Reviewers of the complex of small, white-bellied glossy swiftlets, for which the oldest available name is Collocalia esculenta (Linnaeus, 1758), have given weight to certain phenotypic characters as taxonomic indicators distinguishing species or species groups. One character, first perceived by Wallace (1864), separates ‘plain-tailed’ and ‘spot-tailed’ species. The plain-tailed group occurs from the Andaman Islands, India, through Malaysia and Indonesia to Lombok. All members have glossy upperparts that are uniformly darkcoloured including the rump and tail. ‘Spot-tailed’ populations, from Sulawesi east through and beyond New Guinea, were believed by Wallace (1864), and thereafter by Stresemann (1940) and Somadikarta (1982, 1986), to be differentiated by a white spot on the concealed inner vane of all but the central pair of rectrices. In a review of speciation in the C. esculenta complex integrating phenotypic and molecular data, Rheindt et al. (2017) broadly confirmed Wallace's ideas, with some revision of species boundaries within these two groups.

Figure 1.

The minute feather tuft on the hallux of Collocalia affinis cyanoptila (Earl of Cranbrook)

Among the plain-tailed group of white-bellied swiftlets a second indicator is variation in the green or blue tone to the glossy upperparts. This character has been linked to the presence or absence of a single, small tufted feather on the dorsal side of the hind toe (hallux). Within the western plain-tailed white-bellied swiftlets, excluding those populations in the Philippines, two species groups have been separated by this mix of characters (Somadikarta 1982, 1986). One consists of the species C. a. affinis (sensu Rheindt et al. 2017) on the Andamans, Nicobar and other islands, and C. a. cyanoptila from Sumatra, Peninsular Malaysia and Borneo. Within this group, members of three geographically separated colonies of C. a. cyanoptila in Selangor, Peninsular Malaysia, exhibited individual age- and wear-related change, from the greenish gloss of fresh plumage to blue gloss of old, worn plumage. Although differently glossed, all individuals in the separate populations were genetically uniform (Lim 1994). Members of this group bear a single, very small tufted feather on the hallux (Fig. 1). In the live bird, this feather is moulted around mid-term during primary moult in the wings, and for a period may be totally absent from one foot or both or, if present, sheathed and inconspicuous (Cranbrook et al. 2005). It can also be difficult to confirm presence or absence of this feather on the hallux in specimens. The feet (often tied together) must be separated to allow inspection with a lens, and are often in poor condition due to fungal infestation or other issues of imperfect preservation.

The other group within the western plain-tailed population includes C. linchi of Java, Bali, Lombok and intervening small islands. These birds are characterised by permanently green-glossed upperparts including the tail, and being invariably ‘bare-toed’, i.e., lacking the feather tuft on the hallux. There is more than one instance of sympatry involving the C. linchi superspecies and C. affinis. In Sabah, Malaysia, the endemic C. dodgei, a member of the linchi superspecies (Rheindt et al. 2017), overlaps in daily activity range with the more numerous local population of C. affinis cyanoptila (Cranbrook et al. 2005, Moyle et al. 2008). On Sumatra, Somadikarta (1986) reported overlapping ranges of C. affinis cyanoptila and C. linchi ripleyi, with a mixed colony in a cave at Talangpadang, South Lampung. Separation at species level is confirmed by divergence in Cytb mtDNA sequences of 6.03–7.20% (Table 3 in Rheindt et al. 2017).

On the grounds of exhibiting a spotted tail, Christmas Island white-bellied swiftlet C. natalis was regarded by Stresemann (1940) and Somadikarta (1986) as a geographically anomalous member of the C. esculenta group, lying west of the main boundary of separation, named ‘Stresemann's Line’ by Somadikarta (1986) (Fig. 2). However, mitochondrial Cytb sequence divergence from nominate linchi, at 1.10–1.45%, is ‘shallow’ and this island endemic, which displays a morphology unlike any other Collocalia species, was considered a member of the linchi superspecies (Rheindt et al. 2017). Removal of C. natalis restores Wallace's concept of the Makassar Strait as a natural boundary between western plain-tailed and eastern spot-tailed white-bellied swiftlets. Molecular studies support divergence between the clades separated by this boundary at 4.66–8.59% in mtDNA, regarded as ‘deep’ by Rheindt et al. (2017).

Figure 2.

Locations where field observations and DNA material collection were undertaken, showing ‘Stresemann's Line’ as defined by Somadikarta (1986).

The possibility that plain-tailed white-bellied swiftlets might cross this boundary was raised by Mayr & Camras (1938); who noted a specimen of a plain-tailed, greenish-glossed, ‘apparently young’ bird resembling C. linchi of Lombok, among spot-tailed, blue-glossed C. esculenta manadensis on Sangihe (Sangir) Island, Sulawesi (also discussed by Salomonsen 1983: 31). S. Somadikarta (in Cranbrook et al. 2005) examined six additional specimens from Sangihe in the Bogor Zoological Museum, all of which had spotted tails. Rheindt et al. (2017) did not assess the implications of this single plain-tailed specimen east of Stresemann's Line, but did report a personal communication by Cranbrook, with photographs of four individuals at a single colony on Ternate, North Maluku province, Indonesia. These shared intensely blue upperparts and the lack of a feather tuft on the hallux, but two were plain-tailed and two spot-tailed. This contrasted with 24 specimens from North Maluku seen by Rheindt et al. (2017: 421), all with spotted tails, leading these authors to conclude that: ‘the new, unpublished findings from Ternate suggest that the morphological division across Stresemann's Line may not be as clear-cut as previously assumed’.

Here we present the results of collective efforts from different research employing both field observations and genetic studies, with the aim of testing the taxonomic significance of the phenotypic characters that are conventionally used in Collocalia identification.

Materials and Methods

Field observations.—Scientific nomenclature sensu Rheindt et al. (2017) is followed. Locations are given as coordinates, usually taken directly from GPS readings, and elevation as metres above sea level (m). Fig. 2 shows the locations mentioned in the text.

East of the Makassar Strait, in Papua New Guinea, on nine occasions between August 2000 and August 2005, MT, with students from the Pacific Adventist University, handled 275 C. esculenta nitens at the abandoned copper mines on Mt. Diamond, Central province (c.09°46.4362″S, 147°32.4446″E; 68 m). Smaller samples were taken at Losave Cave, near the Chimbu / Eastern Highlands province border (06°64.8333″S, 145°15.8889″E; 1,425 m); at a cave 5 km west of Kumul Lodge in Western Highlands province (05°77.9853″S, 143°95.9424″E; 2,614 m), and from one bird caught on the western peak of Mt. Hagen (05°76.4172″S, 144°02.1167″E; 3,670 m) (Table 2). From four nests of C. esculenta tametamele in the Japanese tunnels at Sonoma Adventist College, New Britain (c.04°42.7326″S, 152°23.9931″E; 171 m) one adult was caught and measured. At a cave near Lelet village (03°25.1350″S, 151°96.1743″E; 200 m) on New Ireland, 65 C. esculenta heinrothi were caught and examined. All of these birds were released after handling. Additionally, in 2016, four birds, identified on geographical grounds as C. esculenta spilura (see Coates & Bishop 2000) were caught using a mist-net at a nesting colony below a road bridge over a deep and precipitous river gully on the lower flank of Gunung Gamalama (07°97.182″S, 127°36.8507″E; 204 m) on Ternate, North Maluku province, Indonesia.

Plain-tailed swiftlets were sampled as follows: in 2015, one bird, taken from a small colony of Linchi Swiftlets C. l. linchi, at a swiftlet house-farm at Cacaban, Banten Residency, Java, Indonesia (06°10.8828″S, 106°00.1381″E; 61 m); and in 2016, five White-bellied Swiftlets C. affinis cyanoptila at a colony in the garage of Buona Vista (Stephen's Place), Fraser's Hill, Pahang, Peninsular Malaysia (03°71.5903″N, 101°75.0444″E; 1,292 m). As outgroup (for complete mitochondrial genome analysis), one Black-nest Swiftlet Aerodramus maximus was included from the Perhentian Islands, Terengganu, Peninsular Malaysia (05°96.3306″N, 102°68.3333″E; 3 m).

DNA materials.—Samples for genetic analysis were obtained by plucking one feather from each of the left and right wings in the primary tract; preferentially p3 or p4 (numbered centrifugally). Approximately 0.5 cm at the base of the rachis was cut off with fine scissors and immediately preserved in 70% or 90% ethanol. The birds were then released, apart from five, which were humanely killed by compression of the thorax; breast muscle was excised and immediately preserved in 90% ethanol.

Molecular procedures.—DNA was extracted from preserved material using standard procedures—HiYieldPlus DNA Mini Kit (Real Biotech Corporation) and Wizard Genomic DNA Purification Kit (Promega), following the manufacturers' instructions. Polymerase chain reactions (PCRs) were set up following the primers and conditions in Price et al. (2004) for the mitochondrial cytochrome-b (Cytb) and NADH dehydrogenase 2 (ND2), and Thomassen et al. (2005) for the nuclear beta-fibrinogen gene (Fib) region, which spanned the partial exon 8, complete intron 7 and partial exon 7 regions. Direct sequencing was performed commercially by FirstBase Laboratory Sdn. Bhd. Samples for next-generation sequencing were prepared using the Nextera DNA Sample Preparation Kit (Illumina, #FC-121-1031), according to the manufacturer's instructions. Next-generation sequencing was performed with a 4nM library on an Illumina MiSeq 600 bp v3 (2×300 bp) following standard procedures.

Data analysis.—We employed both multi-gene phylogenetic and phylogenomic analyses. The former was performed based on the mitochondrial Cytb and ND2 regions, and the nuclear Fib7 region; published DNA data from other congeners was incorporated into our analyses. Phylogenomic analysis, based on the mitochondrial genomes, was also performed for a subset of samples.

Phylogenetic analysis for Cytb+ND2 and Fib genes.—Twelve DNA sequences of C. affinis, C. esculenta and C. linchi published in previous studies were randomly retrieved from GenBank (Table 1) and added to our phylogenetic analyses based on Cytb and ND2. Two house-farm white-nest swiftlets, Aerodramus sp., and a white-nest swiftlet A. fuciphagus vestitus from a Middle Baram cave, Sarawak, Malaysia (Goh et al. 2018) were included as outgroups. MtDNA data from the other Collocalia taxa deposited by Price et al. (2004) were also included (Table 1). Analysis based on Fib7 employed two outgroup species, i.e., the house-farm swiftlets (161803i, 161703f, 151020f; Goh et al. 2018) which were newly sequenced for this study, and A. maximus M2–M5 (Thomassen et al. 2005). All sequences obtained for this study are deposited in GenBank (accession numbers: MH727218–226). All sequences were aligned using ClustalX2 v.2.1 (Thompson et al. 1997) and manually edited and trimmed in BioEdit (Hall 1999).

TABLE 1

Additional DNA sequences (Price et al. 2004 and Thomassen et al. 2005) included in the present study.

TABLE 2

Morphometric data for Collocalia esculenta nitens from Papua New Guinea.

Maximum Parsimony (MP) analysis was performed using PAUP4.0b10 (Swofford 2002). The strict consensus tree was reconstructed using heuristic search with 100 random-sequence additions, tree bisection reconnection (TBR) branch swapping and 1,000 bootstrap replications. Bootstrap support (BS) values >70% were considered reliable.

The best-fit models (GTR+G for mtDNA and HKY for nuclear DNA) for the Bayesian Inference (BI) were identified using MrModeltest2.2 (Nylander 2004). BI analyses were run in MrBayes3.1 (Huelsenbeck & Ronquist 2001), using two runs of four chains each, and run for 10,000,000 generations with trees sampled every 100 generations. The first 2,500 trees were discarded as burn-in. Posterior probabilities (PP) >0.90 were considered a strong support in this study.

Phylogenomic analysis of mitochondrial genomes.—Next-generation sequence data for phylogenomic analysis was generated using an Illumina MiSeq. A novel mitochondrial genome scaffold was constructed using Integrated Genome Viewer (IGV) v2.3.88 (Robinson et al. 2011, Thorvaldsdóttir et al. 2013) from the consensus sequences of five Collocalia affinis cyanoptila sampled at Fraser's Hill. The hyper-variable D-loop region (517 bp) was excluded from analysis, leaving a mitochondrial sequence length of 15,564 bp. For the four birds sampled on Ternate and five C. a. cyanoptila from Fraser's Hill, MiSeq data was quality-assessed and re-sequenced against this scaffold using the on-instrument Illumina MiSeq Reporter Software. Additionally, two genetic outgroups were included: one C. l. linchi sampled at Cacaban, Java, and one Aerodramus maximus from the Perhentian Islands, Malaysia. The A. maximus sample was re-sequenced against a novel mitochondrial scaffold constructed from house-farmed birds of the region. Re-sequenced mitochondrial genomes were manually curated in IGV to produce a consensus sequence for each bird. Sequences obtained for this study are deposited in GenBank [accession number(voucher)]: MT123507(bd108_162710a), MT123508(bd109_162710b), MT123509(bd110_162710c), MT123510(bd111_162710d), MT123511(bd112_162810a), MT921253(bd072_160803a), MT921254(bd073_160803b), MT921255(bd074_160803c), MT921256(bd075_160803d), MT921257(bd006_15200618), MT921258(bd097_162003a). Mitochondrial genome sequences for the 11 birds were aligned using MUSCLE (Edgar 2004). The best-fit model (GTR+G) for Maximum Likelihood phylogeny was selected and performed using MEGA7 (Tamura 2013), with 1,000 Bootstrap replications. Bootstrap values >70% were considered strong support. Estimates of genetic divergence (uncorrected p-distances) were computed in MEGA7.

Results

Tail spots.—East of Stresemann's Line (Fig. 2), among some 303 glossy swiftlets, C. esculenta subspp. examined, from three cave colonies and one mountain peak in mainland Papua New Guinea, and one cave on New Ireland (Table 2), three birds had no spots on the rectrices. Others had spots on one, two, three or four rectrices between rr2–5 (never on the central pair). Moreover, spots were variable in size, sometimes large, sometimes small (1 mm diameter) and not always equal or present on matching feathers on either side of the tail. One bird from Mt. Diamond had a very small spot on one feather on the right side, but none on any of the left rectrices. Another from the same colony had an extremely small spot on just r3. One bird from the Losave colony also had a very small spot on just one rectrix.

The small sample of four birds on Ternate were all mist-netted in flight and evidently adult. One had lost the right r4, but otherwise the flight feathers were entirely unmoulted, and fresh in appearance. Two conformed to the description in Coates & Bishop (2000) of C. esculenta spilura, being intensely glossed blue on the upperparts including the dorsal tail-coverts, with white spots on the inner webs of all but the central pair of rectrices. The other two (identical to each other) were similarly intensely blue-glossed above, including the dorsal tail-coverts and tail, but all rectrices were uniform black, glossed blue, with no indication of any white spot on the inner vanes (Figs. 3–4). The hind toes of all four birds lacked a small dorsal tuft.

Green to blue gloss transition.—Repeated sampling of C. esculenta nitens at the Mt. Diamond colonies revealed an age-related trend from greenish to blue gloss, as observed for C. affinis cyanoptila by Lim (1994). The greenish gloss of a newly moulted bird (Fig. 5) gives way after a few months to a mixture of green and bluish gloss (Fig. 6). Just prior to moult the dorsal plumage exhibits a deep blue surface, with much less gloss, as the feather transmits less light via refraction and reflects more (Fig. 7). Confirmation that these are the stages of change is observed when, at the same nest, a blue adult is perched alongside its young with a green gloss (Fig. 8).

Presence of feather tuft on the hallux.—At three colonies of C. esculenta in mainland Papua New Guinea and one on New Ireland, 21% of swiftlets examined had a feather tuft on the hind toe.

Molecular phylogenetic analyses: Cytb + ND2 and Fib7 dataset.—The Cytb + ND2 sequence data were aligned into a data matrix of 1,213 characters, of which 204 characters are parsimony-informative. For the Fib7 dataset, 32 of 932 characters are parsimony-informative. For each dataset, phylogenetic analyses using BI and MP methods resulted in largely consistent tree topologies, thus only BI topologies are presented here (Figs. 9–10).

In our phylogenetic tree based on the mitochondrial Cytb and ND2 data (Fig. 10), all five individuals of C. affinis cyanoptila sampled at Fraser's Hill form a clade (PP 1.00 / BS 100%) with the Collocalia individuals collected from Selangor, Sandakan and Lahad Datu (referred to C. esculenta cyanoptila by Price et al. 2004). C. marginata and C. isonata bagobo from the Philippines (named C. esculenta marginata and C. e. bagobo by Price et al. 2004) resolve into a cluster with high support (PP 1.00 / BS 95%). The two plain-tailed C. esculenta sampled on Ternate form a clade (PP 1.00 / BS 100%) with three other C. esculenta, including the two spot-tailed individuals sampled on Ternate by the present study.

Phylogenetic analyses based on the nuclear Fib7 marker (Fig. 10) again showed high affinity (PP 1.00 / BS 100%) between all four Collocalia individuals sampled on Ternate irrespective of tail pattern. They appeared genetically uniform and distant from the C. affinis cyanoptila sampled at Fraser's Hill (where only two representatives could be sequenced for the Fib7 marker).

Mitochondrial genomes.—Phylogenetic analyses based on ‘whole’ mitochondrial genomes (excluding the hypervariable D-loop region) of a subset of individuals are shown in Fig. 11. Exclusion of the D-loop did not significantly affect genetic divergence or phylogeny results, and resulted in a sequence length of 15,564 bp of mtDNA. The four Collocalia individuals sampled on Ternate form a genetically distinct clade with high support (BS 100%), and the five C. affinis cyanoptila individuals at Fraser's Hill form another clade (BS 100%), distinct from Ternate birds. These results corroborate the Cytb and ND2 sequence and nuclear Fib7 marker phylogenies (Figs. 9–10).

Genetic divergence across the mitochondrial genome between the Ternate C. esculenta spilura clade and the Fraser's Hill C. affinis cyanoptila clade was estimated at 4.68% (Table 3). Both clades are equally genetically distant from the C. l. linchi specimen sampled on Java, with 4.74–5.00% divergence. The genetic outgroup for this study, Aerodramus maximus, showed a deep genetic divergence of 9.43–9.86% from all Collocalia specimens tested.

Figure 3.

One of two plain-tailed Collocalia esculenta spilura, Mt. Gamalama road bridge, Ternate, Indonesia, August 2016 (Sian E. W. Davies)

Figure 4.

One of two spot-tailed Collocalia esculenta spilura, Mt. Gamalama road bridge, Ternate, Indonesia, August 2016 (Sian E. W. Davies)

Figure 5.

Adult Collocalia esculenta nitens displaying large tail spots and the green gloss of fresh plumage, Mt. Diamond, Papua New Guinea, October 2000 (Michael Tarburton)

Figure 6.

Another adult Collocalia esculenta nitens midway through moult, displaying a mix of blue and green, Mt. Diamond, Papua New Guinea, September 2003 (Michael Tarburton)

Figure 7.

Adult Collocalia esculenta nitens just prior to moulting, showing its deep blue plumage, Mt. Diamond, Papua New Guinea, August 2001 (Michael Tarburton)

Figure 8.

A blue-glossed adult of Collocalia esculenta nitens with old plumage (on left) perched alongside its young in fresh green plumage, Mt. Diamond, Papua New Guinea, September 1999 (Michael Tarburton)

Figure 9.

Phylogenetic tree based on the mitochondrial Cytb + ND2 regions (1,213 bp) reconstructed using Bayesian Inference. Numbers above the nodes indicate posterior probabilities >0.90. Bootstrap support values >70% in the maximum parsimony analysis were mapped on the Bayesian topology. Sequences retrieved from GenBank (GB) were published in Price et al. (2004). ‘*’ = Collocalia taxa from the Philippines not covered in this paper.

As would be expected from near full-length mitochondrial sequences, genetic differences were observed between individuals within both Collocalia clades, with divergence of 0.40% within the Ternate group and 0.07% in the Fraser's Hill group. There was high support (BS >70%) for the presence of multiple genetic subclades or maternal lineages within the Ternate group. However, these distinctions did not correlate with the presence or absence of tail spots.

Discussion

Tail spots.—Our observations reveal that, among the population of C. esculenta spilura on Ternate, some individuals (50% of the tiny sample of four) were entirely plain-tailed. Yet mtDNA sequencing confirms that these two birds were genetically very close to the two spot-tailed swiftlets at the same colony (genetic divergence of 0.40% among the four). Variation in the phenotypic expression of this character among a larger sample of C. esculenta in Papua New Guinea was very diverse, and included three with plain (unspotted) tails. i.e., c.1% of all birds examined. Although genetic information is not available for these Glossy Swiftlets, the observations are convincing evidence of phenotypic polymorphism among these populations.

TABLE 3

Genetic (nucleotide) divergence across mitochondrial genomes (total length 15,564 bp, excluding the D-loop region). (A) Genetic divergence between groups: percentage divergences (%) are shown in the lower left of the table, with corresponding standard deviations at the upper right. (B) Genetic divergence within groups. Percentages shown are estimates of evolutionary distance (p-distance) over nucleotide sequence pairs, with standard deviations.

Figure 10.

Phylogenetic tree based on the nuclear beta-fibrinogen intron 7 region (935 bp) reconstructed using Bayesian Inference. Numbers above the nodes indicate posterior probabilities >0.90. Bootstrap support values >70% in the maximum parsimony analysis were mapped on the Bayesian topology.

Figure 11.

Phylogenetic tree based on novel assembled ‘full-length’ (with D-loop excluded) mitochondrial genomes (15,564 bp) constructed using Maximum Likelihood with the GTR+G model. Bootstrap support values >70% are indicated above the nodes.

Green and blue gloss.—Between the species (and subspecies) of C. affinis and C. linchi, variation in the colour of dorsal gloss is supported by genomic evidence (Cibois et al. 2018: Fig. 9), and is therefore confirmed as a useful taxonomic indicator. On the other hand, the succession of images of swiftlets at the Mt. Diamond colonies of C. esculenta nitens shows age- and wear-related progress from greenish to blue gloss. It is evident that greenish or blue dorsal gloss cannot be considered a character of taxonomic significance among this member of the C. esculenta group.

Feather tuft on hallux.—The difficulty of confirming the presence or absence of the single feather tuft on the hallux in C. affinis cyanoptila, described by Cranbrook et al. (2005), cautions against generalisations as to the prevalence of this character. If this minute feather is shed about mid-term in the slow progression of the moult of the primaries, its absence in an individual should be checked against the moult state of that bird. Variation in New Guinea swiftlets raises doubts that this character is of general taxonomic significance in the C. esculenta complex.

Genetic relationships.—Phylogenetic analysis of the mitochondrial markers Cytb and ND2 revealed that the maternal genetic lineages of Collocalia may split along geographical lines. This is corroborated by phylogenies produced from the nuclear gene marker Fib7. Cytb and ND2 data placed birds sampled at Fraser's Hill within a super-clade of C. affinis cyanoptila, comprising two distinct subclades corresponding to Peninsular Malaysia and Bornean Malaysia (Fig. 9).

In both Cytb + ND2 and Fib7 datasets, the C. esculenta spilura sampled on Ternate formed a separate clade. This clade did not cluster with Collocalia from the Philippines, which were placed west of Stresemann's Line by Somadikarta (1986), but instead formed a clade with C. esculenta nitens from Papua New Guinea (Fig. 9), confirming that plumage variation in Ternate birds does not represent migration from western populations.

This suggestion was corroborated by analysis of mitochondrial genome sequences from a subset of individuals in this study, including the four Ternate C. e. spilura and five C. a. cyanoptila from Fraser's Hill, in addition to C. linchi and A. maximus outgroups (Table 3). This analysis enabled evaluation of genetic divergence across a larger region (15,564 bp vs. c.2,000 bp for ND2 + Cytb combined), thereby increasing confidence in the calculated divergence values: <0.5% within species, c.5% between species, and c.10% between genera from the sequenced mitochondrial genomes. These are broadly consistent with values from short (c.400 bp) mitochondrial fragments (Rheindt et al. 2017).

There was a very close genetic relationship between all four Ternate swiftlets, with a within-group evolutionary distance of 0.400%, comparable to that among Fraser's Hill birds (0.074%). The Ternate group was equally distant from C. l. linchi on Java (4.739%) and C. a. cyanoptila at Fraser's Hill (4.677%). This distance is equivalent to that between C. a. cyanoptila and C. l. linchi (5.001%), suggesting a species-level divergence of c.5% among Collocalia, and that these three species are equally distinct from each other. In addition, all three were equally distant from the Aerodramus maximus outgroup, each divergent by c.10%. These data imply that tail-spot variation within the Ternate population cannot be attributed to genetic input from another species.

Genetic analyses of mitochondrial and nuclear DNA reveal geographic separation of phylogenies, which do not correlate with the phenotypic characters traditionally used as taxonomic indicators, such as tail spots, green and blue gloss, and hallux feather tuft.

Conclusions

Our observations consolidate the deep divergence at 4.66–8.59% mtDNA, reported by Rheindt et al. (2017) between the clades of Collocalia spp., separated by the Makassar Strait, and further refine it to c.5% (4.677–5.001%) via comparison of long mitochondrial sequences (15,564 bp cf. c.400 bp in Rheindt et al. 2017). Our phylogenetic results support the separation by Rheindt et al. (2017) of the Collocalia esculenta complex into the ‘western linchi’, ‘western white-bellied swiftlet’ (C. affinis) and ‘eastern C. esculenta’ groups. It is, however, now clear that this phylogenetic separation is not expressed phenotypically, by the presence or absence of white spots on the concealed rectrices. The spot-tailed Christmas Island Swiftlet C. natalis is a member of the C. linchi group. Two plain-tailed birds on Ternate were genetically inseparable from two spot-tailed C. esculenta spilura in the same colony, and relatively distantly related to plain-tailed swiftlets C. affinis cyanoptila of Peninsular Malaysia and Borneo. C. esculenta in Papua New Guinea displayed great variation in the size, distribution and, rarely, absence (<1% of the sample) of concealed white spots on the rectrices, and 21% of the sample had a small feather tuft on the hallux. These comparisons support Somadikarta's (1986) contention that swiftlets of the C. esculenta complex cannot be divided into species by single morphological characters.

A comparatively small sample led Rheindt et al. (2017) to conclude that all populations of C. esculenta throughout mainland New Guinea have intensely blue-glossed dorsal plumage and clearly marked tail spots, and usually no feather tuft on the hind toe. Based on our observations of phenotypic variation in C. esculenta subspecies extending from North Maluku through Papua New Guinea to New Ireland, this conclusion needs to be amended, for the benefit of other students of white-bellied swiftlets and to prevent misleading statements in future regional avifaunas.

Genetic data presented by Cibois et al. (2018) from Western Pacific glossy swiftlets showed that the white-rumped taxon albidior of New Caledonia, rather than being a subspecies of Satin Swiftlet C. uropygialis of Vanuatu, was embedded within C. esculenta becki of the Solomon Islands, and C. e. nitens of New Guinea. This result suggests that further work may reveal that rump coloration is also not a taxonomically significant indicator among some C. esculenta species.