Proposal (720) to
Treat White-breasted Wood-Wren Henicorhina leucosticta as two or more species
Henicorhina leucosticta inhabits lowland humid forest from eastern Mexico to central Colombia and the Chocó region of northwestern Ecuador, and also in the western and northern Amazon Basin (Fig. 1). H. leucosticta varies subtly in plumage throughout its distribution, and is classified into a dozen or so subspecies based largely on plumage differences (Table 1). H. leucosticta also shows geographic variation in vocal traits including pitch and repetition of song style (Brewer 2001).
Two phylogenetic studies have examined geographic patterns of mtDNA in H. leucosticta (Dingle et al 2006, Lelevier 2008). Both studies identified three major clades—one in Central America, a second in the Chocó of northwest Ecuador, and a third in the Amazon basin. These three clades are deeply divergent, with estimated 7–12% mtDNA sequence divergence between clades (Lelevier 2008). Additional sampling in Panama in the Lelevier study revealed that birds in the Darien of eastern Panama are sister to the Amazon clade. These Darien birds could be considered part of the Amazon clade or a distinct “Darién” clade; for clarity when discussing different populations, we follow the latter course in this proposal. The geographic distribution of the Darién clade remains uncertain, though it contacts i) the Central American clade in central or eastern Panama and ii) the Chocó clade, likely somewhere in western Colombia. More recently, Smith et al (2014) included H. leucosticta in their large phylogeographic analysis of Neotropical lowland taxa, and, using genomic data, recovered very similar relationships to those previously identified using mtDNA (see screenshot below).
Fig. 1 from Pegan et al 2015. Range map of H. leucosticta illustrating the distributions of four mitochondrial clades (A = Central American, B = Darién, C = Amazon, D = Chocó; the precise geographic distribution of the Darién clade is unknown and marked in dashed lines). The location of La Selva Biological Station in the Caribbean lowlands of Costa Rica, where fieldwork for the Pegan et al. study was completed, is marked by a star. Map from BirdLife International (2014).
Table 1 from Pegan et al. 2015. Summary of subspecies and their clades based on Lelevier (2008). Birds from western Colombia are presumed to be within the Chocó clade pending further evidence.
Close up of Figure S10 from Smith et al. 2014 showing time calibrated gene tree. Note that Smith et al. designate the clade sister to the Amazonian populations found in the Darien region of eastern Panama– that we are calling “Darién” in this proposal— as “Choco”.
Recently, Pegan et al (2015) used vocal trait analysis and field playback experiments to assess premating reproductive isolation in song in H. leucosticta. While the utility of these approaches are well understood for suboscines, divergence in vocalizations between allopatric populations can also be relevant to species limits in oscines.
Vocal trait analyses: Pegan et al. analyzed eight song variables (song length, total note count, mean number of notes per syllable, mean note rate, peak frequency, mean note high frequency, mean note low frequency, and total song frequency range) for a representative song from a total of 49 recordings from three H. leucosticta clades (n = 19 from the Costa Rican population, n = 10 from the Chocó population in northwestern Ecuador and western Colombia, and n = 20 from the Amazon population), then visualized population differences in vocal traits using the first two axes from a principal components analysis (PCA) that quantified variation across the nine vocal characters (see Fig. 2).
Fig. 2 from Pegan et al. 2015. PCA plot of acoustic space.
Larger values on PC1 represent songs that are longer, that have more notes, and that cover a wider range of frequencies. Larger values on PC2 represent songs with faster note rates and smaller frequency ranges. Although variation in these vocal characters is continuous across the entire sample, individual populations are differentiated in acoustic space. In the figure below yellow represents birds from Costa Rica; Amazonian populations are shown in red; and blue denotes samples from Chocó (northwestern Ecuador).
Field playback experiments: Pegan et al. (2015) conducted field experiments on a single population of H. leucosticta at La Selva Biological Station, Heredia province, in the Caribbean lowlands of northeastern Costa Rica. Each experiment contained two treatments; a sympatric treatment of local Costa Rican song playback (H. l. costaricensis) and an allopatric treatment of song playback from a different population. One experiment (hereafter the “Amazon experiment”) used playback of vocalizations from WBWWs from Amazonia as the allopatric treatment (H. l. hauxwelli) whereas the other (hereafter the “Chocó experiment”) used playback of Chocó clade vocalizations (H. l. inornata) as the allopatric treatment. Trials for these two experiments were carried out on different territories at La Selva and analyzed independently.
Fig. 3 from Pegan et al 2015. Behavioral response to playback experiments.
Responses to the Amazon experiment are on the left and responses to the Chocó experiment are on the right. Latency to approach (A), closest approach to speaker (B), total number of songs (C) and latency to vocalize (D). Boxplots illustrate median (horizontal black bar), first and third quartiles (boxes), and minimum and maximum values (points and dotted lines). Raw data are plotted as points in front of the boxplots, with points offset slightly to better display values. Sympatric treatments elicited aggressive responses (fast approaches, close approaches, many songs, low latency to vocalize) in both experiments. Allopatric treatments elicited asymmetric responses: individuals typically responded aggressively to playback of Chocó songs but not to playback of Amazonian songs.
Fig. 4 from Pegan et al. 2015. Percent of individuals approaching the speaker in each experiment.
Responses to the Amazon experiment are on the left and responses to the Chocó experiment are on the right. Note that sympatric trials serve as positive controls (wrens approached the speaker in all sympatric trials).
Additional points to consider:
Pegan et al. did not consider the Darien clade in analysis of vocal traits or playback experiments, as the geographic limits of this clade beyond the Darien region in eastern Panama remain unclear and there are few recordings from the Darien in Panama. Though not quantified, birds from the Darien are vocally similar to Amazonian population, black-capped (as are Amazonian birds), and genetically sister to the combined Amazonian populations (e.g. see Smith et al. topology). In addition, recent (March-April 2016) fieldwork indicates that H. leucosticta in Costa Rica largely ignore songs from the Darien clade.
Pegan et al. did not conduct reciprocal playback analyses. Responses to playback are not always symmetrical between populations (Colbeck et al 2010)—Pegan et al. interpret the lack of Central American response to Amazonian playback as evidence of premating reproductive isolation between these two populations, but did not have the opportunity to measure how Amazonian birds respond to vocalizations from Central American birds. Given the marked differences in vocal traits, however, it seems reasonable to infer that Amazonian birds are unlikely to respond strongly to Central American vocalizations.
Third, Pegan et al. tested a territorial response rather than a mate choice response. In general, research suggests that females are more discriminating than males when it comes to songs, implying that if males discriminate between songs, as we demonstrate, it is likely that females do as well (Cure et al 2010, Danner et al 2011, Nelson et al 2004, Searcy 1990, Searcy and Brenowitz 1988). All vocal trait analyses and playback experiments aimed at determining species limits make the assumption that territorial songs and mate choice are correlated, and though this is generally accepted as a valid assumption, it may not be true in all cases, especially in species with multiple song types where certain types may be used for territoriality and others for mate attraction.
The playback experiments presented in Pegan et al (2015) show that H. leucosticta in Central America do not appear to recognize H. leucosticta song from the Amazon clade as conspecific (but, interestingly, do generally recognize song from the even more genetically divergent Choco clade as conspecific). These field experiments echo the results of a vocal trait analysis that indicates songs from the Amazonian and Central American clades are mostly (but not entirely) separated in acoustic space. Genetic data also shows deep divergence between the Central American and Amazon clades. In sum, available evidence suggests that Amazon and Central American populations of H. leucosticta represent distinct biological species and should not be classified as conspecific.
The suggestion that multiple biological species lurk within H. leucosticta is hardly novel, and we expect future work (e.g., in the zones where the Darien clade contacts the Central American and Choco clades, also within the Amazonian population) will lead to further insights relevant to H. leucosticta taxonomy. We note that H. leucosticta as currently defined is paraphyletic with respect to H. leucoptera, and that this complex remains paraphyletic in Options 1, 2 and 5 (but not 3 and 4).
We support Option 2 (or Option 1, though we believe Option 2 is preferable) as a first step forward in improving the species-level taxonomy of this complex group. Option 3 also is a reasonable interpretation of the available evidence. Options 4 and 5 are offered for the sake of completeness, but in our view are not well-supported at this time. Nomenclature can be considered at a future date if votes favor one of the splitting options.
Option 1: Split Amazonian population of H. leucosticta from all populations found west of the Andes. This reflects the finding of Pegan et al. that H. leucosticta in Costa Rica largely ignore song from the Amazonian population.
Option 2: Split Amazonian and Darien populations of H. leucosticta from all populations found west of the Andes. This reflects the finding of Pegan et al. that H. leucosticta in Costa Rica largely ignore song from the Amazonian population and also the close relationship between Amazonian and Darien populations; these two clades are sister groups (Smith et al 2014), share plumage traits (both black capped; Choco and Central American birds have brown caps), and are similar acoustically (unpublished playback experiments also indicate that Costa Rican H. leucosticta largely ignore song from the Darien population).
Option 3: Split H leucosticta into three – the Central American clade, the Choco clade, and the Amazonian + Darien clade. There is no current behavioral evidence indicating premating reproductive isolation between the Choco and Darien clades, but they are highly genetically divergent and it would not be surprising if the two clades behave as biological species wherever they come into contact in western Colombia.
Option 4: Split H leucosticta into four– the Central American clade, the Choco clade, the Amazonian clade and the Darien clade. This arrangement was proposed to be the species tree of this complex by Smith et al (2014).
Option 5: No action for now – retain the status quo.
BirdLife International and NatureServe. Bird species distribution maps of the world. BirdLife International, Cambridge, UK and NatureServe, Arlington, USA; 2014
Brewer D. Wrens, Dippers, and Thrashers. New Haven: Yale University Press; 2001.
Cadena CD, Caro LM, Caycedo PC, Cuervo AM, Bowie RCK, & Slabbekoorn H. Henicorhina anachoreta (Troglodytidae), another endemic bird species for the Sierra Nevada de Santa Marta, Colombia. Ornitología Colombiana 2015;15.
Colbeck G, Sillett T, Webster M. Asymmetric discrimination of geographical variation in song in a migratory passerine. Anim Behav. 2010;80: 311–318. doi: 10.1016/j.anbehav.2010.05.013
Curé C, Mathevon N, Mundry R, Aubin T. Acoustic cues used for species recognition can differ between sexes and sibling species: evidence in shearwaters. Anim Behav. 2010;84: 239–250. doi: 10.1016/j.anbehav.2012.04.039
Danner JE, Danner RM, Bonier F, Martin PR, Small TW, Moore IT. Female, but not male, tropical sparrows respond more strongly to the local song dialect: implications for population divergence. Am Nat. 2011;178: 53–63. doi: 10.1086/660283. pmid:21670577
Dingle C, Lovette IJ, Canaday C, Smith TB. Elevational zonation and the phylogenetic relationships of the Henicorhina wood-wrens. Auk. 2006;123: 119–134. doi: 10.1642/0004-8038(2006)123[0119:ezatpr]2.0.co;2
Lelevier M. Phylogeography of three widespread Neotropical avian taxa: Rufous-tailed Hummingbird, White-breasted Wood-Wren and Anthracothorax mangos. Ph.D. Thesis, University of Alaska Fairbanks. 2008. Available:
Nelson DA, Soha JA. Male and female white-crowned sparrows respond differently to geographic variation in song. Behaviour. 2004;141: 53–69. doi: 10.1163/156853904772746600
Pegan TM, Rumelt RB, Dzielski SA, Ferraro MM, Flesher LE, et al. Asymmetric Response of Costa Rican White-Breasted Wood-Wrens (Henicorhina leucosticta) to Vocalizations from Allopatric Populations. PLoS ONE 2015; 10(12) e0144949. doi: 10.1371/journal.pone.0144949
Reif J, Jiran M, Reifová R, Vokurková J, Dolata PT, Petrusek A, et al. Interspecific territoriality in two songbird species: potential role of song convergence in male aggressive interactions. Anim Behav. 2015;104: 131–136. doi: 10.1016/j.anbehav.2015.03.016.
McEntee JP. Reciprocal territorial responses of parapatric African sunbirds: species-level asymmetry and intraspecific geographic variation. Behav Ecol. 2014;25: 1380–1394. doi: 10.1093/beheco/aru136.
Searcy WA. Species recognition of song by female red-winged blackbirds. Animal Behaviour. 1990:40; 1119–1127. doi: 10.1016/s0003-3472(05)80178-2
Searcy WA, Brenowitz EA. Sexual differences in species recognition of avian song. Nature. 1988;332: 152–154. doi: 10.1038/332152a0
Seddon N, Tobias JA. Character displacement from the receiver’s perspective: species and mate recognition despite convergent signals in suboscine birds. Proc Biol Sci. 2010;277: 2475–83. doi: 10.1098/rspb.2010.0210. pmid:20375056
Smith, BT, McCormack, JE, Cuervo, AM, Hickerson, MJ, Aleixo, A, et al. The drivers of tropical speciation. Nature 2014;515(7527):406-409.
Benjamin Freeman and Teresa Pegan, May 2016
Comments from Stiles: “NO. This proposal is part of a larger problem for which the solution is not yet at hand. The authors present evidence that the northern form responds differently to its own songs vs. those of an Amazonian form (although the playback experiments were not reciprocal), suggesting that the two could be considered separate species. However, the combination of vouchered specimens with song recordings and DNA samples from across the wide range of the leucosticta complex is yet too incomplete to fully resolve the puzzle. The form or forms intervening between the northern and Amazonian birds are especially poorly sampled. My only personal contribution to the debate is a bird of this group that I recorded in Puerto Solano, NW Colombia that sounded so different from the birds I knew in Costa Rica that I didn’t even recognize its song as that of a wren until I played it back and had the bird screaming it back at me from 3m away! So, I’d put their suggestion on hold for now: suggesting species status for the birds at the extremes of the distribution of the leucosticta complex without knowing what lies between does not seem justified to me at this time.”
Comments from Areta: “NO. As underscored by Gary, I think that there are still too many missing pieces in this puzzle. I would like to see more thorough reciprocal playback experiments using larger sample sizes and better acoustic characterizations of populations being tested. This paper is one step forward in the taxonomy of Henicorhina leucosticta, but more need to be taken before taking a robust decision. With the data at hand (and with the playback experiments suggesting stronger responses to more distantly related populations), we cannot place unsampled (or sampled!) populations confidently to species-level taxa.”
Comments from Robbins: “NO. I fully concur with comments made by Gary. Much more information is needed before we start subdividing this complex.”
Comments from Pacheco: “NO. The consensus is that there is most likely more than one species in this complex. But I agree with the views of colleagues that the evidence is still incomplete for a good decision.”
Comments from Remsen: “YES on Option 2. Evidence from these excellent studies is overwhelming that that a minimum of two species are represented, and I see no real reason to delay at least a 2-way spit until other studies further subdivide those two (other than likely problems with English names as daughters are further subdivided).”
Comments from Claramunt: “NO. I agree that more than one species may be involved, but the data published so far are not sufficient to redefine species limits. The mitochondrial data is suggestive of population structure but by itself is not sufficient to demonstrate the existence of more than one lineage. I’m not impressed by the deep mitochondrial “divergences”, either. Such deep “divergences” could be just old coalescences, which seem common in widespread lowland tropical species, a consequence of large, old, geographically structured, and stable populations. Note also that data in Smith et al. (2014) is also mitochondrial, not genomic/nuclear; therefore, the congruence among datasets is expected (they are all mitochondrial). Congruence among mitochondrial and nuclear datasets would have been a strong test for the existence of more than one lineage, but we don’t have that information so far. Variation in external/morphological traits is apparent but an analysis is needed to determine how these traits vary, whether they vary in a continuous or discrete fashion, whether they define coherent groups, and whether those groups coincide or conflict with the mitochondrial groups. Thus, much more data and work is needed to test whether there are one or more lineages in this complex. Also critical to determine species limits in this group would be to analyze more samples from the Chocó, where the Darien and Chocó populations could be in contact.
“The vocal data are fairly inconclusive. First, variation in song characteristics seems continuous instead of defining discrete groups. As for their geographic position, songs from the Chocó occupy an intermediate position in the multivariate space, suggesting that vocal geographic variation may be clinal. Playback experiments further highlight the potential for the clinal variation in songs to allow for species recognition and interbreeding throughout the range of this species: birds from Costa Rica responded strongly to songs from the Chocó, and to a lesser degree to Amazonian vocalizations. Therefore, it is possible that neighboring populations do recognize each other as conspecific. Then, the relevant test for assessing reproductive isolation (and species status under the “Biological” species concept) would be to analyze a more dense chain of neighboring populations to see if there is some sort of break in species recognition. If there is no break and neighboring populations recognize each other as conspecific across the entire range, there is no intrinsic barrier to gene flow. “Isolation tests” performed on widely allopatric populations (e.g. Costa Rica versus Amazonia) are irrelevant if they are connected by intermediate populations with intermediate characteristics and gene flow can occur across the entire range.”
Comments from Zimmer: “NO”. The evidence presented certainly suggests that more than one biological species is involved, and my own field experience has likewise indicated as such, at least as regards vocal differences. However, I am troubled by the fact that the conclusions of the vocal analysis and playback experiments rest on a comparison of populations near the geographic extremes of the rather extensive distribution, without a clear picture of what’s happening in between. Problems that I see include the following: 1) The playback experiments were one-way, not reciprocal, so that we know the response of the Costa Rican birds to songs of the Amazonian clade and the Chocó clade, but we know nothing of how those populations respond or fail to respond to playback of the Costa Rican birds. 2) The Darién clade is not included in the experiments or the vocal analysis. 3) The geographic limits of the Darién clade are unknown. 4) There are two named subspecies from western Colombia and nw Ecuador whose inclusion in the Chocó clade is only presumed. 5) In playback trials, Costa Rican birds did not respond to playback of Amazonian birds, but did respond strongly to playback of Chocó birds, which are even more genetically divergent from the Central American clade than are the Amazonian birds. On top of all of this, I am even more concerned about the vocal analysis itself. According to the proposal, Pegan et al (2015) analyzed eight song variables, which is great. But I can’t help but wonder how they addressed the issue of song-types. My experience with Henicorhina is that each population of each species has a song repertoire that includes multiple song-types, and that the size of the song repertoire may be individually variable. I vividly remember an experience with a White-breasted Wood-Wren from my early days of Neotropical birding at Palenque (Chiapas, Mexico). I tape-recorded natural (unsolicited) songs of a leucosticta, and played those songs back, in an attempt to see the bird. In response to playback, the wren in question immediately switched to a 2nd, distinctly different, song-type. I recorded song-type #2 and played that back, which stimulated the wren to switch to a 3rd song-type. This was repeated until I recorded song-type #5 and played it back, only to have the wren recycle song-type #1! It would seem that any vocal analysis comparing only one song-type from Population A to one song-type from Population B could produce seemingly diagnosable characters that are, in fact, merely reflecting sampling bias. Conversely, failure to control for song-type discrimination (i.e. including more than one song-type in the same analysis) within the sample from a single population, could introduce all kinds of variation in the measurement of individual characters, such that populations could appear to overlap in characters, when, in fact, it’s a case of apples to oranges. Given that these are oscines that, at least to a certain extent, learn their songs, I would like to see a bit more rigor applied to the vocal analysis, and would like to have a lot more information from the central portions of the geographic distribution (rather than just from the extremes) before attempting to assess species-limits, even though I believe the authors are correct in their assertions that at least two different species are involved.”