Song patterns support species status for some, but possibly not all, island populations of House Wren ( Troglodytes spp. ) in the Lesser Antilles

11 I am grateful to Karen Rendall for field assistance in 2015; to Rochelle Bellemare and Rupert Radix for 12 logistical support at Simla Field Research Station and the Asa Wright Nature Center in Trinidad; to 13 ProAves and OTS, respectively, for visits to Copalinga Reserve in Ecuador and the Las Cruces Biological 14 Research Station in Costa Rica; and to Theresa Burg for several helpful discussions of the topic and 15 comments on the manuscript. Research funding was provided from the Natural Sciences and 16 Engineering Research Council of Canada (NSERC) through the Government of Canada’s Tri-Agency 17 research funding program, the University of Lethbridge, and the University of New Brunswick. 18 Research adhered to guidelines of the Canadian Council on Animal Care and was conducted in 19 accordance with approved animal welfare protocols from the University of Lethbridge (AWC#1023; 20 AWC#1429) and the University of New Brunswick (ACC#18041, ACC#19055). I report no conflicts of 21 interest and am responsible for all aspects of the research reported. In accordance with institutional 22 and Tri-Agency policy, research data will be deposited in the UNB-DATAVERSE. 23 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 2

(279 words) 24 Island populations are special for the study of evolutionary processes and can be a zone of 25 incipient speciation. Recently, several island populations of House Wren in the Lesser Antilles 26 (Dominica, St Lucia, St Vincent, and Grenada), formerly recognized as subspecies of the continental 27 form, were reclassified as distinct species. However, much of the supporting data was fragmentary in 28 its sampling of the different islands or equivocal in the patterns observed. Because song is a core 29 element of mate recognition and choice, and can therefore be a key character in species 30 identification, I report here the first detailed characterization and analysis of song for House Wren on 31 all of the islands of the Lesser Antilles where they remain, including Trinidad and Tobago; and 32 compare song patterns across the different islands as well as to several continental populations. 33

show that song is broadly similar across all of the islands and to continental populations in 34 high-level features of its structure, organization and delivery but is discriminably different among 35 many of them in its more detailed features. The latter differences are consistent with the recent 36 species splits, with the possible exception of Grenada. They also support retention of House Wren on 37 Trinidad and Tobago as subspecies of the continental form. Results also point to the possibility of a 38 central American origin for some of the islands and a south American origin for others, yielding a trait 39 mosaic where islands that putatively share the same geographic origins, and are therefore 40 presumably genetically closest, are not the most similar in patterns of song (or plumage). This pattern 41 would therefore entail multiple intriguing instances of convergent evolutionary divergence among 42 them that warrants further detailed study. 43

House Wren, convergent evolution, island populations, bird song, speciation, taxonomy 44 45 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 3 Lay Summary 46 • I provide the first comprehensive analysis and comparison of song patterns of House Wrens for all 47 of the islands of the Lesser Antilles where they remain, some of which are at risk of extirpation, or 48 even extinction if they represent distinct species. 49 • I use the patterns to interpret the recent taxonomic reclassification of many of these island 50 populations as distinct species. 51 • In their general structure, organization and delivery male song is similar across all of the islands 52 and follows patterns common to contintental forms of House Wren distributed broadly across 53 North, Central and South America. 54 • Songs of the different islands are, however, discriminably different in their more detailed features 55 and these differences are consistent with most, but possibly not all, of the recent species splits. 56 • For the island populations recently reclassified as different species, the distinctiveness of male 57 song is greatest in Dominica and St Vincent and to a lesser extent also St Lucia, and least distinctive 58 in Grenada. Song in Trinidad and Tobago is not substantively different from populations in 59 mainland South America which supports retaining these two island populations as subspecies of 60 the closest continental forms. 61 • Song patterns also point to different possible continental sources for some of the island 62 populations: a source in Central America for Dominica and St Lucia; and a source in South America 63 for the rest. If true, this creates multiple instances of convergent evolutionary divergence in trait 64 patterns across the various islands which merits further study. 65 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 4 Island populations are special for evolution both because their physical isolation restricts gene 66 flow, which allows accumulation of even neutral changes, and because their often distinct ecologies 67 can also yield adaptive diversification. As a result, island populations may differentiate relatively 68 rapidly and therefore can be zones of incipient speciation. Darwin’s finches of the Galapagos Islands 69 are a classic example: a relatively rapid radiation of different finch species on different islands that 70 vary relatively subtly in plumage and overall size, as well as in beak size and shape based on variation 71 in foraging niche. 72 One challenge with definitively establishing species status for island populations – or any 73 population – is knowing how much differentiation is enough because, although there are many 74 proposals, there is no definitive guide to how much variation in typical taxonomic characters, 75 including genes, is required (reviewed in Sukumaran and Knowles 2017; Winker 2021). And the 76 insularity of island populations complicates matters because it effectively precludes interbreeding 77 even if it might occur otherwise. Here song is often given some priority in taxonomic reviews given its 78 role in mate attraction, recognition and mate choice (Remsen 2005; see also Alström and Ranft 2003), 79 and there are many examples of cryptic speciation revealed largely or only on the basis of differences 80 in song patterns (e.g., Irwin 2000; Valderrama et al. 2007; Toews and Irwin 2008). Notably, Podos and 81 colleagues (Podos et al. 2004; Huber & Podos 2006; Podos 2010) documented song variation within 82 and among several species of Darwin’s finches traceable to differences in beak size and shape, 83 illustrating how divergence in feeding niches can have knock-on effects for factors more directly tied 84 to species recognition and status (see also Grant and Grant 1996). 85 The House Wren (Troglydytes spp.) is another species complex well suited to contributing to 86 these evolutionary questions. House Wrens are widely distributed across North, Central and South 87 America. Indeed, they have one of the widest latitudinal distributions of any songbird in the western 88 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 5 hemisphere, breeding from approximately 550 north in Canada to 550 south at the southern tip of 89 South America. Across this vast range, house wrens inhabit a diversity of environments and include 90 virtually continuously distributed continental populations but also several island populations in 91 Mexico, Central and South America, and the Lesser Antilles. Hence, it is a taxon that is well suited to 92 examining how variable environments and ecologies affect migration, social and mating systems, and 93 life-history traits, and there is a large and important literature on many of these topics (see taxon 94 reviews by Fernandez et al. 2024; Johnson 2024). 95 With widely distributed continental forms but also a number of insular island forms, it is a taxon 96 that is also especially well-suited to testing how continuity and discontinuity in distribution can affect 97 population genetic structuring and speciation. This latter important evolutionary issue has long been 98 contested for house wrens, as the classification and species status of populations in this taxon has 99 been complicated and regularly debated. At times, the continental populations have been split into 100 three different species: T. aedon (North America), T. brunneicolis (northern Mexico), and T. musculus 101 (Central and South America), while at other times they have been lumped into a single super-species 102 (T. aedon), with more than 30 different subspecies (see for example, Brumfield and Caparella 1996; 103 Brewer 2001; Kroodsma and Brewer 2005; Dickinson and Christidis 2014). 104 Until recently, the latter framework of a single super-species was formally accepted but not 105 without continuing calls to recognize species status for one or more of the continental forms, as well 106 as several of the island forms. And, indeed, a very recent re-evaluation (Chesser et al. 2024) restored 107 two of the formerly recognized continental species (the northern House Wren: T. aedon; and the 108 southern House Wren: T. musculus) and elevated to species status three of the formerly recognized 109 island subspecies off the coast of Mexico (Isla Cozumel: T. beani; Isla Socorro: T. sissoni; Isla Clarion: T. 110 tanneri) as well as all of the island subspecies in the Lesser Antilles (Dominica: T. martenicensis; St 111 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 6 Lucia: T. mesoleucus; St Vincent: T. musicus; Grenada: T. grenadensis) with the exception of those 112 closest to the mainland on Trinidad (T.m. clarus) and Tobago (T.m. tobagensis). 113 These recent taxonomic changes were based on a mix of genetic, morphological, plumage, and 114 song characteristics (Chesser et al. 2024). The data supporting species status for the two continental 115 forms and the island forms near Mexico appear clear and compelling. The genetic data involved both 116 mitochondrial and nuclear DNA for a large sample of house wrens across their broad continental 117 distribution, as well samples from all of the island populations of Mexico (Klicka et al 2023); while the 118 song data involved a similarly comprehensive sample for continental populations as well as the island 119 forms in Mexico, including experimental responses to song playbacks for the latter island populations 120 (Sosa-López and Mennill 2014a; Sosa-López et al. 2016; see also Sosa-López and Mennill 2014b,c). 121 However, the data in support of species status for each of the island populations in the Lesser 122 Antilles are less clear and complete. For example, the genetic sample for the Lesser Antilles group did 123 not include either St. Lucia or Tobago; and it involved only mitochondrial DNA and not nuclear DNA 124 (Klicka et al 2023). The mtDNA results suggested distinct genetic structuring for the island forms that 125 were sampled, as would be expected given their insularity, and also pointed to an intriguing possible 126 difference in the source of colonization of the different islands either from mainland South America or 127 Central America. However, the authors themselves ultimately favored using their nuclear DNA data to 128 resolve taxonomic relationships in the House Wren complex because the mtDNA produced some odd 129 instances of paraphyly that were inconsistent with traditional taxonomy and the general geography 130 (Klicka et al. 2023). Thus, the authors concluded that, “The greater consistency of the RADseq [nuclear 131 DNA] results and traditional taxonomy and geography suggests it is the more trustworthy 132 phylogenetic pattern … the RADseq data provide the most appropriate basis for classification and 133 understanding House Wren evolution…”. Consequently, these authors recommended recognizing 134 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 7 separate species status for northern and southern House Wren, as well as three of the island forms 135 from Mexico, but they specifically did not make any recommendations concerning the island 136 populations in the Lesser Antilles. 137 The song data for the Lesser Antilles are also limited. Although their songs are often popularly 138 described as distinctive, there has been no systematic analysis and comparison of song across the 139 various islands. The most relevant and comprehensive comparative work cited by the classification 140 committee is the study by Sosa- López and Mennill (2014). While comprehensive in every other 141 respect, that study was limited in its sampling for the Lesser Antilles: it included only two songs from 142 Dominica, one of which was noted as possibly being from St Vincent; there were no song samples 143 included for any of the other islands. Unfortunately, there are few other formal sources that speak 144 compellingly to the matter of song differences among the islands. There are only a handful of short 145 reports, often with a different primary focus, that offer only subjective, impressionistic descriptions of 146 one or a few songs from a single island. For example, Gilardi and John (1998), also cited by the 147 classification committee, performed an important census and study of the general behavior and 148 breeding biology of house wrens in St Lucia, which are seriously threatened on that island. Their 149 paper is most valuable in other respects but provides only a single spectrogram of one song noting 150 that songs are “highly variable”. Similarly, there is an informal report by Barlow (1978) for house 151 wrens on Gaudeloupe, where the birds are now believed extinct, which provides a spectrogram of 152 one song from that island compared to a recording made of a house wren in Florida (which is not 153 typically part of the species’ breeding range) noting that “to my ear, the song of the Gaudeloupe 154 House Wren is the louder, richer and more melodic of the two”. More recently, Cyr et al. (2021) 155 conducted a much more systematic study of song for Grenada; however, that study was focused 156 specifically on the issue of anthropogenic influences on song for birds living in urban (disturbed) 157 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 8 versus rural (undisturbed) locations only on the island of Grenada and involved no data or 158 comparisons to song for any other islands in the Lesser Antilles. Hence, to date there simply is no 159 comprehensive study and comparison of song patterns available for the island group of house wrens 160 in the Lesser Antilles. 161 There is obvious plumage variation among the island forms, as also noted in the recent 162 reclassification, and illustrated in Figure 1. However, here again, the patterns have been described 163 only historically and in impressionistic terms (Oberholser 1904); they have not been studied more 164 recently or systematically, and the patterns of variation are not entirely straightforward. Thus, birds 165 on the islands closest to the mainland in Trinidad and Tobago are most similar to continental forms, 166 with birds in Tobago potentially moderately distinctive in having a more uniform (bleached) white 167 chest and belly, which in continental forms is often more greyish, cream or buff-colored with a pale 168 russet to rose-colored wash on the flanks. The four more remote islands of the Lesser Antilles show a 169 “leap-frog” pattern of plumage variation (sensu Remsen 1984). Birds on the most southern and most 170 northern of the islands (Grenada and Dominica) are most similar and are much darker overall 171 compared to continental forms, with a more uniformly rufescent color all over and therefore lacking 172 the grey-white underbelly common to continental forms. And the two middle island forms (St Vincent 173 and St Lucia) are also most alike and differ from both Grenada and Dominica and from continental 174 forms in showing a more bounded pattern of white ventrum and pale cheeks (auriculars) combined 175 with chestnut-to-russet colored wings and back, and a comparatively heavy and conspicuous white 176 eye-stripe. The latter eye-stripe is present but variable and generally less prominently developed in 177 the other island populations as well as in most continental forms. Hence, while the four more isolated 178 island populations of the Lesser Antilles recently elevated to species status are genuinely distinctively 179 patterned relative to continental forms, the pattern of plumage variation among them does not 180 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 9 suggest four distinctly different color morphs in support of four different species, nor clear clinal 181 variation across them, but rather two clusters with a disjunct distribution. Ultimately, while plumage 182 patterns can often be a useful taxonomic character, they are also famously complicated, inherently 183 subject to environmental variation, and therefore often differing within species with broad 184 distributions (see below). 185 Finally, there is well-documented morphological variation in the island populations of the Lesser 186 Antilles (Sosa-López and Mennill 2014; Wetten 2021). In a particularly comprehensive recent study, 187 Wetten (2021) confirmed that the island populations of house wrens are larger in most standard 188 measures of body and beak size compared to mainland populations, and that these measures also 189 differ among some, but not all, of the core islands of the Lesser Antilles (not including Trinidad or 190 Tobago). However, the pattern of larger body and beak size was not unique to the island populations 191 of the Lesser Antilles but applied also to the island populations in Mexico. It was thus interpreted to 192 be an instance of the broader phenomenon of insular gigantism, where small-bodied species tend to 193 get larger on islands, compared to mainland counterparts, as a function of reduced species diversity 194 and attendant ecological release and the adoption of a more generalist niche (Grant 1965; Cox and 195 Ricklefs 1977; Clegg and Owens 2002). 196 Hence, as with plumage, morphology is an inherently plastic trait subject to considerable 197 environmental variation, even within species; and both characters can thus facilitate but sometimes 198 also complicate species identification. The complications are well illustrated by the example of the 199 North American Fox Sparrow (Passarella iliaca) which parallels closely the situation for House Wren. 200 This taxon, widely distributed across North America, has also been a focus of taxonomic debate. It is 201 currently considered a single species, with many subspecies, organized into four broad types (AOU 202 1998). These four types were once regarded as different species based on variation in body size, beak 203 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 10 size and shape, and plumage patterns (Bailey 1902; Swarth 1920, Linsdale 1928), although there is 204 also considerable variation in these traits within each of the types. More recent mtDNA analyses have 205 reported genetic structuring between the four types but with non-trivial amounts of introgression 206 (hybridization) between them, and the morphs that are most similar in size and plumage are not 207 those most similar genetically, indicating that morphology and plumage in this taxon are 208 environmentally plastic and changing independently of gene flow (Zink 1994; Zink and Blackwell 1996; 209 Zink & Weckstein 2003; Zink 2008; for additional details and review see Weckstein et al. 2020). Avian 210 taxonomy includes many similar examples where genetic, size and plumage characteristics do not 211 align neatly (e.g., Yellow-rumped warbler - Setophaga coronata: see Hunt & Flashpohler 2020; 212 Northern Flicker – Colaptes auratus: see Wiebe & Moore 2024) and where there are therefore no 213 easy taxonomic solutions, only the opportunity that additional data might contribute further insight. 214 To help address these taxonomic and evolutionary issues for the House Wren group and 215 contribute to the broader enterprise for other taxa, I provide here a first comprehensive 216 characterization and quantitative analysis of song patterns across all of the islands in the Lesser 217 Antilles where house wrens remain. To ground the island comparisons, I include several continental 218 populations in the analysis and compare the complete set to comprehensive descriptions of song 219 reported previously for both northern and southern House Wren (e.g., Kroodsma 1977; Tubaro 1990; 220 Rendall and Kaluthota 2013; dos Santos et al. 2016; DiSciullo et al. 2023). 221

222 Populations Studied 223 House wrens in the Lesser Antilles were studied twice over a ten-year period, first in 2015 and 224 again in 2024. On both occasions, birds were sampled and recorded on all of the core islands where 225 they remain, which are from north-to-south and in order of increasing proximity to the mainland: 226 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 11 Dominica, St Lucia, St Vincent and Grenada. Trinidad and Tobago, which are closest to Grenada and a 227 bit further south, were also deliberately included in this study because they too are part of the Lesser 228 Antilles island chain and because they are also comparatively close to the south American mainland, 229 with correspondingly higher avian species diversity, confirming greater historical and likely also 230 continuing connectivity (i.e., gene flow) with continental populations. Hence, the islands of Trinidad 231 and Tobago represent a natural immediate outgroup from which to root comparisons of song 232 patterns for the other core islands of the Lesser Antilles. Birds on Trinidad and Tobago are also not 233 currently regarded as distinct species but rather as subspecies of the primary south American form (T. 234 musculus): the birds in Trinidad belong to the same subspecies as the closest mainland form (T. m. 235 clarus); while the birds in Tobago are classified as their own subspecies (T. m. tobagensis). 236 To further root the analysis of island song patterns with continental out-group comparisons, I 237 include several mainland populations from Ecuador, studied in 2022, that are grouped in the same 238 subspecies (T. m. albicans/clarus) as birds from Trinidad. I also include several populations from Belize 239 and Costa Rica (hereafter referred to as BCR), studied in 2018 and 2016 respectively, representing 240 additional subspecies of the south American continental form (T. m. intermedius and T. m. inquietus; 241 see Table 1 and Figure 1 for sample locations). In addition to helping to ground the comparisons of 242 the song patterns of the various island populations, this set of mainland comparisons also facilitates a 243 test of the intriguing possibility raised by the genetic analysis of Klicka et al. (2023) that there might 244 be a different historical source for colonization of the different islands in the Lesser Antilles. 245 Specifically, Klicka et al. (2023) hypothesized that birds on Dominica might have originated from 246 Central America while those on the other islands had a south American origin. It is possible that the 247 patterns of similarity and difference in songs among the islands, and between the various islands and 248 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 12 the mainland populations, can furnish additional evidence with which to evaluate this hypothesis. 249 Table 1 provides further details on the sample populations and their specific locations. 250 Song Recording 251 House wrens are noted for their variable (complex) song patterns (Kroodsma 1977; Tubaro 252 1990; Rendall and Kaluthota 2013; dos Santos et al 2016). In this study, every effort was therefore 253 made to capture and represent that variability but also to standardize recordings to eliminate as 254 many sources of extraneous or confounding influence as possible. 255 To that end, all research and recording was done by the same individual (DR), using the same 256 equipment and recording parameters: a Sound Devices 702 (or 722) digital recorder and Telinga Twin 257 Science Pro6/8/9-MK2 microphones, with corresponding 24” parabola, using a digital sampling rate of 258 48kHz and 24 bit depth. Research was also conducted at the same time of year for all populations 259 (March – April), in spring, to control as best possible for seasonal and breeding stage variation in song 260 patterns. This is harder to ensure for resident, tropical populations compared to migratory 261 populations in the temperate zone, because tropical populations are often not as seasonally 262 synchronized in their breeding activity. However, previous research on tropical House Wrens in Costa 263 Rica (Young 1994) shows that for this species there is nevertheless a breeding peak in early spring 264 (March – May: Young 1994) similar to that in the north temperate zone. 265 I also focused on singing only during the dawn chorus to further control for variation related to 266 breeding stage but also to minimize diel variation in singing activity. Dawn is when singing is most 267 vigorous and sustained for many passerines, and singing often decreases considerably thereafter. 268 Previous focused research on the northern House Wren confirms this and shows the early morning 269 hours to be the time of day when male singing is at its peak of both performance and complexity and 270 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 13 also when the differences between the song patterns of paired and unpaired males are minimal 271 (Kaluthota et al. 2020). 272 Recordings at every location focused on obtaining a sample that balanced sampling different 273 males in the population but also capturing variation in singing activity within individual males which is 274 known for continental populations of house wrens to be substantial and biologically relevant (Johnson 275 and Kermott 1991; Johnson and Searcy 1996; Deslandes et al. 2014; dos Santos et al. 2018; DiSciullo 276 et al. 2024). 277 Song Sample 278 The song sample selected for detailed analysis in this study was compiled from the larger 279 database of recordings obtained from 92 days of field observation across 20 different sample 280 locations. Songs were selected from this larger database using a number of complementary criteria. 281 First, songs had to be of the highest possible quality lacking background noise and overlapping song or 282 vocalizations from other species. The focus of further selection was then on representing population 283 level variation as best possible by including songs from multiple different males from each location 284 but also multiple songs from individual males. The latter dimension is important because previous 285 work on continental populations of both northern and southern House Wren shows that individual 286 males produce a large repertoire of notes and syllables that they combine into even larger repertoires 287 of different song types but that their often repetitive singing style means that much of this variation is 288 hidden in the short term (Kroodsma 1977; Tubaro 1990; Rendall and Kaluthota 2013; dos Santos et al 289 2016); hence it it takes some time to capture the variation of which individual males are capable. To 290 reflect both dimensions of variation, the sample for each location attempted to balance the 291 representation of different males with contributions from individual males. At the same time, to avoid 292 over-representing any specific males, no male contributed more than five songs to the sample for any 293 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 14 location. The final and important selection criterion to properly represent population level song 294 variation was that every song included in the sample, whether from the same male or a different 295 male, had to be of a different song type, as commonly defined in song research and as implemented 296 in previous detailed characterizations of song for both northern and southern House Wren (Rendall 297 and Kaluthota 2013; dos Santos et al. 2016) as the specific sequence of different notes and syllables. 298 There was no point including multiple versions of the same song type; this would simply involve a 299 form of pseudoreplication that might bias the sample for some locations by under-representing the 300 range of variation. Hence, every song in the sample represents a different song type – a different 301 sequence of notes and syllables. The sample ultimately compiled for detailed analysis involved 293 302 songs from 85 different males. Table 1 provides a more detailed breakdown of the sample by location. 303 Song Analysis 304 Preview of song recordings indicated that songs of house wrens in the Lesser Antilles follow the 305 same basic organizational structure and patterns of delivery that has has been reported in previous 306 detailed studies of other house wren populations (Kroodsma 1977; Tubaro 1990; Rendall and 307 Kaluthota 2013; Sosa-López and Mennill 2014a, b, c; dos Santos et al. 2016). Hence, to facilitate 308 comparisons between studies and among all populations, the analysis of songs studied here followed 309 the same general approach and methods used previously, and involved two different stages. 310 The first stage involved qualitative characterization of the relatively high-level patterns of song 311 structure, organization and delivery, including how songs are constructed from constituent notes and 312 syllables (defined as regularly occurring combinations of 2 or more notes); how they are organized 313 into different song types (defined as a different sequence of the individual notes and syllables); how 314 singing is patterned into bouts; and how frequently successive songs in a bout involve repeating or 315 changing song types. 316 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 15 The second stage focused on the more detailed structure of individual songs and involved 317 quantification of 21 different temporal and spectral features encompassing both the Introduction and 318 Main sections of songs (see Table 2 and Figure 2). Measured song parameters were: the overall 319 duration of each song (SONG-Duration) and the duration of its Introduction (INTRO-Duration) and 320 Main sections (MAIN-Duration); the number of elements in the entire song (SONG-#Elements) and in 321 the Introduction and Main sections (INTRO-#Elements, MAIN-#Elements); the average duration of the 322 elements and the intervals between them in each section (INTRO-ElementDuration; INTRO-323 IntervalDuration; MAIN-ElementDuration; MAIN-IntervalDuration); the rate of element production 324 across the entire song (ElemProdRate); the center frequency of elements in the Introduction and 325 Main sections (INTRO-CenterFreq, MAIN-CenterFreq); the average signal-to-noise ratio of elements in 326 the Introduction (INTRO-SNR); the average entropy of elements in the Introduction and Main sections 327 (INTRO-AvgEntropy, MAIN-AvgEntropy); the average bandwidth of elements in the Main section 328 (MAIN-BW-Element); the mean fundamental frequency (MeanF0) and extent of frequency modulation 329 of each element in the Main section (MeanFM/s); the overall duty cycle of each song, defined as the 330 proportion of each song allocated to active song (sound production) versus silent gaps (SONG-331 DutyCycle); and the proportion of each song represented by the Main section (Main-%Duration). 332 Measurements were made using both Raven Pro (V1.6.1, Cornell Lab of Ornithology 2019) and 333 PRAAT (V6.1.4, Boersma & Weenink 2021) software using the following spectrogram parameters: 334 Hann Window, with 3db (135 Hz) filter bandwidth, 512 point DFT, and 90.2% overlap in successive 335 spectral frames (or a 50 point, or 1.04 ms, “hop” size). 336 Statistical Analysis 337 The first phase of song analysis was qualitative in nature and therefore did not involve 338 application of statistical procedures. For the data in the second, quantitative phase, multivariate 339 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 16 discriminant analysis was used to test the extent to which the detailed features of songs were 340 discriminably different among the various islands of the Lesser Antilles, as well as from mainland 341 forms, in alignment with the recent taxonomic reclassification of the island populations as distinct 342 species. Discriminant analysis is notably sensitive to the number of predictor variables used to 343 discriminate among groups, where a high ratio of predictor variables to groups can produce unstable 344 solutions that ‘over-fit’ the data and thereby inflate discriminability (Klecka 1980, Tabachnick & 345 Fiddell 2007). To avoid this problem, a Principle Components Analysis (PCA) was first conducted on 346 the complete set of 21 measured acoustic features, with no rotation of the factor solution. This 347 precautionary step allowed examination of natural covariation among the original acoustic features 348 according to their associations within and between PCA factors. It also importantly reduced the 349 number of predictor variables to be used in discriminant analysis to a much smaller set of orthogonal 350 multivariate dimensions that nevertheless retained most of the variation in the entire set of original 351 acoustic features. The practical success of the discriminant solution was evaluated using a jacknife, 352 leave-one-out classification procedure which proceeds serially to classify each case in turn based on 353 functions derived from all of the other cases except the one that is currently being classified (the one 354 left out). It is a fairly liberal cross-validation technique but avoids the circularity of including in the 355 derivation of the classification functions the very cases that are then to be classified from those 356 functions. PCA and discriminant analysis were implemented in both SPSS (version 31.0) and NCSS 357 (2023). 358

359 General Patterns of Song Structure, Organization and Delivery 360 In general patterns of structure, organization and delivery, male song is similar across all 361 sampled populations and follows closely patterns described previously for continental forms of 362 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 17 northern and southern House Wren. Thus, it is a fast-paced jumble of variably structured notes that is 363 nevertheless predictably organized in two distinct sections: an Introductory section of broadband 364 notes that can be harsh and unstructured, or more tonal with harmonic overtones or a noisy overlay, 365 many resembling or replicating common call notes produced in other contexts; and a terminal or 366 Main section comprising much more distinctive and clearly structured tonal and frequency-modulated 367 notes and syllables. Introductory elements are often repeated multiple times, creating a 'stuttering' 368 effect, and there is often, but not always, a distinctive and emphatic ‘buzzy’ note at the junction 369 between the Introduction and Main sections of each song (see Figure 2). The terminal or Main section 370 of the song is typically considerably louder than the Introduction and can be heard over much longer 371 distances. 372 While most songs conform to this standard Introduction-Main template, males in all populations 373 also show a capacity to deviate from it. Thus, songs occasionally omit one or other section, or involve 374 one or more introductory-type notes appended to the end of the Main section of a song, or an entire 375 second Main section appended following a very short gap only marginally longer than the typical 376 interval between syllables within a song (what might be labelled a double-Main song). Songs can also 377 entail a protracted concatentation of two or three complete Introductory-Main sections with virtually 378 no gap demarcating individual songs, what might be labelled song ‘doublets’ or ‘triplets’. The capacity 379 for such variety is present in all populations, but its frequency of use appears to vary among some of 380 them. 381 Males in all populations also have a large repertoire of Introductory and Main notes and 382 syllables that they recombine to produce a much larger repertoire of different song types. In bouts of 383 singing, males typically repeat the same song type many times before switching to a different type, 384 although they commonly vary the number of repetitions of particular syllabe types within each song 385 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 18 from one rendition to the next. And when they switch song types, they tend to do so incrementally by 386 the addition, deletion, substitution, or movement of only one or two syllable types at a time. The 387 common pattern of song delivery then is one of ‘eventual variety’ where the diversity of a male’s 388 repertoire of different song types is displayed only gradually (sensu Kroodsma 1977). However, males 389 in all populations also show a capacity for more wholesale changes in successive song types and for 390 more rapid switching amomg them, thereby sometimes manifesting a pattern of ‘immediate variety’ 391 in their singing (see Figure 3 for example; see also Figure 1 in Juárez et al. 2025). The latter delivery 392 style is more common when males are close to and courting a female, or responding to song from 393 other males. 394 The diel pattern of singing is also similar across all sampled populations and follows a pattern 395 typical of tropical species generally. It involves a relatively short period of vigorous singing at dawn, 396 lasting 10-30 minutes, sometimes preceded by a few tentative songs (of lower amplitude, and 397 relatively short duration with longer intervals between songs) in the hour pre-dawn. Bouts of singing 398 at dawn typically involve high rates of song production (8-12 songs/minute). For the remainder of the 399 day, males produce short bouts of 2-4 songs sporadically as they move about their territory in the 400 course of routine activities. The exception is that males can sing far more vigorously and continously 401 throughout the day when actively courting a female. 402 While broadly similar in the above-noted general patterns of song organization and delivery, 403 there is nevertheless some variation in their manifestation among populations. Thus songs of males in 404 St Vincent appear to adhere much more strictly to the basic Introduction-Main template and manifest 405 fewer deviations from it, and appear aslo to involve an exaggerated level of song type repetition: for 406 example, one unpaired male was observed (and recorded) to repeat the same song type 102 times in 407 succession over a 13-min interval, and to produce only four slightly different song types over the 408 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 19 course of 214 successive songs; on another occasion, he repeated the same song type 79 times before 409 switching. In contrast, the organization of song by males in St Lucia appears to be more flexible with 410 less delineation between, and more fluid mixing among, Introductory and Main type notes and 411 syllables. Males in St Lucia also sometimes sing protracted and highly variable sequences of 412 Introductory and Main type notes and syllables with no detectable gaps demarcating individual songs 413 in a pattern of continuous delivery reminiscient of some members of the Mimidae, such as the north 414 American Catbird (Dumetellus). 415 There is also variation in the form and diversity of notes and syllables among some of the 416 populations. Male song in St Vincent is notably distinctive in this respect: introductory elements 417 include noisy broadband notes as elsewhere but a greater prevelance of higher-frequency notes that 418 often entail two distinct tones that are harmonically unrelated to each other, which gives the 419 introductory section of many songs a uniquely shrill and raspy, or even discordant, quality. Notes in 420 the Main section of the song in St Vincent are also notably distinct in being limited in the number and 421 variety of forms they take (see example in Figure 4). These include a few virtually pure-tone whistles, 422 which are not common in other populations, and a limited variety of only subtly different narrow-423 band, chevron and inverted chevron-shaped notes. The latter are typically produced in pairs with 424 modest frequency shifts, or offsets, between notes in a pair (e.g., ^ v; ^ ^ ; v v ) which creates a highly 425 distinctive alternating high-low, sing-song quality ("see-soo, see-soo") that may motivate the latin 426 name for the species (musicus). At the same time, there are few broad-band frequency sweeps or 427 protracted trills in the Main section of song in St Vincent which are relatively common features of 428 song in the other populations. The latter broad-band frequency sweeps are especially common and 429 elaborated in the Main section of songs in Dominica which tend to be longer and to involve a 430 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 20 preponderance of such sweeps, many spanning 5 kHz or more, organized in extended trills (see 431 example in Figure 5). 432 Detailed Song Structure 433 Mean values for each of the 21 detailed song features are provided for each population in Table 434 2. Inspection of these values confirms some of the differences among populations just noted: for 435 example, that songs in St Vincent have notably shorter, simpler Main sections, comprised of very few 436 elements that also have the lowest levels of frequency modulation (FM/s); while songs in Dominica 437 are notably longer than all of the other populations, with the longest Main section comprised of 438 elements with comparatively high levels of frequency modulation. 439 Principle Components Analysis on this set of 21 acoustic features produced a smaller set of 440 seven orthogonal factors, or components, with eigenvalues greater than one that cumulatively 441 accounted for and retained 83.9% of the variation in the complete dataset (Table 3). Fully two-thirds 442 of this variation was accounted for by the first three components. The association of original variables 443 with the different components is detailed in Table 4. 444

of Discriminant Analysis using components from PCA are provided in Table 5. The 445 analysis produced seven discriminant functions (one for each component from PCA), all but the last of 446 which contributed some degree of discrimination among the populations. However, the bulk of the 447 discrimination was carried by the first two functions (DF1 and DF2) which each had eigenvalues > 1.0, 448 markedly higher Canonical Correlations and lower Wilk’s Lambda scores than the other functions, and 449 together accounted for 84.3% of the variation in all of the components from PCA. DF3 and DF4 450 provided notably less, but not trivial, levels of discrimination. The associations of individual 451 components from PCA with each discriminant function are shown in the structure matrix in Table 6. 452 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 21 Discriminant scores for each song are shown in bivariate plots for the first three discriminant 453 functions in Figure 6. The first panel, plotting DF1 against DF2, encompasses almost 85% of the 454 variation in the data and shows some separation of BCR, Dominica, and St Vincent and to a much 455 lesser extent Grenada, with the remaining populations clustered centrally with highly overlapping or 456 virtually coincident distributions: Ecuador at the center and Trinidad and Tobago immediately 457 adjacent to that and co-located. The second panel, plotting DF1 against DF3, shows that DF3 helps to 458 separate St Lucia from the central nexus. It also highlights a higher-level pattern of population 459 clustering that groups BCR, Dominica and St Lucia together and separately from a clustering of the 460 remaining populations. The third panel (DF2 vs DF3) shows that there is little further separation of 461 populations gained from additional functions. 462 Classification success from discriminant analysis using all seven functions is shown in Table 7 463 which confirms and more clearly quantifies the patterns of overlap versus separation among the 464 populations observed in the discriminant score plots. Overall, 65.9% of songs were correctly classified, 465 which represents a significant degree of overall discrimination among populations and a very 466 considerable improvement on random (chance) classification which would be approximately 12.5% 467 for this sample. There was also considerable variation in levels of classification success for the 468 different populations. Of the island populations, St Vincent (85.7% correct) and Dominica (80.0%) 469 were correctly classified at the highest levels, followed by St Lucia (65.6%) and Grenada (60.0%), while 470 Tobago (33.3%) and Trinidad (25.8%) showed the lowest levels of successful classification. Both 471 mainland populations were classified at relatively high levels (BCR: 77.1%; Ecuador 79.7%). 472 Table 7 also tabulates a “Certainty Own” statistic which repesents the probability that a song 473 from a given population will be assigned to itself, as opposed to a different population, averaged 474 across all songs for that population. The statistic is similar but not equivalent to % correct because any 475 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 22 given song can be assigned correctly (or incorrectly) with varying certainty based on the relative 476 probabilities that it could belong to one or more of the other populations. So the statistic shows, for 477 example, that songs from Dominica and St. Vincent are not only correctly classified at high levels (80-478 85%) but also with comparatively high certainty (>75%), indicating that there is little possibility that 479 they belong to another population, which further underscores the relative distinctiveness of these 480 two populations from all the others. In contrast, songs from Grenada are classified moderately well 481 (60.0%) but with relatively low certainty (47%) confirming their broader similarity to other 482 populations. 483 Finally, Table 7 also shows that misclassified cases were not randomly distributed among the 484 populations. The small number of cases for Dominica that were misclassified were not assigned to any 485 of the other island populations but rather only to either BCR or Ecuador. A similar bias held for St 486 Lucia with more misclassified cases assigned to either BCR or Dominica and only one or two assigned 487 to any of the other populations. The remaining populations showed a reciprocal bias with 488 misclassified cases from Grenada, Tobago and Trinidad all being assigned either to one another or to 489 Ecuador but not to St Lucia, Dominica or BCR; and misclassified cases from Ecuador were assigned 490 primarily to Trinidad and Tobago. St Vincent had very few misclassified cases, but they were also all 491 assigned to either Grenada, Tobago, Trinidad or Ecuador and not to St Lucia, Dominica or BCR. 492 These patterns in misclassification are summarized in the final rows of Table 7, which tally the 493 number and percentage of all misclassified cases that were assigned to each of the different 494 populations. This serves as another measure of the relative distinctiveness of each population by 495 capturing how likely a given population is to be mistaken for other populations. These tallies further 496 confirm the distinctiveness of Dominica, St Lucia and St Vincent: not only were these three 497 populations successfully classified at the highest levels, they were also not often mistaken for any of 498 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 23 the other populations. By comparison Grenada, Tobago and Trinidad were more often mistaken for 499 another population in addition to being successfully classified at lower levels. 500 The overall patterns of similarity and difference among the populations are distilled in the final 501 panel in Figure 6 in the form of a heat map and clustered dendrogram. The heat map plots the 502 centroid values for each population on each of the seven discriminant functions represented by the 503 variable grey-scale shading of each box. Boxes that are more similarly shaded have similar centroid 504 values (i.e., are closer to each other in multivariate space) and the grey-scale patterns here simply 505 replicate what is graphically evident also in the discriminant score plots in the preceding three panels. 506 The clustered dendrogram distills the overall results pattern by grouping populations based on the 507 cumulative similarity or difference between their centroid values across all seven functions. The 508 dendrogram groups BCR, St Lucia and Dominica together and separates them from the other 509 populations as a distinct cluster. It then groups Tobago with Trinidad, followed by Grenada, and then 510 Ecuador, with St Vincent as the final and more distant member of the second cluster. 511

512 Summary Findings 513 In their general structure, organization and delivery the song patterns of island populations of 514 House Wren in the Lesser Antilles are all very similar and similar also to continental forms of northern 515 and southern House Wren (Kroodsma 1977; Tubaro 1990; Rendall and Kaluthota 2013; Sosa-López 516 and Mennill 2014a, b; dos Santos et al. 2016; DiSciullo et al. 2023) and notably also to some of the 517 other island populations in Mexico whose song has been comprehensively described (Sosa-López and 518 Mennill 2014c). Thus, they all show the same basic organizational template, with songs structured in 519 two discrete sections, an Introduction section of lower amplitude often noisy elements, and a Main 520 section of louder, tonal and frequency modulated elements; they all show the same pattern of song 521 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 24 type delivery where bouts of songs involve serial repetition of the same song type many times before 522 switching and then successive song types tend to differ only minimally by the addition, deletion, 523 substitution or movement of a single syllable type at a time; and they all also show a similar capacity 524 to deviate from both of these patterns. Thus, songs sometimes omit one or other of the Introduction 525 or Main sections; at other times they involve concatenating a series of them together into song 526 ‘doublets’ or ‘triplets’. Similarly, sequences of songs are often very repetitious, but song sequences at 527 other times can be far more variable and involve rapid switching between different song types with 528 more wholesale changes in the note and syllable contents of successive songs. The latter sorts of 529 deviation from the basic structural and organizational templates may be developed and used to a 530 greater or lesser extent in some of the populations, but in these higher-level characteristics, the song 531 patterns of all populations are generally more similar than they are different. 532 There are however diagnosable differences in the more detailed features of individual songs. 533 Thus, songs in Dominica are markedly longer in overall duration, with the longest Main section, the 534 longest intervals between elements, and thus also one of the slowest rates of element production. 535 They also entail a preponderance of protracted broad-band frequency sweeps that yield the highest 536 average center frequency and markedly higher averages for the fundamental frequency (F0) and 537 bandwidth of elements in the Main section. Collectively, the impression is of a kind of relaxed, or 538 largo, delivery style with songs produced at a slower tempo with widely spaced, broadband trilled 539 elements. These particular findings agree well with the song analysis by Sosa-López and Mennill 540 (2014a). Although Dominica was the only island population from the Lesser Antilles that was 541 represented in that broad-scale study of house wrens, it was described to have the most distinctive 542 song with specific distinguishing characteristics very much in agreement with those identified here. 543 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 25 Songs from St Vincent are also notably distinctive. They have the highest average center 544 frequency for the Introduction section and the shortest Main section, with the fewest number of 545 elements of markedly lowest bandwidth, frequency modulation, and average entropy (the latter 546 measure capturing the relatively pure tone nature of elements in the Main section). Subjectively, the 547 impression is of a relatively high frequency and often shrill, or even discordant, introductory section 548 paired with a short simple Main section with comparatively few steretoytped, narrowband elements 549 that are slightly offset in frequency creating a high-low, sing-song quality. 550 Songs from StLucia are – subjectively – more typically House Wren but may be discriminably 551 different in having a relatively short Introduction section with fewer and more tonal elements (with 552 lowest average entropy and highest signal-to-noise ratio) combined with a proportionally longer Main 553 section containing more elements. The impression is virtually the opposite of that for Dominica, 554 entailing a more rapid fire – staccato – delivery style with a relatively long and complex Main section. 555 Songs from Grenada are potentially distinctive in having the shortest Introduction section with 556 the fewest number of elements, contributing to the shortest overall song duration where the Main 557 section represents a greater proportion of the song relative to other populations. Otherwise, they 558 bear the most resemblance to canonical House Wren song, being intermediate in almost every 559 dimension and most similar to Tobago, Trinidad and Ecuador. 560 Songs from Tobago and Trinidad are not exceptional on any of the measured acoustic features, 561 although Main section elements have comparatively low bandwidth, and the two populations are 562 similar to each other in this and most other features. 563 Taken together, the picture that emerges for song of the various island populations of the 564 Lesser Antilles is of a broadly conserved and over-arching organizational template common to all 565 house wrens, with population-specific differences arising in the more detailed features of song 566 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 26 structure, where Dominica and St Vincent are most distinctive, followed by St Lucia and Grenada, with 567 Trinidad and Tobago minimally distinctive from mainland populations. 568 Taxonomic Implications: How Many Species? 569 Overall, these findings provide some important empirical support for the recent taxonomic 570 reclassification that elevated to species status the four most distant islands of the Lesser Antilles, 571 namely Dominica, St Lucia, St Vincent and Grenada (Chesser et al. 2024). In discriminant analysis of 572 detailed song features, these four populations were discriminated at levels significantly above chance 573 and substantially higher than levels of success for either Trinidad or Tobago. Of course, statistical 574 discrimination is not by itself definitive proof of species differentiation. Nevertheless, it does have an 575 important bearing on the issue because, as reviewed in the Introduction, each of the major forms of 576 data available previously and used in support of the taxonomic splits were either fragmentary or 577 incomplete in their sampling of the different island populations, or equivocal in the patterns manifest. 578 In such situations, some weight in taxonomic exercises is often attached to song given its role in mate 579 recognition and choice (Remsen 2005). Hence, the findings here, based on a comprehensive sampling 580 and analysis of song across all of the islands, represents an important contribution to the taxonomy of 581 this group and provides support for most of the splits. The cases are strongest for Dominica and St 582 Vincent whose songs are most distinctive, and to a somewhat lesser extent also for St Lucia. The case 583 for species status – at least based on song – is arguably weakest for Grenada which was successfully 584 classified at the lowest levels of the four islands in question (though considerably better than either 585 Tobago or Trinidad) and was also more frequently mistaken for one of the other populations. 586 Importantly also, the discriminant analysis specifically confirmed the similarity of songs between 587 Trinidad and Tobago, and the similiarity of both to songs from mainland South America, as 588 represented by populations in Ecuador. This similarity in song patterns among all three populations 589 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 27 may be evidence of some continuing contact (and gene flow) among them which would not be 590 unexpected given the proximity of Trinidad and Tobago to each other and to mainland South America. 591 It is therefore consistent with, and further supports, the recent taxonomic exercise which identifies 592 them all as subspecies of the mainland south American form (T. musculus; Chesser et al. 2024) 593 Evolutionary Implications 594 An additional novel result of this study, with important evolutionary implications, is the 595 summary finding from discriminant analysis of two distinct clusters of island populations: one cluster 596 that groups Dominica and St Lucia (the two most northern islands in the chain) as distinct from the 597 others and linked most closely to BCR as the most similar mainland population; and another cluster 598 that groups Tobago, Trinidad and Grenada (the three most southern islands) linked most closely to 599 Ecuador, with St Vincent (in the middle of the island chain) an outlier to this second cluster (Figure 6). 600 This pattern provides some support for, and also extends, the intriguing but tentative hypothesis from 601 the genetic study by Klicka et al. (2023) that there might be different geographic sources for the 602 various islands in the lesser Antilles, specifically that birds in Dominica might originate from Central 603 America, while birds on the other islands likely have a south American origin. This hypothesis was 604 tentative given the limitations of mtDNA data (reviewed in Winker 2021) and the absence of 605 complementary nuclear DNA results for the Lesser Antilles populations; and the possibility of a central 606 American source was suggested only for Dominica because Klicka et al. (2023) did not have any 607 genetic samples for St Lucia. So the results here for song data are productive in that they align with 608 the broader point of the Klicka et al. hypothesis that the island populations of the Lesser Antilles 609 might have different continental sources, and the results here specifically extend that proposal to 610 include St Lucia in those with a possible central American origin. 611 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 28 This is important because, if true, the link of both Dominica and St Lucia – at the north end of 612 the island chain – to a mainland source from Central America, distinct from the other islands with a 613 likely source from South America, makes all the more interesting the patterns of distinctive variation 614 in other traits across the islands. For example, the most noticeable (and commonly noted) pattern of 615 island variation is in plumage, where the four islands recently raised to species status are all different 616 in plumage from mainland forms but themselves actually sort into two pairs of very similarly colored 617 morphs: one pair links Dominica and Grenada which are at the northern and southern extremes of 618 this group of four islands; and the other pair links St Lucia and St Vincent which are in the middle (see 619 Figure 1 and the Introduction section for more details). Hence, if Dominica and St Lucia (at the north 620 end) share a central American source, while St Vincent and Grenada (at the south end) share a south 621 American source, then the leap-frog pattern of plumage similarity and difference among them would 622 require two independent instances of convergent plumage divergence: in other words, two separate 623 instances where the divergence in plumage within each of the two pairs resulted in convergence on 624 very similar plumage patterns across them. It is not immediately obvious what set of factors might 625 naturally explain such convergent divergence but it does not map naturally onto the song patterns 626 among them either: Dominica and Grenada are similar in plumage but not in song; St Lucia and St 627 Vincent are likewise similar in plumage but not in song; while Grenada, Trinidad and Tobago are most 628 similar in song but not in plumage. 629 Female Song 630 Adding to this already interesting character mix is the pattern of female song across the islands, 631 in so far it is known, which is very poorly. Although not a focus of study, my own observations (and 632 other informal reports) confirm the presence of female song on all of the islands except St Lucia and 633 St Vincent. This is curious because female song is common in tropical songbird species generally, and 634 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 29 it is also known, though possibly variably used, in continental populations of both northern and 635 southern House Wren (Johnson and Kermott 1990; Krieg and Getty 2016; Keck et al. 2025; reviewed 636 in Fernandez et al. 2024; Johnson 2024). Where it occurs among house wrens in the Lesser Antilles, 637 female song is similar to continental counterparts and often takes the form of a short sequence of 638 relatively harsh unstructured notes, similar to some introductory notes of male song and some 639 common non-song vocalizations. It can, however, also include more tonal, frequency modulated 640 notes similar to some of the more structured elements of the Main section of male song (for an 641 example from Dominica, see Figure 1 in Johnson et al. 2025). And female song can at times be highly 642 synchronized to the male’s song, appended immediately to the end of his song almost as an 643 accentuation or echo of it; at other times it overlaps male song and can be difficult to distinguish from 644 it; and yet other times it can lead or occur completely independently of male song. 645 This is a very cursory and preliminary characterization of female song, unfortunately, and the 646 matter deserves much more detailed study because, while not historically a focus of research to the 647 same extent as male song, very recent comprehensive comparative studies and evolutionary path 648 analysis have shown that female song is both more widespread across songbird taxa than previously 649 appreciated and likely also represents the ancestral state for many of them (Odom et al. 2014; see 650 also Odom and Benedict 2018). This work further shows that, across tropical and temperate zone 651 taxa, the absence of female song likely reflects evolutionary loss associated with a shift away from 652 year-round territoriality and biparental care (Odom et al. 2025). Hence, if female song is truly absent 653 in the Lesser Antilles islands of St Lucia and St Vincent, that have putatively different continental 654 sources, it would not only represent an additional instance of convergent evolutionary divergence – 655 the independent loss of song in two island populations with different original sources – but it might 656 also point to associated evolutionary changes in other core behavioral and life-history traits. Further, 657 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 30 if effective mate choice and pairing involves coordinating song contributions between prospective 658 male and female partners, then the absence of female song in St Lucia and St Vincent could be an 659 important taxonomic character as an effective premating isolating mechanism. 660 These latter possibilities remain speculative but there currently appear to be multiple possible 661 instances of character divergence and convergence among plumage, male song and possibly also 662 female song (and potentially other behavioral and life-history traits) across the islands of the Lesser 663 Antilles. Collectively, this intriguing mix of characters therefore stands as a strong impetus to further 664 detailed study to bring additional light to the evolutionary relationships among these island 665 populations of house wrens and to the historical processes that have yielded the trait mosaic they 666 now manifest. 667 Summary and Future Research Directions 668 1. In general structure, organization and delivery the song of males on all of the islands of the 669 Lesser Antilles is similar and follows patterns common to house wrens broadly, but songs of the 670 different islands are discriminably different in their more detailed features. 671 2.The latter differences in song among the island populations are consistent with most of the 672 recently recognized species splits, with the possible exception of Grenada where song is the least 673 distinctive and comparatively closely aligned with that of Trinidad, Tobago and Ecuador. Birds in 674 Grenada are distinct from continental forms (and Trinidad and Tobago) in plumage but this difference 675 is of uncertain classification value given the environmental plasticity of plumage (and size) generally 676 and the fact that birds in Grenada are most similar in plumage to birds at the other end of the island 677 chain in Dominica which genetic and song analysis both link to a different geographic route of 678 incursion. 679 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 31 3.Additional genetic research, especially including nuclear DNA, would help to resolve the 680 relationships among island populations and clarify their potential central versus south American 681 origins, which would in turn bring additional clarity to the character mosaic manifest among them 682 that currently implies multiple independent instances of evolutionary convergence and divergence. 683 4.Female song should be a particular focus of future research to confirm or revise its presence 684 on each of the islands and to more fully characterize its features and possible functions, including 685 potential coordination with male song. 686 LITERATURE CITED 687 Alström, Per & Ranft, Richard. (2003). The use of sounds in avian systematics and the importance of 688 bird sound archives. Bull. B.O.C.. 123. 689 American Ornithologist’s Union (1983). Check-list of North American Birds, 6th ed. Am, Ornithol. 690 Union, Washington, D.C. 691 American Ornithologist’s Union (1998). Check-list of North American Birds, 7th ed. Am, Ornithol. 692 Union, Washington, D.C. 693 Bailey, F.M. (1902). Handbook of Birds of the Western United States. Houghton, Mifflin and Co., 694 Cambridge, Massachusetts. 695 Barlow, J.C. (1978). Another colony of the Guadeloupe house wren. The Wilson Bulletin 90 (4): 6356-696 637 697 Boersma, P. , and D. Weenink (2021). PRAAT Software, version 6.1.40 698 Brewer, D. (2001). Wrens, Dippers and Thrashers. Yale University Press. 699 Brumfield, R. T., and A. P. Capparella. (1996). Genetic differentiation and taxonomy in the house wren 700 species group. Condor 98:547-556. 701 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 32 Center for Conservation Bioacoustics (2019). Raven Pro: Interactive Sound Analysis Software 702 (Version 1.6.1) Ithaca, NY: The Cornell Lab of Ornithology 703 Chesser, R. T., S. M. Billerman, K. J. Burns, C. Cicero, J. L. Dunn, B. E. Hernández-Baños, R. A. Jiménez, 704 O. Johnson, A. W. Kratter, N. A. Mason, P. C. Rasmussen, and J. V. Remsen, Jr. (2024). Sixty-fifth 705 Supplement to the American Ornithological Society’s Check-list of North American Birds. 706 Ornithology 141: 1–20. https://doi.org/10.1093/ornithology/ukae019 707 Clegg, S. M., and I. P. F. Owens (2002). The ‘island rule’ in birds: Medium body size and its ecological 708 explanation. Proceedings of the Royal Society of London B 269: 1359-1365.53 709 Cox, G. W. and R. E. Ricklefs (1977). Species diversity and ecological release in Caribbean land bird 710 faunas. Oikos 28(2): 113-122. 711 Cyr, M.E., K. Wetten, M.H. Warrington, and N. Koper (2021). Variation in song structure of house 712 wrens living in urban and rural areas in a Caribbean small island developing state. Bioacoustics 30 713 (5): 594-607. DOI: 10.1080/09524622.2020.1835538 714 Deslandes, V., L. R. R. Faria, M. E. Borges and M. R. Pie. (2014). The structure of an avian syllable 715 syntax network. Behavioural Processes 106:53-59. 716 DiSciullo, R.A., S. K. Sakaluk, and C. F. Thompson (2023). Song structure of male Northern House 717 Wrens and patterns of song production and delivery across the nesting cycle. J Ornithology 718 https://doi.org/10.1007/s10336-023-02098-0 719 DiSciullo, R.A., A.M. Forsman, R. R. Fitak, J. Hunt, P. Nietlisbach, C. F. Thompson, and S. K. Sakaluk 720 (2024). Male song structure predicts offspring recruitment to the breeding population in a 721 migratory bird. Evolution 78(6): 1054-1066. 722 Dickinson, E. C., and L. Christidis, Editors (2014). The Howard and Moore Complete Checklist of the 723 Birds of the World. Volume 2. 4th edition. Aves Press, Eastbourne, UK. ISBN 978-0-9568611-2-2. 724 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 33 Fernández, G.J., M.E. Carro, and L. S. Johnson (2024). Southern House Wren (Troglodytes musculus), 725 version 1.0. In Birds of the World (R. Juárez, B. K. Keeney, and S. M. Billerman, Editors). Cornell Lab 726 of Ornithology, Ithaca, NY, USA. https://doi.org/10.2173/bow.houwre4.01 727 Gilardi, J. D., and C. L. John (1998). Conservation of the St. Lucia House Wren Troglodytes aedon 728 mesoleucus: Distribution, abundance and breeding biology. Dodo, Jersey Wildlife Preservation Trust 729 34:91–102. 730 Grant, P. R. (1965). The adaptive significance of some size trends in island birds. Evolution 19(3): 355-731 367. 732 Grant, B. R., and P. R. Grant (1996). Cultural inheritance of song and its role in the evolution of 733 Darwin’s finches. Evolution 50(6): 2471-2487 734 Huber, S., and J. Podos (2006). Beak morphology and song features covery in a population of Darwin’s 735 finshes (Geospiza fortis). Biological Journal of the Linnean Society 88: 489-498. 736 Hunt, P. D., and D. J. Flaspohler (2020). Yellow-rumped Warbler (Setophaga coronata), version 1.0. In 737 Birds of the World (P. G. Rodewald, Editor). Cornell Lab of Ornithology, Ithaca, NY, 738 USA. https://doi.org/10.2173/bow.yerwar.01 739 Irwin, D.E. (2000). Song variation in an avian ring species. Evolution 54: 998–1010. 740 Johnson, L. S. (2024). Northern House Wren (Troglodytes aedon), version 1.1. In Birds of the World (B. 741 K. Keeney, A. F. Poole, M. G. Smith, and S. M. Billerman, Editors). Cornell Lab of Ornithology, 742 Ithaca, NY, USA. https://doi.org/10.2173/bow.houwre.01.1 743 Johnson, L. S., and L. H. Kermott (1990). The structure and context of female song in a north-744 temperate population of House Wrens. Journal of Field Ornithology 61:273–284. 745 Johnson, L. S., and L. H. Kermott (1991). The functions of song in male House Wrens (Troglodytes 746 aedon). Behaviour 116: 190–209. 747 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 34 Johnson, L. S., and W. A. Searcy. (1996). Female attraction to male song in House Wrens (Troglodytes 748 aedon). Behaviour 133:357–366. 749 Johnson, L. S., R. Juárez, and D. Rendall (2025). Kalinago Wren (Troglodytes martinicensis), version 1.1. 750 In Birds of the World (B. K. Keeney, S. M. Billerman, J. Gerbracht, and A. J. Spencer, Editors). Cornell 751 Lab of Ornithology, Ithaca, NY, USA https://doi.org/10.2173/bow.cinwre1.01.1 752 Juárez, R., L. S. Johnson, and D. Rendall (2025). St. Lucia Wren (Troglodytes mesoleucus), version 1.1. 753 In Birds of the World (B. K. Keeney, S. M. Billerman, and A. J. Spencer, Editors). Cornell Lab of 754 Ornithology, Ithaca, NY, USA. https://doi.org/10.2173/bow.houwre8.01.1 755 Kaluthota, C., B. E. Brinkman, E. B. dos Santos, and D. Rendall (2016). Transcontinental latitudinal 756 variation in song performance and complexity in house wrens (Troglodytes aedon). Proceedings of 757 the Royal Society of London B. 283:20152765. 758 Kaluthota, C.D, Logue, D.M., and Rendall, D. (2020). Conventional and network analyses of song 759 organization and complexity in northern House Wrens (Troglodytes aedon parkmanii). Journal of 760 Field Ornithology. https://doi.org/10.1111/jofo.12347 761 Keck, SM. A. Barnhill , S. Q. Smeele, M. H. Hsu, A. Dolson-Fazio, C. Bergler, G. F. Ball , C. A. Krieg and K. 762 J. Odom (2025) Deep machine learning allows classification of male and female songbird songs in a 763 species with highly variable repertoires Behaviour DOI: 10.1163/1568539X-bja10324 764 Klecka, W.R. (1980). Discriminant Analysis Series: Quantitative Applications in the Social Sciences 765 (Sage, London). 766 Klicka, J., K. Epperly, B. T. Smith, G. M. Spellman, J. A. Chaves, P. Escalante, C. C. Witt, R. Canales-del-767 Castillo, and R. M. Zink (2023). Lineage diversity in a widely distributed New World passerine bird, 768 the House Wren. Ornithology 140:1–13. https://doi.org/10.1093/ornithology/ukad018 769 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 35 Krieg, C.A, and T. Getty (2016). Not just for males: females use song against male and female rivals in 770 a temperate zone songbird. Animal Behaviour 113: 39-47 771 Kroodsma, D.E. (1977). Correlates of song organization among North American wrens. American 772 Naturalist 109: 995-1008. 773 Kroodsma, D. E., and D. Brewer (2005). Family Troglodytidae (wrens). In Handbook of the Birds of the 774 World. Volume 10: Cuckoo Shrikes to Thrushes (J. del Hoyo, A. Elliott, and D. A. Christie, Editors). 775 Lynx Edicions, Barcelona, Spain. pp. 356–447. 776 Linsdale, J. M. (1928). Variations in the Fox Sparrow (Passerella iliaca) with reference to natural 777 history and osteology. University of California Publications in Zoology 30:251-392. 778 Valderrama, S., J. Parra, and D.J. Mennill (2007). Species differences in the songs of the critically 779 endangered Niceforo’s Wren and the related Rufous-and-white Wren. Condor 109(4): 871–878. 780 NCSS 2023 Statistical Software (2023). NCSS, LLC. Kaysville, Utah, USA, ncss.com/software/ncss 781 Oberholser, H. C. (1904). A review of the wrens of the genus Troglodytes. Proceedings of the United 782 States National Museum 27:197–210. 783 Odom, K.J., and L. Benedict (2018). A call to document female bird songs: applications for diverse 784 fields. Auk 135: 314-325. DOI:10.5751 785 Odom, K.J., Hall, M.L., Riebel, K., Omland, K.E., and Langmore, N.E. (2014). Female song is widespread 786 and ancestral in songbirds. Nature Communications 5: 3379. DOI:10.1038/ncomms4379. 787 Odom, KJ, M Araya-Salas, L Benedict, K Lim, J Dale, WH Webb, C Sheard, J.A. Tobias, G. F. Ball, M.L. 788 Hall, N.E. Langmore, M.S. Webster,and K. Riebel (2025). Global incidence of female birdsong is 789 predicted by territoriality and biparental care in songbirds Nature Communications 16 (1), 6157 790 Podos, J., J.A. Southall, and M.R. Rossi-Santos (2004). Vocal mechanics in Darwin's finches: correlation 791 of beak gape and song frequency Journal Experimental Biology (2004) 207 (4): 607–619. 792 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 36 Podos, J. (2010). Acoustics discrimination of sympatric morphs in Darwin’s finches: a behavioral 793 mechanism for assorative mating? Phil Trans Royal Society, B 365: 1031-1039. 794 Remsen, J.V. (1984). High incidence of" leapfrog" pattern of geographic variation in Andean birds: 795 implications for the speciation process. Science, 224 (4645): 171-173. 796 Remsen, J.V. (2005). Pattern, process and rigor meet classification. The Auk, 122(2): 403-413 797 Rendall, D., and C. D. Kaluthota. (2013). Song organization and variability in Northern House Wrens 798 (Troglodytes aedon parkmanii) in western Canada. Auk 130 (4):617-628. 799 dos Santos, E.B., P.E. Llambías, and D. Rendall (2016). The structure and organization of song in 800 Southern House Wrens (Troglodytes aedon chilensis). Journal of Ornithology. 157(1): 289–301. 801 dos Santos, E.B., Llambías, P., and D. Rendall (2018). Male song diversity and its relation to breeding 802 success in Southern House Wrens (Troglodytes aedon chilensis). Journal of Avian Biology DOI: 803 10.1111/jav.01606 804 Sosa-López, J. R., and D. J. Mennill (2014a). Continent-wide patterns of divergence in acoustic and 805 morphological traits in the House Wren species complex. The Auk 131(1):41–54. 806 Sosa-López, J. R., and D. J. Mennill (2014b). The vocal behavior of the Brown-Throated Wren 807 (Troglodytes brunneicollis): song structure, repertoires, sharing, syntax, and diel variation. J 808 Ornithology 115: 435-446. 809 Sosa-López, J. R., and D. J. Mennill (2014c). Vocal behavior of the island-endemic Cozumel Wren 810 (Troglodytes aedon beani): song structure, repertoires, and song sharing. J Ornithology 115: 337-811 346. 812 Sosa-López, J. R., J. E. Martínez Gómez, and D. J. Mennill (2016). Divergence in mating signals 813 correlates with genetic distance and behavioural responses to playback. Journal of Evolutionary 814 Biology 29:306–318. https://doi.org/10.1111/jeb.12782 815 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 37 Sukumaran, J., and L. Knowles (2021). Multispecies coalescent delimits structure, not species. Proc 816 Nat Acad Sci 114: 1607-1612. www.pnas.org/cgi/doi/10.1073/pnas.1607921114 817 Swarth, H. W. (1920). Revision of the avian genus Passerella with special reference to the distribution 818 and migration of the races in California. University of California Publications in Zoology 21:75-224. 819 Tabachnick, B.G., and L.S. Fiddell (2007). Using Multivariate Statistics, 5th edition. Pearson: New York. 820 Toews, D.P.L, and D.E. Irwin (2008). Cryptic speciation in a Holarctic passerine revealed by genetic and 821 bioacoustic analyses. Molecular Ecology 17: 2691–2705. 822 Tubaro, P. L. (1990). Song description of the House Wren (Troglodytes aedon) in two populations of 823 eastern Argentina, and some indirect evidences of imitative vocal learning. Hornero 13:111-16. 824 Weckstein, J. D., D. E. Kroodsma, and R. C. Faucett (2020). Fox Sparrow (Passerella iliaca), version 1.0. 825 In Birds of the World (A. F. Poole and F. B. Gill, Editors). Cornell Lab of Ornithology, Ithaca, NY, 826 USA. https://doi.org/10.2173/bow.foxspa.01 827 Wetten, K. N. (2021). Morphological divergence in the House Wren (Troglodytes aedon) species 828 complex: A study of island populations with a focus on the Grenada House Wren (T. a. grenadensis). 829 M.Sc. thesis, University of Manitoba, Canada. 830 Wiebe, K. L., and W. S. Moore (2024). Northern Flicker (Colaptes auratus), version 2.1. In Birds of the 831 World (P. G. Rodewald and B. K. Keeney, Editors). Cornell Lab of Ornithology, Ithaca, NY, 832 USA. https://doi.org/10.2173/bow.norfli.02.1 833 Winker, K. (2021). An overview of speciation and species limints in birds. Ornithology 138: 1-27. DOI: 834 10.1093/ornithology/ukab006 835 Young, B.E. (1994). The effects of food, nest predation and weather on the timing of breeding in 836 tropical house wrens. Condor 96: 341-353 837 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 38 Zink, R. M. (1994). The geography of mitochondrial DNA variation, population structure, hybridization, 838 and species limits in the Fox Sparrow (Passerella iliaca). Evolution 48:96–111. 839 Zink, R.M. (2008). Microsatellite and mitochondrial DNA variation in the Fox Sparrow. The Condor 110 840 (3): 482-492. 841 Zink, R. M., and R. C. Blackwell (1996). Patterns of allozyme, mitochondrial DNA, and morphometric 842 variation in four sparrow genera. The Auk 113(1):59–67. 843 Zink, R. M., and J. D. Weckstein (2003). Recent evolutionary history of the Fox Sparrows (Genus: 844 Passerella). The Auk 120(2):522–527. 845 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 1. Summary of song sample. Population Species/subspecies Dates Males Songs Latitude Longitude Elevation (m) Belize-Belmopan T. m. intermedius Mar 7 - Mar 9, 2018 6 8 17.22 -88.85 72 CostaRica-LaSuiza T. m. intermedius Apr 10 - Apr 14, 2016 2 8 9.85 -83.61 960 CostaRica-Orosi T. m. intermedius Apr 5 - Apr 9, 2016 7 16 9.78 -83.84 1,250 CostaRica-LasCruces T. m. inquietus Apr 15 - Apr 19, 2016 4 16 8.78 -82.95 1,218 Dominica-Rosalie Dominica-Portsmouth T. martinicensis Apr 21 - Apr 26, 2015 Apr 5 - Apr 8, 2024 6 30 15.37 -61.26 29 StLucia-GrosPiton StLucia-VillaPiton T. mesoleucus Apr 16 - Apr 20, 2015 Mar 31 - Apr 4, 2024 8 32 13.81 -61.06 293 StVincent-Buccament Valley StVincent-MtVernon T. musicus Apr 11 - Apr 15, 2015 Mar 25 - Mar 30, 2024 6 28 13.19 -61.24 206 Grenada-StGeorge’s Grenada-StDavids T. grenadensis Apr 6 - Apr 10, 2015 Mar 21 - Mar 24, 2024 9 30 12.10 -61.74 256 Tobago-Castara Tobago-Castara T. m. tobagensis Apr 1 - Apr 5, 2015 Mar 17 - Mar 20, 2024 8 30 11.28 -60.69 95 Trinidad-Maravel Valley Trinidad-Arima Valley T. m. clarus Apr 27 - Apr 30, 2015 Mar 13 – Mar 16, 2024 10 31 10.69 -61.29 245 Ecuador-SantoDomingo T. m. albicans Apr 7 - Apr 8, 2022 2 10 0.27 -79.14 611 Ecuador-Maquipacuna T. m. albicans Mar 23 – Mar 25, 2022 3 8 0.13 -78.63 1,367 Ecuador-Mindo T. m. albicans Mar 26 – Mar 28, 2022 Apr 8 – Apr 10, 2022 5 22 0.03 -78.80 1,141 Ecuador-Milpes T. m. albicans Mar 29 – Mar 30, 2022 2 8 0.03 -78.87 1,150 Ecuador-CopaLinga T. m. clarus Apr 1 – Apr 4, 2022 7 16 -4.09 -78.96 1,017 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 2. Means for song variables by population. SONG VARIABLES BCR Dominica StLucia StVincent Grenada Tobago Trinidad Ecuador Overall INTRO-#Elements 10.38 10.67 6.25 7.61 5.70 8.30 9.35 8.73 8.53 INTRO-Duration (s) 1.44 1.76 0.94 1.32 0.85 1.15 1.30 1.29 1.27 INTRO-ElementDuration (s) 0.06 0.09 0.08 0.13 0.10 0.09 0.09 0.10 0.09 INTRO-IntervalDuration (s) 0.08 0.08 0.07 0.05 0.05 0.06 0.05 0.06 0.06 INTRO-CenterFreq (Hz) 4,212 5,211 4,533 5,280 4,096 4,350 4,423 5,241 4,701 INTRO-AvgEntropy 3.97 3.77 3.57 3.91 3.85 4.09 4.00 3.82 3.87 INTRO-SNR 27.36 26.64 27.86 23.95 23.08 23.48 23.80 24.45 25.17 MAIN-#Elements 12.38 10.97 12.34 5.86 9.07 8.00 10.52 10.83 10.28 MAIN-Duration (s) 1.23 1.48 1.36 0.92 1.19 1.03 1.23 1.17 1.20 MAIN-ElementDuration (s) 0.08 0.11 0.08 0.13 0.11 0.10 0.09 0.09 0.10 MAIN-IntervalDuration (s) 0.03 0.05 0.04 0.03 0.03 0.03 0.03 0.03 0.03 MAIN-CenterFreq (Hz) 4,068 4,848 4,222 3,700 3,682 3,916 3,753 4,259 4,083 MAIN-AvgEntropy 2.95 2.76 2.70 1.84 2.44 2.24 2.41 2.56 2.53 MAIN-BW-Element (Hz) 3,988 4,166 3,516 2,255 3,141 2,870 3,092 3,473 3,380 SONG-#Elements 22.75 21.63 18.59 13.46 14.77 16.30 19.87 19.56 18.82 SONG-Duration (s) 2.68 3.24 2.30 2.24 2.05 2.18 2.52 2.46 2.47 ElemProdRate (elements/s) 8.59 6.70 8.18 6.02 7.21 7.47 7.89 7.90 7.63 SONG-DutyCycle 0.59 0.63 0.63 0.76 0.76 0.69 0.72 0.70 0.68 MAIN-%Duration 0.46 0.46 0.58 0.41 0.58 0.48 0.49 0.47 0.49 MeanFM/s 60,337 45,729 38,672 20,016 34,531 32,565 36,332 50,986 42,554 MeanF0 3,476 4,444 3,864 3,594 3,424 3,622 3,441 3,789 3,703 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 3. Summary of principle components analysis. Component Eigenvalue % of Variance Cumulative % 1 5.584 26.591 26.591 2 3.465 16.501 43.092 3 2.693 12.823 55.914 4 1.958 9.324 65.238 5 1.458 6.943 72.182 6 1.280 6.096 78.277 7 1.186 5.650 83.927 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 4. Variable loadings on each component from principle components analysis. Acoustic variables Component* 1 2 3 4 5 6 7 INTRO-#Elements 0.436 0.609 -0.555 0.166 -0.011 0.074 -0.246 INTRO-Duration 0.318 0.770 -0.427 0.283 0.015 0.014 0.106 INTRO-ElementDuration -0.647 0.142 0.135 0.141 -0.097 0.263 0.394 INTRO-IntervalDuration 0.417 0.031 0.309 -0.006 0.176 -0.571 0.509 INTRO-CenterFreq -0.157 0.292 0.197 -0.002 -0.207 0.276 0.570 INTRO-AvgEntropy 0.013 -0.070 -0.131 0.080 0.772 0.085 0.202 INTRO-SNR 0.212 0.263 0.116 -0.286 -0.624 -0.337 -0.055 MAIN-#Elements 0.737 -0.523 0.125 0.280 -0.147 0.137 0.149 MAIN-Duration 0.489 -0.325 0.512 0.597 -0.045 0.095 -0.074 MAIN-ElementDuration -0.637 0.425 0.288 0.228 0.240 -0.069 -0.264 MAIN-IntervalDuration -0.112 0.466 0.566 0.299 -0.023 -0.270 -0.145 MAIN-CenterFreq 0.342 0.457 0.505 -0.336 -0.048 0.342 -0.071 MAIN-AvgEntropy 0.744 0.050 0.220 -0.273 0.306 0.080 -0.034 MAIN-BW-Element 0.365 0.357 0.526 -0.255 0.250 0.043 -0.349 SONG-#Elements 0.876 -0.039 -0.250 0.333 -0.127 0.158 -0.035 SONG-Duration 0.609 0.402 0.013 0.657 -0.019 0.079 0.035 ElemProdRate 0.620 -0.509 -0.364 -0.286 -0.153 0.118 -0.116 SONG-DutyCycle -0.756 -0.160 -0.034 0.135 -0.018 0.505 -0.122 MAIN-%Duration 0.082 -0.739 0.605 0.174 -0.016 0.043 -0.130 MeanFM/s 0.695 0.082 -0.007 -0.429 0.257 0.176 0.162 MeanF0 0.199 0.450 0.442 -0.238 -0.210 0.343 0.107 *Bolded values indicate for each acoustic variable the component with which they are most strongly associated. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 5. Summary results of discriminant analysis on principle component scores. Function Eigenvalue % Variance Cumulative % Variance Canonical Correlation Wilk’s Lambda F- Value P-value DF1 3.346 63.1 63.1 0.877 0.052 22.9 <0.001 DF2 1.123 21.2 84.3 0.727 0.226 13.8 <0.001 DF3 0.385 7.3 91.5 0.527 0.480 9.1 <0.001 DF4 0.266 5.0 96.5 0.458 0.664 7.7 <0.001 DF5 0.143 2.7 99.2 0.353 0.841 5.6 <0.001 DF6 0.039 0.7 100.0 0.194 0.961 2.9 0.0226 DF7 0.002 0.0 100.0 0.041 0.998 0.5 0.4871 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 6. Structure matrix from discriminant analysis showing the correlations of individual principle components with each discriminant function. Component* Discriminant Function 1 2 3 4 5 6 7 PCA1 0.627 0.028 0.405 0.446 0.199 -0.375 0.252 PCA2 0.086 0.564 0.319 -0.359 -0.007 0.210 -0.632 PCA3 0.117 0.281 -0.614 0.084 0.411 0.522 0.285 PCA4 -0.108 0.236 0.061 -0.374 0.431 -0.614 0.476 PCA5 -0.021 -0.215 0.514 -0.248 0.363 0.624 0.327 PCA6 -0.256 0.250 0.302 0.821 0.322 0.026 -0.052 PCA7 -0.022 0.277 0.138 0.047 -0.678 0.174 0.641 *Bolded values indicate the component most highly correlated with each discriminant function. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Table 7. Classification results from discriminant analysis Population N BCR Dominica StLucia StVincent Grenada Tobago Trinidad Ecuador % correct Certainty Own* BCR 48 37 2 4 0 1 2 0 2 77.1% 0.72 Dominica 30 2 24 0 0 0 0 0 4 80.0% 0.76 StLucia 32 4 2 21 0 1 1 1 2 65.6% 0.62 StVincent 28 0 0 0 24 1 1 1 1 85.7% 0.77 Grenada 30 0 0 1 3 18 2 2 4 60.0% 0.47 Tobago 30 1 0 0 1 6 10 5 7 33.3% 0.33 Trinidad 31 0 0 0 2 6 4 8 11 25.8% 0.31 Ecuador 64 2 0 1 1 1 4 4 51 79.7% 0.63 Totals: 293 65.9% #Misclassified to: 6 4 6 7 16 14 13 31 %Misclassified to: 6.7% 4.3% 6.7% 7.3% 18.2% 17.5% 16.9% 35.6% * Certainty own repesents the probability that a song from a given population will be assigned to itself, as opposed to a different population, averaged across all songs for that population. Average values for each population express the confidence of correct self-assignment. Misclassified totals and % values for each population are based on the total of the misclassified cases that could possibly be assigned to them, which therefore excludes their own misclassifed songs that were assigned to other populations. It is an additional measure of distinctiveness that captures how likely a given population is to be mistaken for other populations. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 39 Figure Legends 842 Figure 1: Map of study locations with spectrograms of representative songs and photographs of 843 plumage patterns for each population. Note the ‘leap-frog’ pattern of plumage similarity and 844 difference for the four core islands of the Lesser Antilles that were recently reclassified as separate 845 species, where plumage is similar between Dominica and Grenada at the northern and southern 846 extremes of this group, and also similar between St Lucia and St Vincent in the middle. 847 Figure 2: Spectrogram of standard house wren song variant. This example is from Grenada and 848 illustrates the standard structural organization of song common to all populations involving two 849 distinct sections: an Introductory section of primarily noisy broad-band notes, or more tonal notes 850 with a noisy overlay; and a Main section of more distinctive, tonal and frequency modulated notes 851 and syllables. Many, but not all, songs include a conspicuous buzzy note (o) at or near the 852 transition between the two sections. Some of the other measured features of song are also 853 labelled. 854 Figure 3: A short sequence of four songs from a longer bout of singing from St Lucia. This short 855 sequence illustrates several notable features of song organization and song type variation. The 856 differently sized and colored boxed segments are used to highlight a specific note or note 857 combination (syllable) to help illustrate the patterns. The typical pattern of song organization and 858 delivery in St Lucia and elsehere commonly entails repeating the same song type many times 859 before switching (eventual variety) and then changing song types only gradually by the addition, 860 deletion, substitution, or movement of one or two notes or syllables at a time. However, this 861 example illustrates the capacity for more immediate variety in, and some wholesale changes 862 between, different song types that can occur in all populations but may be elaborated to a greater 863 degree in St Lucia. Thus, S1 and S2 are completely different in both their Introduction and Main 864 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 40 sections. S3 uses the first three notes from the Introduction of S1 (blue box) and adds a buzzy 865 element (yellow box) to the Introduction and then introduces another different Main section with 866 a rising buzz note (grey box) at the end of the Main section; it then also appends Introductory 867 notes from S2 (orange and green boxes) to the end of the first canonical song after a very short 868 gap. S4 reverts to the Introductory notes and most of the Main section from S2, adds some new 869 buzz and short tick notes and the rising buzz note from S3 (grey box), and then appends to the end 870 of the song, after a long gap, the Introduction from S3 (blue and yellow boxes), the first part of 871 which (blue box) was also the Introduction to S1. In addition to thereby illustrating more 872 immediate variety in successive song types, this example also illustrates how songs (particularly in 873 St Lucia) sometimes involve greater continuous mixing of Introduction and Main type notes (S4) or 874 simply appending Introduction type notes to the end of the Main section of a song after a short 875 gap (as in S3 and S4). 876 Figure 4: A representative song from St Vincent illustrating some of the distinctive features of song 877 from this island population. These include higher frequency elements in the Introduction section 878 with harmonically unrelated tonal components (tall rectangular boxes) and a comparatively short 879 and simple Main section comprised of narrow band, chevron shaped notes (^v) often produced in 880 pairs and offset slightly in frequency producing a distinctive high-low (‘sing-song’) percept. 881 Figure 5: A representative song from Dominica illustrating the more widely spaced elements and 882 preponderance of very wide-band frequency sweeps in the Main section. 883 Figure 6. Bivariate plots of discriminant scores for the songs of each population on the first three 884 discriminant functions (a-c) and a heat map and clustered dendrogram (d) summarizing the 885 relationships among populations based on the multivariate distances between them. The first 886 panel (a) shows some separation of BCR, Dominica, St Vincent and to a lesser extent Grenada on 887 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint 41 the first two discriminant functions, while the second panel (b) shows that the third discriminant 888 function helps to separate St Lucia and it also highlights (with large elipses) the distinct clustering 889 of BCR, St Lucia and Dominica seaparately from Grenada, Trinidad, Tobago, and St Vincent. The 890 third panel (c) shows no further separation of the populations. The fourth panel (d) is a heat map 891 of population centroid values on each of the seven functions from discriminant analysis and a 892 clustered dendrogram that groups the populations based on similarity and difference in the 893 cumulative euclidean distance between their centroid values across all seven functions using the 894 group averaging clustering method. Values at the nodes of the dendrogram represent the average 895 distance between the branches they connect, expressed in units defined by the discriminant space: 896 the absolute values therefore have no wider meaning but they can be interepreted in relative 897 terms, such that the relative differences in the magnitude of the node values capture the relative 898 magnitude of the song differences between the populations. So, for example, the difference in 899 songs linking St Lucia and BCR is greater than the difference in songs that links the entire cluster of 900 populations of Grenada, Tobago, Trinidad, and Ecuador. 901 902 903 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Dominica St Lucia St Vincent Tobago Trinidad Grenada Belize Ecuador Costa Rica Figure 1 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Time (s) 0 2 20Frequency (kHz) MainIntroduction Center Frequency Element Bandwidth F0 Frequency Modulation Figure 2 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint S1 S2 S3 S4 S1 S2 S3 S4 Figure 3 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Frequency-shifted (^ v) note pairs Time (s)0 3 20Frequency (kHz) Figure 4 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Time (s)0 4 20Frequency (kHz) Figure 5 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Discriminant Function 1 Discriminant Function 2 BCR Dominica Ecuador Grenada St Lucia St Vincent Tobago Trinidad Figure 6a .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Discriminant Function 1 Discriminant Function 3 BCR Dominica Ecuador Grenada St Lucia St Vincent Tobago Trinidad Figure 6b .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Discriminant Function 2 Discriminant Function 3 BCR Dominica Ecuador Grenada St Lucia St Vincent Tobago Trinidad Figure 6c .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint Discriminant Function Centroid Values 1 2 3 4 5 6 7 Ecuador Trinidad Tobago Grenada Dominica St Lucia BCR St Vincent 2.5 3.2 3.7 4.4 5.75.9 6.9 Figure 6d .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 20, 2026. ; https://doi.org/10.64898/2026.01.15.699782doi: bioRxiv preprint