval of the simulated distribution.
We implemented 2 methods of Bayesian phylogenetic inference to explore evolutionary relationships among sampled individuals (sensu Berv and Prum 2014). To assess whether or not populations from New Providence and Great Inagua represent reciprocally monophyletic sister clades, we built individual gene trees and a single concatenated gene tree using MrBayes version 3.2.2 (Ronquist et al. 2012). Further, we generated a time-calibrated species tree using *BEAST version 1.8.0 (Drummond et al. 2012). For these analyses, we included closely related representatives from Calothorax, Archilochus, and Mellisuga (McGuire et al. 2014). Finally, we chose nucleotide substitution models using PartitionFinder version 1.1.1 and the Bayesian Information Criterion (Lanfear et al. 2012, 2014).
For analysis with MrBayes, we used the MCMCMC settings, priors, and convergence diagnostics described in Berv and Prum (2014). For *BEAST analyses, alignment partitions, molecular clock calibrations, and priors were taken from McGuire et al. (2014)—we performed 4 separate analyses of 2.2 × 108 generations, with trees sampled every 2.0 × 104 generations. The log files generated by *BEAST were examined in Tracer version 1.6 (Rambaut et al. 2014) to ensure that the Markov chain Monte Carlo (MCMC) chain had run long enough and that the effective sample sizes for all statistics were >200 (most were >1,000). After discarding the first 103 trees from each of 4 analyses as burn-in, we randomly sampled 2.5 × 103 trees from each, produced a combined tree log file using LogCombiner version 1.8.0, and generated a maximum clade-credibility tree using TreeAnnotator version 1.8.0 with default settings (see Acknowledgments). Results are presented as means ± SD.
Morphology and Plumage
We investigated the morphology and plumage of the Bahama Woodstar across most of the major islands within its range (Figure 1). This included several individuals from the Turks and Caicos, which were described as C. e. salita by Greenway (1936) on the basis of tail coloration and iridescent forecrown feathers. Currently, salita is synonymized with evelynae (Gill and Donsker 2013). We found no significant differences between salita and the nominate evelynae in tail-feather coloration, shape, or size. One male individual of salita (FMNH 42913) had a small patch of iridescent feathers on the anteriormost portion of the forecrown, but this was not observed in other males from this population. These data support that salita is invalid, and we included Turks and Caicos specimens with nominate evelynae populations for subsequent morphological comparisons to lyrura.
Adult males of both evelynae and lyrura have a completely iridescent gorget and a forked tail with narrow feathers (Figure 2A–2C). Females from both populations lack an iridescent gorget and have a rounded tail with broader feathers (Figure 2A, 2C). Immature males are similar in appearance to females before they begin to molt in adult male tail feathers and iridescent feathers. Nevertheless, they can still be differentiated from females by a relatively shorter R1 (Figure 2A and Table 1) and by the presence of more extensive black on the outer vanes of R3–R5. This gives the overall impression of a blurred black band in immature male tails versus a sharply defined black subterminal band in female tails (Figure 2A).
Morphology of adult male, immature male, and female Calliphlox evelynae lyrura and C. e. evelynae (means ± SD, with ranges below).
Discriminant function analysis discovered significant differences between evelynae and lyrura adult males (λ = 0.15, P < 0.01), immature males (λ = 0.27, P < 0.01), and females (λ = 0.83, P = 0.05). Adult male lyrura from the Inagua islands are diagnosed from adult male evelynae of other islands by the presence of a fully iridescent forecrown, lyre-shaped outer tail feathers, and a more strongly forked tail (Figures 1 and 2; Table 1). Adult male lyrura have a fully iridescent forecrown of the same color as the gorget, whereas most evelynae have no iridescent feathers on the forecrown (Figure 1). We found 3 male evelynae with a few iridescent feathers on the forecrown. Two males had 1 or 2 iridescent feathers at the base of the bill, and a third male from Caicos (FMNH 42913) had a single row of iridescent feathers along the upper margin of the bill. The tail of adult male lyrura is more strongly forked than that of evelynae as a result of graduated tail-feather lengths (Figures 1 and 2). Adult male lyrura have significantly shorter R1 and significantly longer R5 (P < 0.01), with no overlap in R5 length between populations (Table 1). Additionally, we found no evidence of a north–south clinal transition in either R1 or R5 length in evelynae (Figure 3).
Female evelynae and lyrura were similar to each other in overall appearance (Figure 2A, 2C; Table 1), and only folded wing chord was significantly different between the 2 populations (P < 0.01). However, folded-wing-chord range overlapped between the 2 populations (Table 1), and we found no discrete morphological characters that could reliably be used to diagnose female lyrura from evelynae. Immature male evelynae and lyrura were also similar in appearance. Nevertheless, immature male lyrura were diagnosable from immature male evelynae by their significantly shorter R1 (P < 0.01) and significantly longer R5 (P < 0.01).
Male and female evelynae on New Providence were most common in residential areas where gardens and hummingbird feeders provided sources of food. Females were also common in the pine woodland near the airport (Figure 1: PW), where males were less common. In December 2009, female evelynae were engaged in all stages of nesting, including gathering of nesting material (spiderwebs), incubating of eggs, and feeding of fledglings. In March 2012, we did not happen to find any active nests of evelynae.
In both December and March, male evelynae guarded courtship territories, sang, and displayed. Individual males spent the majority of their time within a given territory of approximately 25 × 25 m (core) and utilized 5–10 perches. The density of males was variable between locations and times. In December 2009, we found 10 or more males along a short stretch of road at Fort Winton (Figure 1: FW), but only 4 males along the same road in March 2012. Elsewhere, we found 1 lone male holding a territory in the pine woodland.
Male evelynae readily displayed to females that naturally entered their territory, and to caged females placed within their territories. During these displays, it was common for 1 or more additional males to perch close by as onlookers and, occasionally, interrupt the displaying male. Males infrequently displayed to other adult males that entered their territory before chasing them away. Once, we observed a male evelynae display and attempt to copulate with a fledgling.
Male and female lyrura were common on Great Inagua, and their behavior differed between the dry season (February) and the rainy season (October). During the dry season, lyrura were abundant in the flowering gardens of Matthew Town (Figure 1: MT), where they visited flowers of Cordia sebastina, Aloe vera, Carica papaya, Passiflora sp., Opuntia sp., Bougainvillea sp., and other unidentified flowers; we estimated that there were 400 birds km−2 in Matthew Town in February. In surveys, we observed a few additional birds in dune scrub near Northwest Point and in freshwater riparian habitat but found no hummingbirds in mangrove or coppice scrub, which had few flowers during the dry season.
We did not find signs of breeding lyrura during the dry season, and some birds were in body or tail molt. Both males and females guarded small territories (about 2 × 2 m) around dense nectar resources (e.g., Aloe vera), which they defended against other males and females. Both sexes defended territories, using scolding calls extensively in agonistic interactions with other individuals. We also observed 2 territorial males singing; these birds did not perform displays.
By contrast, we found very few hummingbirds in Matthew Town during the rainy season, despite the presence of blooming Cordia sp. Instead, hummingbirds were common in coppice just east of town, where several plants were in bloom, and we estimated ≥20 birds km−2. We found evidence of breeding lyrura during the rainy season: Two males held adjacent courtship territories in the low coppice just east of Matthew Town, and both performed displays and songs. Both territories revolved around 1 or 2 primary exposed perches, in the dead branches at the very top of a tall plant, as well as 2 or 3 lower perches, partially exposed and 5 to 10 m from the main perch. The territories were similar in size to those of evelynae. In the surrounding habitat, females were observed hunting insects, as is typical during breeding.
Both evelynae on New Providence and lyrura on Great Inagua produced a repertoire of vocalizations that included calls, scolding (agonistic) calls, and song (Figures 4 and 5; Supplemental Material Audio Files A, B, C, D, E, F, G, H, I, and J). In both populations, males and females produced “chip” calls when feeding and flying about (Figure 4A, 4B). These calls were composed of a single syllable that was repeated in a series. Calls varied in both the number and rate of syllables given in a series, from a single “chip” to a lengthy, rapid-fire sequence. Two additional calls were recorded only from evelynae on New Providence (Figure 4B). Male evelynae on New Providence occasionally gave a quiet “spurt” call during shuttle displays or fights. Fledglings gave “cheep” calls to their mothers near a nest, and adult birds also infrequently produced this call.
Male evelynae, and both sexes of lyrura, also gave loud scolding calls during agonistic interactions such as chases and fights (Figure 4C, 4D). Scolding calls were highly variable in both the length of the calls and the pattern of syllables and phrases. The scolding calls of evelynae and lyrura were primarily composed of 2 species-specific syllables given in the phrase “abb” that was repeated a variable number of times, or simply a single “a” syllable followed by a variable number of “b” syllables. The scolding calls of lyrura were more variable than evelynae, and lyrura frequently gave an additional scolding call that was composed of an “a” followed by a variable number of “c” syllables, and sometimes ending with a variable number of “b” and “abb” phrases.
Calls and scolding calls qualitatively differed between lyrura and evelynae, both in fundamental frequency and in length of syllables. The fundamental frequency of lyrura “chip” calls and scolding calls was between 1.5 and 3 kHz (Figure 4A, 4C). By contrast, evelynae “chip” calls and scolding calls had a much higher fundamental frequency, ranging between 7 and 9 kHz (Figure 4B, 4D).
Male lyrura and evelynae sang either from a perch (Figure 5) or as a part of their shuttle display (Figure 6). Within individual males, there were no noticeable differences between songs that were sung while perched or while displaying. In both taxa, songs consisted of a single stereotyped phrase that showed little or no variation in syllable composition (Figure 5). Song length varied within an individual solely though variation in the number of repetitions of this phrase.
The songs of lyrura were relatively short, lasting ≤5 s, and relatively simple (Figure 5A). Song phrases were composed of a single, broad-frequency syllable “a” given in pairs, and songs consisted of 1–4 repeated phrases. The songs of lyrura sounded similar to the sound of wet, squeaking shoes. Most distinctively, the songs were also very quiet. Unlike evelynae, lyrura songs were audible only within ∼5 m of the male under the quietest conditions, similar to the “whisper songs” reported for other hummingbird species (Skutch 1973).
The songs of evelynae were fast, high-pitched, and relatively long, lasting 3–27 s. They were also louder and could be heard 20 or 30 m from the bird. We recorded 2 distinct song phrases from evelynae that were apparently segregated by locality (Figure 5B). Song phrase 1 was recorded on the eastern side of New Providence (Figure 1: FW), whereas song phrase 2 was recorded on the western side of New Providence (Figure 1: CH, PW, and WB). The 2 song phrases were similar in overall quality and differed in both the type and cadence of syllables given in a repeated phrase.
Both male evelynae and lyrura readily performed shuttle displays to wild or caged females, and occasionally to wild males (Figure 6). More observations and field recordings were made of evelynae than of lyrura, so we present a detailed description of the evelynae shuttle display, followed by a comparison with the lyrura display.
We collected audio recordings of shuttle displays from at least 7 male evelynae (3 males in 2009 and 4 males in 2012) and 15 high-speed videos from 1 male. A male flapped his wings throughout the shuttle display and faced the female in an upright posture with gorget flared and tail widely spread and depressed (Figure 6A, 6B). The female actively followed the movements of the male with her head throughout the display.
The shuttle display of evelynae had 3 kinematically and acoustically distinct forms: the “initial” shuttle, the “typical” shuttle, and the “alternate” shuttle. Displays sometimes began with an “initial” shuttle, which was a brief, subtle behavior performed at the onset of the display. During the “initial” shuttle, a male performed a short shuttle segment (see below) and then paused to hover and look at the female. The “typical” form of the shuttle consisted of rapid, repeated, side-to-side (lateral) flights in which the male accelerated from rest, flew a distance past the female, and then decelerated (stopping only momentarily), all while facing the female (i.e. much of the motion was sideways). We term each individual lateral flight a “segment” because of their discrete, periodic nature. Shuttle segments were performed at a rate of 3.2 ± 0.24 Hz (segments per second, n = 7 males).
Similar to Calothorax and Archilochus, the shuttle segments were comparatively wide at the beginning of the display, up to 0.5–1 m, and decreased in amplitude with repetition to 20–30 cm, as the display bout progressed. From one segment to the next, the male often reversed direction, flying back the way he had come in the previous segment, or he sometimes continued at a large angle in relation to the previous segment. As a result, the display sometimes processed around the recipient in stages (Figure 6A).
At the end of each lateral shuttle segment, males produced a sharp buzzing sound with a trill rate of 74.5 ± 3.28 Hz (n = 7 males) that corresponded to the wingbeat frequency (74.3 ± 3.8; n = 8 videos of 1 male), indicating that this sound was a wing trill (Figure 6D; Supplemental Material Audio File K). The tail was held spread, with the stripes on R3 and R4 highly visible, as an apparent visual signal. Males also flicked their tails 1–3 times to the side, usually toward the second half of a shuttle segment (Figure 6C). High-speed videos of the shuttle display showed that the wings occasionally struck the outer rectrices during a tail flick, but that this did not occur frequently enough to explain the buzzing sound.
During the alternate shuttle, a male paused after a typical shuttle segment to hover in front of the female and sing with bill wide open (Figure 6B, 6D). In many displays, the male alternated singing with slowly drifting upward while producing a purring sound with bill closed (Figure 6B, 6D). There was a roughly 2:1 correspondence between the hovering wingbeat frequency (55.6 ± 3.8 Hz; n = 1 male) and the trill rate of the purring sound (31.9 ± 1.3; n = 6 males), suggesting that the wings also produced this sound. However, neither the precise mechanism nor the feathers responsible for the sound are clear from our data.
The time spent on the typical and alternate shuttles varied; some displays included only typical flight, whereas others mostly consisted of hovering after a brief period of lateral shuttling. Displays performed to females lasted a relatively long time (≤3 min) and included both typical and alternate forms. These bouts ended with the male pursuing the female or perching nearby or giving chase to an intruding male. If the female remained, males sometimes returned after a few minutes to perform another shuttle display. Displays to males were brief (≤10 s), mostly included typical shuttle motions, and inevitably ended in a chase.
We obtained audio recordings, but not video, of 3 shuttle displays from 2 male lyrura in October 2012. The first male performed a natural shuttle display that lasted 3 min, apparently to a female, but the entire display was out of our sight on the far side of a bush, so we were unable to observe the display kinematics. The second male performed a brief display with short shuttle segments, ∼20 cm in length, to a hatch-year male. The observed kinematics of this 1 male were generally similar to those of evelynae, with no obvious differences, other than a slightly increased shuttle cycle rate. The pattern of sounds produced during the display were also similar to those of evelynae: a series of buzz sounds with a trill rate of 74.2 ± 3.84 (n = 2 males) produced at the end of each shuttle segment, which were performed at a rate of 3.8 ± 0.48 Hz (Figure 6D; Supplemental Material Audio File L). Periods of shuttling were interspersed with periods of hovering and song. None of these displays included a purring sound (though this sound was not always present in displays of evelynae either).
Both male evelynae and lyrura infrequently performed display dives (Figure 7). We saw or heard ∼12 dives from 5 or 6 male evelynae, and we obtained audio recordings of 6 display dives from 3 males and high-speed video of 2 dives from 1 male. All dives were performed either to wild or to caged females (none to males), and they were usually performed immediately after a shuttle display. Males performed a single dive per bout that appeared to be oriented toward their main perch. A dive began with a male ascending 15–20 m high while zig-zagging slightly from side to side and then, without pausing, turning and diving in a J- or L-shaped trajectory (Figure 7A). At the bottom of the dive, the male would spread his tail repeatedly 6–9 times at a rate of 21.3 ± 0.6 Hz (n = 1 male) and produce the dive sound (Figure 7B; Supplemental Material Audio File M). Then, after passing over the female, the male would fly away while undulating from side to side with the tail widely spread. Males flapped their wings continuously throughout the display with a wingbeat frequency of 65 ± 2.1 Hz (n = 1 male).
At the bottom of a dive, males produced a dive sound that consisted of a series of 6 to 9 tonal notes with an average pitch of 0.94 ± 0.03 kHz (n = 3 males; Figure 7B). The notes were produced at a rate of 22.3 ± 0.4 Hz (n = 3 males), and each spread of the tail at the bottom of the dive corresponded to 1 note. The entire dive sound was relatively quiet and brief, lasting an average of 0.34 ± 0.05 s (n = 3 males).
We heard ≤10 dives, and obtained audio recordings of 3 display dives, from 1 male lyrura in October 2012. We saw only a portion of 2 dives, which prevented a detailed comparison of kinematics. The dive sound of lyrura was similar to that of evelynae, except that the average pitch of the notes was slightly higher, at 1.5 ± 0.2 kHz (n = 1 male; Figure 7B; Supplemental Material Audio File N). The lyrura dive sound also included an additional 1–3 longer notes produced later during the undulation phase. The undulation notes had an average pitch, similar to the notes produced at the bottom of the dive. The male spread his tail and produced notes at the bottom of the dive at a rate of 16.8 ± 0.3 Hz (n = 1 male), and the undulation notes measured from 1 dive had a rate of 2.7 Hz. The presence of additional undulating notes on the lyrura recordings, and not on the evelynae recordings, is probably due to greater sensitivity of the parabola used to record lyrura (vs. shotgun microphone for evelynae), rather than a biological difference.
When placed in the wind tunnel, outer rectrices (R5) of evelynae and lyrura produced sounds with a tip mode of flutter corresponding in frequency to their respective dive sounds (Figure 7C). Rectrix 5 of evelynae fluttered at 0.97 ± 0.01 kHz (n = 3 feathers), and lyrura fluttered at 1.2 ± 0.06 kHz (n = 3 feathers). The lyrura R5 produced sounds at significantly higher frequencies than evelynae R5 (t-test, P = 0.021). The frequency of sound varied only slightly with airspeed in both evelynae and lyrura (Figure 7C).
We sequenced 4 nuclear loci and 2 mitochondrial genes in individuals of evelynae from New Providence and lyrura from Great Inagua. This was supplemented by genetic data from 2 individuals of evelynae collected previously (Appendix Table 6). Loci ranged in length from 508 to 1,049 base pairs, and all loci exhibited within-population variation (Table 2). Estimates of θw per site ranged from 0.00032 to 0.0021, with the highest estimates found in MT-ND4 for both evelynae and lyrura. Measures of nucleotide diversity (π) within each population ranged from 0.028% for MUSK I3 in evelynae to 0.18% for MT-ND2 and AK1 I5 in evelynae (Table 2).
Genetic variation within populations of Calliphlox evelynae evelynae (from New Providence) and C. e. lyrura (from Great Inagua).
We also calculated average pairwise divergence between evelynae from New Providence and lyrura from Great Inagua (Table 3). Mitochondrial loci exhibited the largest divergence, 3.0% in MT-ND2 and 2.3% in MT-ND4, with 29 and 18 fixed differences, respectively. Average pairwise divergence between populations at nuclear loci ranged from 0.11% (ODC1) to 0.92% (MUSK I3), with 2 of the 4 nuclear loci exhibiting fixed differences (1 in FGB I7 and 5 in MUSK I3). We found no sites where polymorphisms were shared between both populations. Further, consistent with assumptions of the infinite-sites model, we found no evidence of any sites exhibiting multiple mutations.
Genetic divergence between populations of Calliphlox evelynae evelynae (from New Providence) and C. e. lyrura (from Great Inagua).
Our tests for significance of Fu's Fs and Tajima's D in every combination of locus and population found only 1 case in which these values were significant. The locus FGB I7 in evelynae had Fu's Fs and Tajima's D values that were significantly lower then expected under neutral evolution (P = 0.04 for both values). This may indicate that this locus, or a nearby linked region, is under selection.
Both lyrura and evelynae populations were recovered as reciprocally monophyletic sister populations in our concatenated analysis (Figure 8A). Further, their placement as sister taxa was supported by a posterior probability (PP) of 1.0 in the *BEAST species tree (Figure 8B). The 95% highest posterior density (HPD) *BEAST estimate for the divergence of the Great Inagua and New Providence populations was 0.41–0.96 mya (median = 0.69 mya). The lyrura–evelynae clade is estimated to have diverged from its most recent common ancestor with Mellisuga 1.3–2.4 mya (median = 1.8 mya).
All relationships were supported by a posterior probability (PP) of 1.0, with the exception of the clade containing C. evelynae and Mellisuga minima, which was resolved with a posterior probability of 0.51. This low probability appears to be a consequence of topological conflict among the sampled genetic markers. In gene trees generated by *BEAST, mitochondrial DNA resolves Mellisuga as sister to Archilochus (PP = 0.54); AK1 I5 resolves Mellisuga as sister to Calliphlox (PP = 0.97); FGB I7 resolves Mellisuga as nested within Archilohus (PP = 0.92); ODC1 resolves Mellisuga as sister to Archilochus (PP = 0.43); and MUSK I3 resolves Mellisuga as sister to Calliphlox (PP = 0.92). The gene-tree topologies recovered by *BEAST were otherwise consistent with our final species tree estimate. Individual gene trees estimated by MrBayes varied in their resolution but were generally consistent with those generated by *BEAST (data not shown).
Our analysis of morphological, acoustic, and genetic variation among populations of lyrura from the Inagua islands and among evelynae from the remaining islands of the Bahama Archipelago discovered consistent and diagnosable differences between the 2 populations. Our results confirm that the morphological characteristics originally used to describe lyrura as a separate species—iridescent forecrown and lyre-shaped tail feathers—are valid characters for distinguishing adult male lyrura from adult male evelynae (Figures 1 and 2). Iridescent head feathering and, possibly, tail shape serve as visual signals during the courtship displays that males of both populations perform to females. Moreover, kinematic and wind-tunnel evidence indicates that R5 in both lyrura and evelynae produces the dive sound (Figure 7). The sounds produced by lyrura tail feathers were significantly higher pitched than those of evelynae, which suggests that the different feather shapes between the 2 populations are responsible for the small divergence in the acoustic signals of the dive display.
Both sexes of evelynae and lyrura are diagnosed to their respective populations by their distinct calls and scolding calls (Figure 4). Male lyrura are additionally diagnosed from evelynae by distinctly different song (Figure 5). Although most bird vocalizations appear to be determined genetically, hummingbirds learn their songs (Gahr 2000). Cultural transmission via song learning may help explain the large differences in song between the recently isolated populations of lyrura and evelynae, as well as the potential dialects that we recorded from the evelynae of New Providence.
Breeding behaviors of evelynae and lyrura were similar. Like other members of the bee hummingbird clade, male evelynae and lyrura hold courtship territories during the breeding season. From these territories, males sing songs and perform 2 different elaborate aerial courtship displays, shuttles, and dives. Similar to Lucifer Hummingbirds and Slender Sheartails, males relied primarily on a complex and lengthy shuttle display that included song when displaying to a female on their territory, and only infrequently performed a single display dive (Figures 6 and 7; Scott 1994, Diza-Valenzuela et al. 2011). Males also infrequently performed shuttle displays to other intruding males, but these displays were generally shorter than those performed to females and did not include song.
Phylogenetic reconstruction based on nuclear and mitochondrial DNA sequences revealed that populations of evelynae from New Providence and lyrura from Great Inagua are reciprocally monophyletic, which is consistent with reproductive isolation (Figure 8). Although we only sampled genetic data from 2 islands (no other tissues were available), we found no evidence of a cline in morphology in evelynae or of individuals with an intermediate phenotype (Figures 2 and 3). We also recovered an average mitochondrial pairwise divergence of ∼2.7% between the 2 populations. This level of divergence is comparable to that found among other closely related sister species (Klicka and Zink 1997) and is greater than the divergence we estimated for other named sister taxa included in our phylogenetic reconstruction (i.e. D. eliza and D. enicura: ∼2.1% mtDNA divergence).
Furthermore, we estimate that divergence between evelynae and lyrura populations occurred between 0.41 and 0.96 mya. Fluctuations in sea level during this time may have played a role in preventing reproductive barriers from evolving between populations on most northern islands. The isolation of the Inagua bank from the rest of the Bahama Archipelago during times of sea-level minima has presumably maintained a geographic barrier between these 2 taxa. Thus, data from morphology, behavior, genetics, and geology support the recognition of lyrura and evelynae as separate species.
Numerous criteria for species delimitation exist (de Queiroz 2007). The two that are most commonly applied in delimiting avian species are the biological species concept (BSC) and the phylogenetic species concept (PSC). Under the PSC, species are defined as diagnosable evolutionary lineages (Cracraft 1983). Our results indicate that lyrura and evelynae fulfill this criterion. In addition to fixed genetic differences in both mitochondrial and nuclear loci, and reciprocal monophyly in our phylogenetic analysis, the 2 populations can be diagnosed by morphology, vocal repertoire, and the mechanical sounds produced by tail feathers.
Under the BSC, species must be reproductively isolated (Mayr 1942). This criterion is difficult to apply in the case of allopatric populations such as evelynae and lyrura. Only indirect evidence can be used to infer reproductive isolation, such as divergence in sexually selected traits that could facilitate the existence of premating isolation barriers (Alström and Ranft 2003). We have found evidence that adult male evelynae and lyrura diverge in several characters associated with courtship displays. These potentially sexually selected traits include songs, dive sounds, tail-feather shape, and extent of iridescent plumage. Divergence in both acoustic and morphological traits associated with courtship display is consistent with reproductive isolation.
Our findings suggest that lyrura is best considered a full species. However, the most appropriate generic placement of evelynae and lyrura is unclear. Anatomical (Zusi 2013) and molecular phylogenetic analyses (McGuire et al. 2014) indicate that Calliphlox is polyphyletic, and that evelynae and lyrura are not woodstars and instead fall within a well-supported clade containing Archilochus and Mellisuga minima (Figure 8). However, within this clade, relationships among genera are weakly supported. One possibility is to place evelynae and lyrura in a resurrected Nesophlox Ridgway (1910), for which evelynae is the type. Alternatively, evelynae, lyrura, and the 2 members of Archilochus could be placed in Mellisuga Brisson (1760), which has nomenclatural priority over Archilochus. Finally, evelynae and lyrura have previously been placed in a clade with Calothorax and Doricha based on phenotypic similarity (Schuchmann 1999, Howell 2002), but this hypothesis currently lacks molecular phylogenetic support (McGuire et al. 2014).
Several common names have previously been given to lyrura: Lyre-shaped Woodstar (Gould 1887), Lyre-tailed Hummingbird (Cory 1880), Inagua Woodstar (Cory 1918), and Inagua Sheartail (Howell 2002). “Lyre-tailed” refers to the uniquely shaped tail feathers of males, which originally characterized the species, whereas “Inagua” is the appropriate toponym. Given the unique, outwardly curving tail feathers of males and their endemic Inaguan geography, we recommend the common name Inaguan Lyretail.
The geographic range of lyrura is small (restricted to Great and Little Inagua), so a discussion of our limited data relevant to conservation is warranted. A significant fraction of Great Inagua is mangrove and saltwater lake, which appeared to be largely unsuitable habitats. Assuming a density of 20 birds km−2 in low coppice, the population may be as low as a few thousand birds. In the dry season the birds appeared to be food limited, in that they were scarce in natural habitats but abundant in gardens in town, whereas in the rainy season the opposite was true. This suggests that limited additional development could actually benefit lyrura if it increased this resource base, as has happened in several North American species of hummingbirds (Zimmerman 1973, Clark and Mitchell 2013). Present human activities are unlikely to have a direct negative impact on the species, and we did not identify any apparent short-term threats to the population. One clear long-term threat is sea-level rise, because the Inagua islands, like much of the Bahamas, are close to sea level. The Bahamas National Trust protects a significant fraction of the island, and lyrura has protection under Bahamian law, so it does not appear to meet the IUCN criteria for a designation of “vulnerable” at this time. Additional surveys and data on the population status of lyrura, particularly in eastern Great Inagua (which we did not survey, because of inaccessibility), would provide a better baseline for future monitoring of this Inaguan endemic.
We appreciate assistance provided by C. Wardle, Mrs. Blackwell, A. Hepburn, J. Marks, A. Varma, H. Nixon, T. Rahming, and the Bahamas National Trust. We thank R. Prum and two anonymous reviewers for valuable comments on the manuscript. National Science Foundation (NSF) grant IOS-0920353 and the W.R. Coe Funds from Yale University provided funding for this research. This material is based on work supported by the NSF Graduate Research Fellowship Program under grant DGE-1122492. This project was conducted under permits from the Bahamas Environment, Science and Technology commission (BEST). A. Varma kindly allowed use of two photos. The following programs are available online: Raven version ( http://www.birds.cornell.edu/raven), LogCombiner ( http://beast.bio.ed.ac.uk/logcombiner), and TreeAnnotator ( http://beast.bio.ed.ac.uk/treeannotator).
Specimens for which genetic data were collected in this study.
The PCR primers used in this study.
Specimens from McGuire et al. (2014) used in this study.